Tag Archives: 6G

AI in RAN – Evolution, Opportunities, and Risks

Posted on by
Reply

INTRO.

On September 10, at the Berlin Open RAN Working Week (BOWW) (a public event arranged by Deutsche Telekom AG’s T-Labs), I will give a talk about AI in Open RAN and RAN in general. The focus of the talk will be on how AI in RAN can boost the spectral efficiency. I have about 20 minutes, which is way too short to convey what is happening in this field at the moment. Why not write a small piece on the field as I see it at the moment? So, enjoy and feel free to comment or contact me directly for one-on-one discussions. If you are at the event, feel free to connect there as well.

LOOKING BACK.

The earliest use of machine learning and artificial intelligence in the Radio Access Network did not arrive suddenly with the recent wave of AI-RAN initiatives. Long before the term “AI-native RAN” (and even the term AI) became fashionable, vendors were experimenting with data-driven methods to optimize radio performance, automate operations, and manage complexity that traditional engineering rules could no longer handle well or at all. One of the first widely recognized examples came from Ericsson, which worked with SoftBank in Japan on advanced coordination features that would later be branded as Elastic RAN. By dynamically orchestrating users and cell sites, these early deployments delivered substantial throughput gains in dense environments such as Tokyo Station (with more than half a million passengers daily). Although they were not presented as “AI solutions,” they relied on principles of adaptive optimization that anticipated later machine learning–based control loops.

Nokia, and previously Nokia-Siemens Networks, pursued a similar direction through Self-Organizing Networks. SON functions, such as neighbor list management, handover optimization, and load balancing, increasingly incorporate statistical learning and pattern recognition techniques. These capabilities were rolled out across 3G and 4G networks during the 2010s and can be seen as some of the earliest mainstream applications of machine learning inside the RAN. Samsung, Huawei, and ZTE also invested in intelligent automation at this stage, often describing their approaches in terms of network analytics and energy efficiency rather than artificial intelligence, but drawing on many of the same methods. Around the same time, startups began pushing the frontier further: Uhana, founded in 2016 (acquired by VMware in 2019), pioneered the use of deep learning for real-time network optimization and user-experience prediction, going beyond rule-based SON to deliver predictive, closed-loop control. Building on that trajectory, today’s Opanga represents a (much) more advanced, AI-native and vendor-agnostic RAN platform, addressing long-standing industry challenges such as congestion management, energy efficiency, and intelligent spectrum activation at scale. In my opinion, both Uhana and Opanga can be seen as early exemplars of the types of applications that later inspired the formalization of rApps and xApps in the O-RAN framework.

What began as incremental enhancements in SON and coordination functions gradually evolved into more explicit uses of AI. Ericsson extended its portfolio with machine-learning-based downlink link adaptation and parameter optimization; Nokia launched programs to embed AI into both planning and live operations; and other vendors followed suit. By the early 2020s, the industry had begun to coalesce around the idea of an AI-RAN, where RAN functions and AI workloads are tightly interwoven. This vision took concrete form in 2024 with the launch of the AI-RAN Alliance, led by NVIDIA and comprising Ericsson, Nokia, Samsung, SoftBank, T-Mobile, and other partners.

The trajectory from SON and early adaptive coordination toward today’s GPU-accelerated AI-RAN systems underscores that artificial intelligence in the RAN has been less a revolution than an evolution. The seeds were sown in the earliest machine-learning-driven automation of 3G and 4G networks, and they have grown into the integrated AI-native architectures now being tested for 5G Advanced and beyond.

Article content
Figure: Evolution of Open RAN architectures — from early X-RAN disaggregation (2016–2018) to O-RAN standardization (2018–2020), and today’s dual paths of full disaggregated O-RAN and vRAN with O-RAN interfaces.

AI IN OPEN RAN – THE EARLIER DAYS.

Open RAN as a movement has its roots in the xRAN Forum (founded in 2016) and the O-RAN Alliance (created in early 2018 when xRAN merged with C-RAN Alliance). While the architectural thinking and evolution around what has today become the O-RAN Architecture (with its 2 major options) is interesting and very briefly summarized in the above figure. The late 2010s were a time when architectural choices were made in a climate of enormous enthusiasm for cloud-native design and edge cloud computing. At that time, “disaggregation for openness” was considered an essential condition for competition, innovation, and efficiency. I also believe that when xRAN was initiated around 2016, the leading academic and industrial players came predominantly from Germany, South Korea, and Japan. Each of these R&D cultures has a deep tradition of best-in-breed engineering, that is, the idea that the most specialized team or vendor should optimize every single subsystem, and that overall performance emerges from integrating these world-class components. Looking back today, with the benefit of hindsight, one can see how this cultural disposition amplified the push for the maximum disaggregation paradigm, even where integration and operational realities would later prove more challenging. It also explains why early O-RAN documents are so ambitious in scope, embedding intelligence into every layer and opening almost every possible interface imaginable. What appeared to be a purely technical roadmap was, in my opinion, also heavily shaped by the R&D traditions and innovation philosophies of the national groups leading the effort.

However, although this is a super interesting topic (i.e., how culture and background influence innovation, architectural ideas, and choices), it is not the focus of this paper. AI in RAN is the focus. From its very first architectural documents, O-RAN included the idea that AI and ML would be central to automating and optimizing the RAN.

The key moment was 2018, when the O-RAN Alliance released its initial O-RAN architecture white paper (“O-RAN: Towards an Open and Smart RAN”). That document explicitly introduced the concept of the Non-Real-Time (NRT) RIC (rApps) and the Real-Time (RT) RIC (xApps) as platforms designed to host AI/ML-based applications. The NRT RIC was envisioned to run in the operator’s cloud, providing policy guidance, training, and coordination of AI models at timescales well above a second. In contrast, the RT RIC (i.e., the official name is RT RIC, which is unfortunate for abbreviations among the two RICs) would host faster-acting control applications within the 10-ms to 1-s regime. These were framed not just as generic automation nodes but explicitly as AI/ML hosting environments. The idea of a dual RIC structure, breaking up the architecture in layers of relevant timescales, was not conceived in a vacuum. It is, in many ways, an explicit continuation of the ideas introduced in the 3GPP LTE Self-Organizing Network (SON) specifications, where optimization functions were divided between centralized, long-horizon processes running in the network management system and distributed, faster-acting functions embedded at the eNodeB. In the LTE context, the offline or centralized SON dealt with tasks such as PCI assignment, ANR management, and energy saving strategies at timescales of minutes to days. At the same time, the online or distributed SON reacted locally to interference, handover failures, or outages at timescales of hundreds of milliseconds to a few seconds. O-RAN borrowed this logic but codified it in a much more rigid fashion: the Non-RT RIC inherited the role of centralized SON, and the RT RIC inherited the role of distributed SON, with the addition of standardized interfaces and an explicit role as AI application platforms.

Figure: Comparison between the SON functions defined by 3GPP for LTE (right) and the O-RAN RIC architecture (left). The LTE model divides SON into centralized offline (C-SON, in OSS/NMS, working on minutes and beyond) and distributed online (D-SON, at the edge, operating at 100 ms to seconds) functions. In contrast, O-RAN formalized this split into the Non-RT RIC (≥1 s) and Near-RT RIC (10 ms–1 s), embedded within the SMO hierarchy. The figure highlights how O-RAN codified and extended SON’s functional separation into distinct AI/ML application platforms.

The choice to formalize this split also had political dimensions. Vendors were reluctant to expose their most latency-critical baseband algorithms to external control, and the introduction of an RT RIC created a sandbox where third-party innovation could be encouraged without undermining vendor control of the physical layer. At the same time, operators sought assurances that policy, assurance, and compliance would not be bypassed by low-latency applications; therefore, the Non-RT RIC was positioned as a control tower layer situated safely above the millisecond domain. In this sense, the breakup of the time domain was as much a governance and trust compromise as a purely technical necessity. By drawing a clear line between “safe and slow” and “fast but bounded,” O-RAN created a model that felt familiar to operators accustomed to OSS hierarchies, while signaling to regulators and ecosystem players that AI could be introduced in a controlled and explainable manner.

Article content
Figure: Functional and temporal layering of the O-RAN architecture — showing the SMO with embedded NRT-RIC for long-horizon and slow control loops, the RT-RIC for fast loops, and the CU, DU, and RU for real-time through instant reflex actions, interconnected via standardized O-, A-, E-, F-, and eCPRI interfaces.

The figure above shows the O-RAN reference architecture with functional layers and interfaces. The Service Management and Orchestration (SMO) framework hosts the Non-Real-Time RIC (NRT-RIC), which operates on long-horizon loops (greater than 1 second) and is connected via the O1 interface to network elements and via O2 to cloud infrastructure (e.g., NFVI and MANO). Policies, enrichment information, and trained AI/ML models are delivered from the NRT-RIC to the Real-Time RIC (RT-RIC) over the A1 interface. The RT-RIC executes closed-loop control in the 10-ms to 1-s domain through xApps, interfacing with the CU/DU over E2. The 3GPP F1 split separates the CU and DU, while the DU connects to the RU through the open fronthaul (eCPRI/7-2x split). The RU drives active antenna systems (AAS) over largely proprietary interfaces (AISG for RET, vendor-specific for massive MIMO). The vertical time-scale axis highlights the progression from long-horizon orchestration at the SMO down to instant reflex functions in the RU/AAS domain. Both RU and DU operate on a transmission time interval (TTI) between 1 ms and 625 microseconds.

The O-RAN vision for AI and ML is built directly into its architecture from the very first white paper in 2018. The alliance described two guiding themes: openness and intelligence. Openness was about enabling multi-vendor, cloud-native deployments with open interfaces, which was supposed to provide for much more economical RAN solutions, while intelligence was about embedding machine learning and artificial intelligence into every layer of the RAN to deal with growing complexity (i.e., some of it self-inflicted by architecture and system design).

The architectural realization of this vision is the hierarchical RAN Intelligent Controller (RIC), which separates the control into different time domains and couples each to appropriate AI/ML functions:

  • Service Management and Orchestration (SMO, timescale > 1 second) – The Control Tower: The SMO provides the overarching management and orchestration framework for the RAN. Its functions extend beyond the Non-RT RIC, encompassing lifecycle management, configuration, assurance, and resource orchestration across both network functions and the underlying cloud infrastructure. Through the O1 interface (see above figure), the SMO collects performance data, alarms, and configuration information from the CU, DU, and RU, enabling comprehensive FCAPS (Fault, Configuration, Accounting, Performance, Security) management. Through the O2 interface (see above), it orchestrates cloud resources (compute, storage, accelerators) required to host virtualized RAN functions and AI/ML workloads. In addition, the SMO hosts the Non-RT RIC, meaning it not only provides operational oversight but also integrates AI/ML governance, ensuring that trained models and policy guidance align with operator intent and regulatory requirements.
  • Non-Real-Time RIC (NRT RIC, timescale > 1 second) – The Policy Brain: Directly beneath, embedded in the SMO, lies the NRT-RIC, described here as the “policy brain.” This is where policy management, analytics, and AI/ML model training take place. The non-RT RIC collects large volumes of data from the network (spatial-temporal traffic patterns, mobility traces, QoS (Quality of Service) statistics, massive MIMO settings, etc.) and uses them for offline training and long-term optimization. Trained models and optimization policies are then passed down to the RT RIC via the A1 interface (see above). A central functionality of the NRT-RIC is the hosting of rApps (e.g., Python or Java code), which implement policy-driven use cases such as energy savings, traffic steering, and mobility optimization. These applications leverage the broader analytic scope and longer timescales of the NRT-RIC to shape intent and guide the near-real-time actions of the RT-RIC. The NRT-RIC is traditionally viewed as an embedded entity within the SMO (although in theory, it could be a standalone entity)..
  • Real-Time RIC (RT RIC, 10 ms – 1 second timescale) – The Decision Engine: This is where AI-driven control is executed in closed loops. The real-time RT-RIC hosts xApps (e.g., Go or C++ code) that run inference on trained models and perform tasks such as load balancing, interference management, mobility prediction, QoS management, slicing, and per-user (UE) scheduling policies. It maintains a Radio Network Information Base (R-NIB) fed via the E2 interface (see above) from the DU/CU, and uses this data to make fast control decisions in near real-time.
  • Centralized Unit (CU): Below the RT-RIC sits the Centralized Unit, which takes on the role of the “shaper” in the O-RAN architecture. The CU is responsible for higher-layer protocol processing, including PDCP (Packet Data Convergence Protocol) and SDAP (Service Data Adaptation Protocol), and is therefore the natural point in the stack where packet shaping and QoS enforcement occur. At this level, AI-driven policies provided by the RT-RIC can directly influence how data streams are prioritized and treated, ensuring that application- or slice-specific requirements for latency, throughput, and reliability are respected. By interfacing with the RT-RIC over the E2 interface, the CU can dynamically adapt QoS profiles and flow control rules based on real-time network conditions, balancing efficiency with service differentiation. In this way, the CU acts as the bridge between AI-guided orchestration and the deterministic scheduling that occurs deeper in the DU/RU layers. The CU operates on a real-time but not ultra-tight timescale, typically in the range of tens of milliseconds up to around one second (similar to the RT-RIC), depending on the function.
  • DU/RU layer (sub-1 ms down to hundreds of microseconds) – The Executor & Muscles: The Distributed Unit (DU), located below the CU, is referred to as the “executor.” It handles scheduling and precoding at near-instant timescales, measured in sub-millisecond intervals. Here, AI functions take the form of compute agents that apply pre-trained or lightweight models to optimize resource block allocation and reduce latency. At the bottom, the Radio Unit (RU) represents the “muscles” of the system. Its reflex actions happen at the fastest time scales, down to hundreds of microseconds. While it executes deterministic signal processing, beamforming, and precoding, it also feeds measurements upward to fuel AI learning higher in the chain. Here reside the tightest loops, on a Transmission Time Interval (TTI) time scale (i.e., 1ms – 625 µs), such as baseband PHY processing, HARQ feedback, symbol scheduling, and beamforming weights. These functions require deterministic latencies and cannot rely on higher-layer AI/ML loops. Instead, the DU/RU executes control at the L1/L2 level, while still feeding measurement data upward for AI/ML training and adaptation.
Article content
Figure: AI’s hierarchical chain of command in O-RAN — from the SMO as the control tower and NRT-RIC as the policy brain, through the RT-RIC as the decision engine and CU as shaper, down to DU as executor and RU as muscles. Each layer aligns with guiding timescales, agentic AI roles, and contributions to spectral efficiency, balancing perceived SE gains, overhead reductions, and SINR improvements.

The figure above portrays the Open RAN as a “chain of command” where intelligence flows across time scales, from long-horizon orchestration in the cloud down to sub-millisecond reflexes in the radio hardware. To make it more tangible, I have annotated the example of spectral efficiency optimization use case on the right side of the figure. The cascading structure, shown above, highlights how AI and ML roles evolve across the architecture. For instance, the SMO and NRT-RIC increase perceived spectral efficiency through strategic optimization, while the RT-RIC reduces inefficiencies by orchestrating fast loops. Additionally, the DU/RU contribute directly to signal quality improvements, such as SINR gains. The figure thus illustrates Open RAN not as a flat architecture, but as a hierarchy of brains, decisions, and muscles, each with its own guiding time scale and AI function. Taken together, the vision is that AI/ML operates across all time domains, with the non-RT RIC providing strategic intelligence and model training, the RT RIC performing agile, policy-driven adaptation, and the DU/RU executing deterministic microsecond-level tasks, while exposing data to feed higher-layer intelligence. With open interfaces (A1, E2, open fronthaul), this layered AI approach allows multi-vendor participation, third-party innovation, and closed-loop automation across the RAN.

From 2019 onward, O-RAN working groups such as WG2 (Non-RT RIC & A1 interface) and WG3 (RT RIC & E2 interface) began publishing technical specifications that defined how AI/ML models could be trained, distributed, and executed across the RIC layers. By 2020–2021, proof-of-concepts and plugfests showcased concrete AI/ML use cases, such as energy savings, traffic steering, and anomaly detection, running as xApps (residing in RT-RIC) and rApps (residing in NRT-RIC). Following the first O-RAN specifications and proof-of-concepts, it becomes helpful to visualize how the different architectural layers relate to AI and ML. You will find a lot of the standardization documents in the reference list at the end of the document.

rAPPS AND xAPPS – AN ILLUSTRATION.

In the Open RAN architecture, the system’s intelligence is derived from the applications that run on top of the RIC platforms. The rApps exist in the Non-Real-Time RIC and xApps in the Real-Time RIC. While the RICs provide the structural framework and interfaces, it is the apps that carry the logic, algorithms, and decision-making capacity that ultimately shape network behavior. rApps operate at longer timescales, often drawing on large datasets and statistical analysis to identify trends, learn patterns, and refine policies. They are well-suited to classical machine learning processes such as regression, clustering, and reinforcement learning, where training cycles and retraining benefit from aggregated telemetry and contextual information. In practice, rApps are commonly developed in high-level languages such as Python or Java, leveraging established AI/ML libraries and data processing pipelines. In contrast, xApps must execute decisions in near-real time, directly influencing scheduling, beamforming, interference management, and resource allocation. Here, the role of AI and ML is to translate abstract policy into fast, context-sensitive actions, with an increasing reliance on intelligent control strategies, adaptive optimization, and eventually even agent-like autonomy (more on that later in this article). To meet these latency and efficiency requirements, xApps are typically implemented in performance-oriented languages like C++ or Go. However, Python is often used in prototyping stages before critical components are optimized. Together, rApps and xApps represent the realization of intelligence in Open RAN. One set grounded in long-horizon learning and policy shaping (i.e., Non-RT RIC and rApps), the other in short-horizon execution and reflexive adaptation (RT-RIC and xApps). Their interplay is not only central to energy efficiency, interference management, and spectral optimization but also points toward a future where classical ML techniques merge with more advanced AI-driven orchestration to deliver networks that are both adaptive and self-optimizing. Let us have a quick look at examples that illustrate how these applications work in the overall O-RAN architectural stack.

Figure: Energy efficiency loop in Open RAN, showing how long-horizon rApps set policies in the NRT-RIC, xApps in the RT-RIC execute them, and DU/RU translate these into scheduler and hardware actions with continuous telemetry feedback.

One way to understand the rApp–xApp interaction is to follow a simple energy efficiency use case, shown in the figure below. At the top, an energy rApp in the Non-RT RIC learns long-term traffic cycles and defines policies such as ‘allow cell muting below 10% load.’ These policies are then passed to the RT-RIC, where an xApp monitors traffic every second and decides when to shut down carriers or reduce power. The DU translates these decisions into scheduling and resource allocations, while the RU executes the physical actions such as switching off RF chains, entering sleep modes, or muting antenna elements. The figure above illustrates how policy flows downward while telemetry and KPIs flow back up, forming a continuous energy optimization loop. Another similarly layered logic applies to interference coordination, as shown in the figure below. Here, an interference rApp in the Non-RT RIC analyzes long-term patterns of inter-cell interference and sets coordination policies — for example, defining thresholds for ICIC, CoMP, or power capping at the cell edge. The RT-RIC executes these policies through xApps that estimate SINR in real time, apply muting patterns, adjust transmit power, and coordinate beam directions across neighboring cells. The DU handles PRB scheduling and resource allocation, while the RU enacts physical layer actions, such as adjusting beam weights or muting carriers. This second loop shows how rApps and xApps complement each other when interference is the dominant concern.

Figure: Interference coordination loop in Open RAN, where rApps define long-term coordination policies and xApps execute real-time actions on PRBs, power, and beams through DU/RU with continuous telemetry feedback.

Yet these loops do not always reinforce each other. If left uncoordinated, they can collide. An energy rApp may push the system toward contraction, reducing Tx power, muting carriers, and blanking PRBs. In contrast, an interference xApp simultaneously pushes for expansion, raising Tx power, activating carriers, and dynamically allocating PRBs. Both act on the same levers inside the CU/DU/RU, but in opposite directions. The result can be oscillatory behaviour, with power and scheduling thrashing back and forth, degrading QoS, and wasting energy. The figure below illustrates this risk and underscores why conflict management and intent arbitration are critical for a stable Open RAN.

Figure: Example of conflict between an energy-saving rApp and an interference-mitigation xApp, where opposing control intents on the same CU/DU/RU parameters can cause oscillatory behaviour.

Beyond the foundational description of how rApps and xApps operate, it is equally important to address the conflicts and issues that can arise when multiple applications are deployed simultaneously in the Non-RT and RT-RICs. Because each app is designed with a specific optimization objective in mind, it is almost inevitable that two or more apps will occasionally attempt to act on the same parameters in contradictory ways. While the energy efficiency versus interference management example is already well understood, there are broader categories of conflict that extend across both timescales.

Conflicts between rApps occur when long-term policy objectives are not aligned. For instance, a spectral efficiency rApp may continuously push the network toward maximizing bits per Hertz by advocating for higher transmit power, more active carriers, or denser pilot signaling. At the same time, an energy-saving rApp may be trying to mute those very carriers, reduce pilot density, and cap transmit power to conserve energy. Both policies can be valid in isolation, but when issued without coordination, they create conflicting intents that leave the RT-RIC and lower layers struggling to reconcile them. Even worse, the oscillatory behavior that results can propagate into the DU and RU, creating instability at the level of scheduling and RF execution. The xApps, too, can easily find themselves in conflict when they react to short-term KPI fluctuations with divergent strategies. An interference management xApp might impose aggressive PRB blanking patterns or reduce power at the cell edge. At the same time, a mobility optimization xApp might simultaneously widen cell range expansion parameters to offload traffic. The first action is designed to protect edge users, while the second may flood them with more load, undoing the intended benefit. Similarly, an xApp pushing for higher spectral efficiency may keep activating carriers and pushing toward higher modulation and coding schemes, while another xApp dedicated to energy conservation is attempting to put those carriers to sleep. The result is rapid toggling of resource states, which wastes signaling overhead and disrupts user experience.

The O-RAN Alliance has recognized these risks and proposed mechanisms to address them. Architecturally, conflict management is designed to reside in the RT-RIC, where a Conflict Mitigation and Arbitration framework evaluates competing intents from different xApps before they reach the CU/DU. Policies from the Non-RT RIC can also be tagged with priorities or guardrails, which the RT-RIC uses to arbitrate real-time conflicts. In practice, this means that when two xApps attempt to control the same parameter, the RT-RIC applies priority rules, resolves contradictions, or, in some cases, rejects conflicting commands entirely. On the rApp side, conflict resolution is handled at a higher abstraction level by the Non-RT RIC, which can consolidate or harmonize policies before they are passed down through the A1 interface.

The layered conflict mitigation approach in O-RAN provides mechanisms to arbitrate competing intents between apps. It can reduce the risk of oscillatory behavior, but it cannot guarantee stability completely. Since rApps and xApps may originate from different sources and vary in design quality, careful testing, certification, and continuous monitoring will remain essential to ensure that application diversity does not undermine network coherence. Equally important are policies that impose guardbands, buffers, and safety margins in how parameters can be tuned, which serve as a hedge against instabilities when apps are misaligned, whether the conflict arises between rApps, between xApps, or across the rApp–xApp boundary. These guardbands provide the architectural equivalent of shock absorbers, limiting the amplitude of conflicting actions and ensuring that, even if multiple apps pull in different directions, the network avoids catastrophic oscillations.

Last but not least, the risks may increase as rApps and xApps evolve beyond narrowly scoped optimizers into more agentic forms. An agentic app does not merely execute a set of policies or inference models. It can plan, explore alternatives, and adapt its strategies with a degree of autonomy (and agency). While this is likely to unlock powerful new capabilities, it also expands the possibility of emergent and unforeseen interactions. Two agentic apps, even if aligned at deployment, may drift toward conflicting behaviors as they continuously learn and adapt in real time. Without strict guardrails and robust conflict resolution, such autonomy could magnify instabilities rather than contain them, leading to system behavior that is difficult to predict or control. In this sense, the transition from classical rApps and xApps to agentic forms is not only an opportunity but also a new frontier of risk that must be carefully managed within the O-RAN architecture.

IS AI IN RAN ALL ABOUT “ChatGPT”?

I want to emphasize that when I address AI in the RAN, I generally do not refer to generative language models, such as ChatGPT, or other large-scale conversational systems built upon a human language context. Those technologies are based on Large Language Models (LLMs), which belong to the family of deep learning architectures built on transformer networks. A transformer network is a type of neural network architecture built around the attention mechanism, which allows the model to weigh the importance of different parts of an input sequence simultaneously rather than processing it step by step. They are typically trained on enormous human-based text datasets, utilizing billions of parameters, which requires immense computational resources and lengthy training cycles. Their most visible purpose today is to generate and interpret human language, operating effectively at the scale of seconds or longer in user interactions. In the context of network operations, I suspect that GPT-like LLMS will have a mission in the frontend where humans will need to interact with the communications network using human language. That said, the notion of “generative AI” is not inherently limited to natural language. The same underlying transformer-based methods can be adapted to other modalities (information sources), including machine-oriented languages or even telemetry sequences. For example, a generative model trained on RAN logs, KPIs, and signaling traces could be used to create synthetic telemetry or predict unusual event patterns. In this sense, generative AI could provide value to the RAN domain by augmenting datasets, compressing semantic information, or even assisting in anomaly detection. The caveat, however, is that these benefits still rely on heavy models with large memory footprints and significant inference latency. While they may serve well in the Non-RT RIC or SMO domain, where time scales are relaxed and compute resources are more abundant, they are unlikely to be terribly practical for the RT RIC or the DU/RU, where deterministic deadlines in the millisecond or microsecond range must be met.

By contrast, the application of AI/ML in the RAN is fundamentally about real-time signal processing, optimization, and control. RAN intelligence focuses on tasks such as load balancing, interference mitigation, mobility prediction, traffic steering, energy optimization, and resource scheduling. These are not problems of natural human language understanding but of strict scheduling and radio optimization. The time scales at which these functions operate are orders of magnitude shorter than those typical of generative AI. From long-horizon analytics in the Non-RT RIC (greater than one second) to near-real-time inference in the RT-RIC (i.e., 10 ms–1 s), and finally to deterministic microsecond loops in the DU/RU. This stark difference in time scales and problem domains explains why it appears unlikely that the RAN can be controlled end-to-end by “ChatGPT-like” AI. LLMs, whether trained on human language or telemetry sequences, are (today at least) too computationally heavy, too slow in inference, and are optimized for open-ended reasoning rather than deterministic control. Instead, the RAN requires a mix of lightweight supervised and reinforcement learning models, online inference engines, and, in some cases, ultra-compact TinyML implementations that can run directly in hardware-constrained environments.

In general, AI in the RAN is about embedding intelligence into control loops at the right time scale and with the right efficiency. Generative AI may have a role in enriching data and informing higher-level orchestration. It is difficult to see how it can efficiently replace the tailored, lightweight models that drive the RAN’s real-time and near-real-time control.

As O-RAN (and RAN in general) evolves from a vision of open interfaces and modular disaggregation into a true intelligence-driven network, one of the clearest frontiers is the use of Large Language Models (LLMs) at the top of the stack (i.e., frontend/human-facing). The SMO, with its embedded Non-RT RIC, already serves as the strategic brain of the architecture, responsible for lifecycle management, long-horizon policy, and the training of AI/ML models. This is also the one domain where time scales are relaxed, measured in seconds or longer, and where sufficient compute resources exist to host heavier models. In this environment, LLMs can be utilized in two key ways. First, they can serve as intent interpreters for intent-driven network operations, bridging the gap between operator directives and machine-executable policies. Instead of crafting detailed rules or static configuration scripts, operators could express high-level goals, such as prioritizing emergency service traffic in a given region or minimizing energy consumption during off-peak hours. An LLM, tuned with telecom-specific knowledge, can translate those intents into precise policy actions distributed through the A1 interface to the RT RIC. Second, LLMs can act as semantic compressors, consuming the vast streams of logs, KPIs, and alarms that flow upward through O1, and distilling them into structured insights or natural language summaries that humans can easily grasp. This reduces cognitive load for operators while ensuring (at least we should hope so!) that the decision logic remains transparent, possibly explainable, and auditable. In both roles, LLMs do not replace the specialized ML models running lower in the architecture. Instead, they enhance the orchestration layer by embedding reasoning and language understanding where time and resources permit.

WHAT AI & ML ARE LIKELY TO WORK IN RAN?

This piece assumes a working familiarity with core machine-learning concepts, models, training and evaluation processes, and the main families you will encounter in practice. If you want a compact, authoritative refresher, most of what I reference is covered, clearly and rigorously, in Goodfellow, Bengio, and Courville’s Deep Learning (Adaptive Computation and Machine Learning series, MIT Press). For hands-on practice, many excellent Coursera courses walk through these ideas with code, labs, and real datasets. They are a fast way to build the intuition you will need for the examples discussed in this section. Feel free to browse through my certification list, which includes over 60 certifications, with the earliest ML and AI courses dating back to 2015 (should have been updated by now), and possibly find some inspiration.

Throughout the article, I use “AI” and “ML” interchangeably for readability, but formally, they should be regarded as distinct. Artificial Intelligence (AI) is the broader field concerned with building systems that perceive their environment, reason about it, and act to achieve goals, encompassing planning, search, knowledge representation, learning, and decision-making. Machine Learning (ML) is a subset of AI that focuses specifically on data-driven methods that learn patterns or policies from examples, improving performance on a task through experience rather than explicit, hand-crafted rules, which is where the most interesting aspects occur.

Article content
Figure: Mapping of AI roles, data flows, and model families across the O-RAN stack — from SMO and NRT-RIC handling long-horizon policy, orchestration, and training, to RT-RIC managing fast-loop inference and optimization, down to CU and DU/RU executing near-real-time and hardware-domain actions with lightweight, embedded AI models.

Artificial intelligence in the O-RAN stack exhibits distinct characteristics depending on its deployment location. Still, it is helpful to see it as one continuous flow from intent at the very top to deterministic execution at the very bottom. So, let’s go with the flow.

At the level of the Service Management and Orchestration, AI acts as the control tower for the entire system. This is where business or human intent must be translated into structured goals, and where guardrails, audit mechanisms, and reversibility are established to ensure compliance with regulatory oversight. Statistical models and rules remain essential at this layer because they provide the necessary constraint checking and explainability for governance. Yet the role of large language models is increasing rapidly, as they provide a bridge from human language into structured policies, intent templates, and root-cause narratives. Generative approaches are also beginning to play a role by producing synthetic extreme events to stress-test policies before they are deployed. While synthetic data for rare events offers a powerful tool for training and stress-testing AI systems, it may carry significant statistical risks. Generative models can fail to represent the very distributions they aim to capture, bias inference, or even introduce entirely artificial patterns into the data. Their use therefore requires careful anchoring in extremes-aware statistical methods, rigorous validation against real-world holdout data, and safeguards against recursive contamination. When these conditions are met, synthetic data can meaningfully expand the space of scenarios available for training and testing. Without the appropriate control mechanisms, decisions or policies based on synthetic data risk becoming a source of misplaced confidence rather than resilience. With all that considered, the SMO should be the steward of safety and interpretability, ensuring that only validated and reversible actions flow down into the operational fabric. If agentic AI is introduced here, it could reshape how intent is operationalized. Instead of merely validating human inputs, agentic systems might proactively (autonomeously) propose actions, refine intents into strategies, or initiate self-healing workflows on their own. While this promises greater autonomy and resilience, it also raises new challenges for oversight, since the SMO would become not just a filter but a creative actor in its own right.

At the top level, rApps (which reside in the NRT-RIC) are indirectly shaped by SMO policies, as they inherit intent, guardrails, and reversibility constraints. For example, when the SMO utilizes LLMs to translate business goals into structured intents, it essentially sets the design space within which rApps can train or re-optimize their models. The SMO also provides observability hooks, allowing rApp outputs to be audited before being pushed downstream.

The Non-Real-Time RIC can be understood as the long-horizon brain of the RAN. Its function is to train, retrain, and refine models, conduct long-term analysis, and transform historical and simulated experience into reusable policies. Reinforcement learning in its many flavors is the cornerstone here, particularly offline or constrained forms that can safely explore large data archives or digital twin scenarios. Autoencoders, clustering, and other representation learning methods uncover hidden structures in traffic and mobility, while supervised deep networks and boosted trees provide accurate forecasting of demand and performance. Generative simulators extend the scope by fabricating rare but instructive scenarios, allowing policies to be trained for resilience against the unexpected. Increasingly, language-based systems are also being applied to policy generation, bridging between strategic descriptions and machine-enforceable templates. The NRT-RIC strengthens AI’s applicability by moving risk away from live networks, producing validated artifacts that can later be executed at speed. If an agentic paradigm is introduced here, it would mean that the NRT-RIC is not merely a training ground but an active planner, continuously setting objectives for the rest of the system and negotiating trade-offs between coverage, energy, and user experience. This shift would make the Non-RT RIC a more autonomous planning organ, but it would also demand stronger mechanisms for bounding and auditing its explorations.

Here, at the NTR-RIC, rApps that are native to this RIC level are the central vehicle for model training, policy generation, and scenario exploration. They consume SMO intent and turn it into reusable policies or models for the RT-RIC. For example, a mobility rApp could use clustering and reinforcement learning to generate policies for user handover optimization, which the RT-RIC then executes in near real time. Another rApp might simulate mMIMO pairing scenarios offline, distill them into simplified lookup tables or quantized policies, and hand these artifacts down for execution at the DU/RU. Thus, rApps act as the policy factories. Their outputs cascade into xApps, at the RT-RIC, CU parameter sets, and lightweight silicon-bound models deeper down.

The Real-Time RIC is where planning gives way to fast, local action. At timescales between ten milliseconds and one second, the RT-RIC is tasked with run-time inference, traffic steering, slicing enforcement, and short-term interference management. Because the latency budget is tight, the model families that thrive here are compact and efficient. Shallow neural networks, recurrent models, and temporal CNN-RNN hybrids are all appropriate for predicting near-future load and translating context into rapid actions. Decision trees and ensemble methods remain attractive because of their predictable execution and interpretability. Online reinforcement learning, in which an agent interacts with its environment in real-time and updates its policy based on rewards or penalties, together with contextual bandits, a simplified variant that optimizes single-step decisions from observed contexts, both enable adaptation in small, incremental steps while minimizing the risk of destabilization. In specific contexts, lightweight graph neural networks (GNNs), which are streamlined versions of GNNs designed to model relationships between entities while keeping computational costs low, can capture the topological relationships between neighboring cells. In the RT-RIC, models must balance accuracy with predictable execution under tight timescales. Shallow neural networks (simple feedforward models capturing non-linear patterns), recurrent models (RNNs that retain memory of past inputs), and hybrid convolutional neural network–recurrent neural network (CNN–RNN) models (combining spatial feature extraction with temporal sequencing) are well-suited for processing fast-evolving time series, such as traffic load or interference, delivering near-future predictions with low latency. Decision trees (rule-based classifiers that split data hierarchically) and ensemble methods (collections of weak learners, such as random forests or boosting) add value through their lightweight, deterministic behavior and interpretability, making them reliable for regulatory oversight and stable actuation. Online reinforcement learning (RL) and contextual bandits further allow the system to adapt incrementally to changing conditions without risking destabilization. In more complex contexts, lightweight GNNs capture the topological structure between neighboring cells, supporting coordination in handovers or interference management while remaining efficient enough for real-time use. The RT-RIC thus embodies the point where AI policies become immediate operational decisions, measurable in KPIs within seconds. When viewed through the lens of agency, this layer becomes even more dynamic. An agentic RT-RIC could weigh competing goals, prioritize among multiple applications, and negotiate real-time conflicts without waiting for external intervention. Such an agency might significantly improve efficiency and responsiveness but would also blur the boundary between optimization and autonomous control, requiring new arbitration frameworks and assurance layers.

At this level, xApps, native to the RT-RIC, execute policies derived from rApps and adapt them to live network telemetry. An xApp for traffic steering might combine a policy from the Non-RT RIC with local contextual bandits to adjust routing in the moment. Another xApp could, for example, use lightweight GNNs to coordinate interference management across adjacent cells, directly influencing DU scheduling and RU beamforming. This makes xApps the translators of long-term rApp insights into second-by-second action, bridging the predictive foresight of rApps with the deterministic constraints of the DU/RU.

The Centralized Unit occupies an intermediate position between near-real-time responsiveness and higher-layer mobility and bearer management. Here, the most useful models are those that can both predict and pre-position resources before bottlenecks occur. Long Short-Term Memory networks (LSTMs, recurrent models designed to capture long-range dependencies), Gated Recurrent Units (GRUs, simplified RNNs with fewer parameters), and temporal Convolutional Neural Networks (CNNs, convolution-based models adapted for sequential data) are natural fits for forecasting user trajectories, mobility patterns, and session demand, thereby enabling proactive preparation of handovers and early allocation of network slices. Constrained reinforcement learning (RL, trial-and-error learning optimized under explicit safety or policy limits) methods play an important role at the bearer level, where they must carefully balance Quality of Service (QoS) guarantees against overall resource utilization, ensuring efficiency without violating service-level requirements. At the same time, rule-based optimizers remain well-suited for more deterministic processes, such as configuring Packet Data Convergence Protocol (PDCP) and Radio Link Control (RLC) parameters, where fixed logic can deliver predictable and stable outcomes in real-time. The CU strengthens applicability by anticipating issues before they materialize and by converting intent into per-flow adjustments. If agency is introduced at this layer, it might manifest as CU-level agents negotiating mobility anchors or bearer priorities directly, without relying entirely on upstream instructions. This could increase resilience in scenarios where connectivity to higher layers is impaired. Still, it also adds complexity, as the CU would need a framework for coordinating its autonomous decisions with the broader policy environment.

Both xApps and rApps can influence CU functions as they relate to bearer management and PDCP/RLC configuration. For example, a QoS balancing rApp might propose long-term thresholds for bearer prioritization. At the same time, a short-horizon xApp enforces these by pre-positioning slice allocations or adjusting bearer anchors in anticipation of predicted mobility. The CU thus becomes a convergence point, where rApp strategies and xApp tactics jointly shape mobility management and session stability before decisions cascade into DU scheduling.

At the very bottom of the stack, the Distributed Unit and Radio Unit function under the most stringent timing constraints, often in the realm of microseconds. Their role is to execute deterministic PHY and MAC functions, including HARQ, link adaptation, beamforming, and channel state processing. Only models that can be compiled into silicon, quantized, or otherwise guaranteed to run within strict latency budgets are viable in this layer of the Radio Access Network. Tiny Machine Learning (TinyML), Quantized Neural Networks (QNN), and lookup-table distilled models enable inference speeds compatible with microsecond-level scheduling constraints. As RU and DU components typically operate under strict latency and computational constraints, TinyML and low-bit QNNs are ideal for deploying functions such as beam selection, RF monitoring, anomaly detection, or lightweight PHY inference tasks. Deep-unfolded networks and physics-informed neural models are particularly valuable because they can replace traditional iterative solvers in equalization and channel estimation, achieving high accuracy while ensuring fixed execution times. In advanced antenna systems, neural digital predistortion and amplifier linearization enhance power efficiency and spectral containment. At the same time, sequence-based predictors can cut down channel state information (CSI) overhead and help stabilize multi-user multiple-input multiple-output (MU-MIMO) pairing. At this level, the integration of agentic AI must, in my opinion, be approached with caution. The DU and RU domains are all about execution rather than deliberation. Introducing agency here could compromise determinism. However, carefully bounded micro-agents that autonomously tune beams or adjust precoders within strict envelopes might prove valuable. The broader challenge is to reconcile the demand for predictability with the appeal of adaptive intelligence baked into hardware.

At this layer, most intelligence is “baked in” and must respect microsecond determinism timescales. Yet, rApps and xApps may still indirectly shape the DU/RU environment. The DU/RU do not run complex agentic loops themselves, but they inherit distilled intelligence from the higher layers. Micro-agents, if used, must be tightly bound. For example, an RU micro-agent may autonomously choose among two or three safe precoding matrices supplied by an xApp, but never generate them on its own.

Taking all the above together, the O-RAN stack can be seen as a continuum of intelligence, moving from the policy-heavy, interpretative functions at the SMO to the deterministic, silicon-bound execution at the RU. Agentic AI has the potential to change this continuum by shifting layers from passive executors to active participants. An agentic SMO might not only validate intents but generate them. An agentic Non-RT RIC might become an autonomous planner. An agentic RT-RIC could arbitrate between conflicting goals independently. And even the CU or DU might gain micro-agents that adjust parameters locally without instruction. This greater autonomy promises efficiency and adaptability but raises profound questions about accountability, oversight, and control. If the agency is allowed to propagate too deeply into the stack, the risk is that millions of daily inferences are taken without transparent justification or the possibility of reversal. This situation is unlikely to be considered regulatory acceptable and would be in direct violation of the European Artificial Intelligence Act, violating core provisions of the EU AI Act. The main risks are a lack of adequate human oversight (Article 14), inadequate record-keeping and traceability (Article 12), failures of transparency (Article 13), and the inability to provide meaningful explanations to affected users (Article 86). Together, these gaps would undermine the broader lifecycle obligations on risk management and accountability set out in Articles 8–17. To mitigate that, openness becomes indispensable: open policies, open data schemas, model lineage, and transparent observability hooks allow agency to be exercised without undermining trust. In this way, the RAN of the future may become not only intelligent but agentic, provided that its newfound autonomy is balanced by openness, auditability, and human authority at the points that matter most. However, I suspect that reaching that point may be a much bigger challenge than developing the AI Agentic framework and autonomous processes.

While the promise of AI in O-RAN is compelling, it is equally important to recognize where existing functions already perform so effectively that AI has little to add. At higher layers, such as the SMO and the Non-RT RIC, the complexity of orchestration, policy translation, and long-horizon planning naturally creates a demand for AI. These are domains where deterministic rules quickly become brittle, and where the adaptive and generative capacities of modern models unlock new value. Similarly, the RT-RIC benefits from lightweight ML approaches because traffic dynamics and interference conditions shift on timescales that rule-based heuristics often struggle to capture. As one descends closer to execution, however, the incremental value of AI begins to diminish. In the CU domain, many bearer management and PDCP/RLC functions can be enhanced by predictive models. Still, much of the optimization is already well supported by deterministic algorithms that operate within known bounds. The same is even more pronounced at the DU and RU levels. Here, fundamental PHY and MAC procedures such as HARQ timing, CRC checks, coding and decoding, and link-layer retransmissions are highly optimized, deterministic, and hardware-accelerated. These functions have been refined over decades of wireless research, and their performance approaches the physical and information-theoretical limits. For example, beamforming and precoding illustrate this well. Linear algebraic methods such as zero-forcing and MMSE are deeply entrenched, efficient, and predictable. AI and ML can sometimes enhance them at the margins by improving CSI compression, reducing feedback overhead, or stabilizing non-stationary channels. Yet it is unlikely to displace the core mathematical solvers that already deliver excellent performance. Link adaptation is similar. While machine learning may offer marginal gains in dynamic or noisy conditions, conventional SINR-based thresholding remains highly effective and, crucially, deterministic. It is worth remembering that simply and arbitrarily applying AI or ML functionality to an architectural element does not necessarily mean it will make a difference or even turn out to be beneficial.

This distinction becomes especially relevant when considering the implications of agentic AI. In my opinion, agency is most useful at the top of the stack, where strategy, trade-offs, and ambiguity dominate. In the SMO or Non-RT RIC, agentic systems can propose strategies, negotiate policies, or adapt scenarios in ways that humans or static systems could never match. At the RT-RIC, a carefully bound agency may improve arbitration among competing applications. But deeper in the stack, particularly at the DU and RU, the agency adds little value and risks undermining determinism. At microsecond timescales, where physics rules and deadlines are absolute, autonomy may be less of an advantage and more of a liability. The most practical role of AI here is supplementary, enabling anomaly detection, parameter fine-tuning, or assisting advanced antenna systems in ways that respect strict timing constraints. This balance of promise and limitation underscores a central point. AI is not a panacea for O-RAN, nor should it be applied indiscriminately.

Article content
Figure: Comparative view of how AI transforms RAN operations — contrasting classical vendor-proprietary SON approaches, Opanga’s vendor-agnostic RAIN platform, and O-RAN implementations using xApps and rApps for energy efficiency, spectral optimization, congestion control, anomaly detection, QoE, interference management, coverage, and security.

The Table above highlights how RAN intelligence has evolved from classical vendor-specific SON functions toward open O-RAN frameworks and Opanga’s RAIN platform. While Classical RAN relied heavily on embedded algorithms and static rules, O-RAN introduces rApps and xApps to distribute intelligence across near-real-time and non-real-time control loops. Opanga’s RAIN, however, stands out as a truly AI-native and vendor-agnostic platform that is already commercially deployed at scale today. By tackling congestion, energy reduction, and intelligent spectrum on/off management without reliance on DPI (which is, anyway, a losing strategy as QUIC becomes increasingly used) or proprietary stacks, RAIN directly addresses some of the most pressing efficiency and sustainability challenges in today’s networks. It also appears straightforward for Opanga to adapt its AI engines into rApps or xApps should the Open RAN market scale substantially in the future, reinforcing its potential as one of the strongest and most practical AI platforms in the RAN domain today.

A NATIVE-AI RAN TEASER.

Native-AI in the RAN context means that artificial intelligence is not just an add-on to existing processes, but is embedded directly into the system’s architecture, protocols, and control loops. Instead of having xApps and rApps bolted on top of traditional deterministic scheduling and optimization functions, a native-AI design treats learning, inference, and adaptation as first-class primitives in the way the RAN is built and operated. This is fundamentally different from today’s RAN system designs, where AI is mostly externalized, invoked at slower timescales, and constrained by legacy interfaces. In a native-AI architecture, intent, prediction, and actuation are tightly coupled at millisecond or even microsecond resolution, creating new possibilities for spectral efficiency, user experience optimization, and autonomous orchestration. A native-AI RAN would likely require heavier hardware at the edge of the network than today’s Open (or “classical”) RAN deployments. In the current architecture, the DU and RU rely on highly optimized deterministic hardware such as FPGAs, SmartNICs, and custom ASICs to execute PHY/MAC functions at predictable latencies and with tight power budgets. AI workloads are typically concentrated higher up in the stack, in the NRT-RIC or RT-RIC, where they can run on centralized GPU or CPU clusters without overwhelming the radio units. However, by contrast, a native-AI design pushes inference directly into the DU and even the RU, where microsecond-scale decisions on beamforming, HARQ, and link adaptation must be made. This implies the integration of embedded accelerators, such as AI-optimized ASICs, NPUs, or small-form-factor GPUs, into radio hardware, along with larger memory footprints for real-time model execution and storage. The resulting compute demand and cooling requirements could increase power consumption substantially beyond today’s SmartNIC-based O-RAN nodes. An effect that would be multiplied by millions of cell sites worldwide should such a design be chosen. This may (should!) raise concerns regarding both CapEx and OpEx due to higher costs for silicon and more demanding site engineering for power and heat management.

Article content
Figure: A comparison of the possible differences between today’s Open RAN and the AI-Native RAN Architecture. I should point out that the AI-Native RAN architecture is my own depiction and may not be how it may eventually look.

A native-AI RAN promises several advantages over existing architectures. By embedding intelligence directly into the control loops, the system can achieve higher spectral efficiency through ultra-fast adaptation of beamforming, interference management, and resource allocation, going beyond the limits of deterministic algorithms. It also allows for far more fine-grained optimization of the user experience, with decisions made per device, per flow, and in real-time, enabling predictive buffering and even semantic compression without noticeable delay. Operations themselves become more autonomous, with the RAN continuously tuning and healing itself in ways that reduce the need for manual intervention. Importantly, intent expressed at the management layer can be mapped directly into execution at the radio layer, creating continuity from policy to action that is missing in today’s O-RAN framework. Native-AI designs are also better able to anticipate and respond to extreme conditions, making the system more resilient under stress. Finally, they open the door to 6G concepts such as cell-less architectures, distributed massive MIMO, and AI-native PHY functions that cannot be realized under today’s layered, deterministic designs.

At the same time, the drawbacks of the Native-AI RAN approach may also be quite substantial. Embedding AI at microsecond control loops makes it almost impossible to trace reasoning steps or provide post-hoc explainability, creating tension with regulatory requirements such as the EU AI Act and NIS2. Because AI becomes the core operating fabric, mistakes, adversarial inputs, or misaligned objectives can cascade across the system much faster than in current architectures, amplifying the scale of failures. Continuous inference close to the radio layer also risks driving up compute demand and energy consumption far beyond what today’s SmartNIC- or FPGA-based solutions can handle. There is a danger of re-introducing vendor lock-in, as AI-native stacks may not interoperate cleanly with legacy xApps and rApps, undermining the very rationale of open interfaces. Training and refining these models requires sensitive operational and user data, raising privacy and data sovereignty concerns. Finally, the speed at which native-AI RANs operate makes meaningful human oversight nearly impossible, challenging the principle of human-in-the-loop control that regulators increasingly require for critical infrastructure operation.

Perhaps not too surprising, NVIDIA, a founding member of the AI-RAN Alliance, is a leading advocate for AI-native RAN, with strong leadership across infrastructure innovation, collaborative development, tooling, standard-setting, and future network frameworks. Their AI-Aerial platform and broad ecosystem partnerships illustrate their pivotal role in transitioning network architectures toward deeply integrated intelligence, especially in the 6G era. The AI-Native RAN concept and the gap it opens compared to existing O-RAN and classical RAN approaches will be the subject of a follow-up article I am preparing based on my current research into this field.

WHY REGULATORY AGENCIES MAY END THE AI PARTY (BEFORE IT REALLY STARTS).

Article content
Figure: Regulatory challenges for applying AI in critical telecom infrastructure, highlighting transparency, explainability, and auditability as key oversight requirements under European Commission mandates, posing constraints on AI-driven RAN systems.

We are about to “let loose” advanced AI/ML applications and processes across all aspects of our telecommunication networks. From the core all the way through to access and out to consumers and businesses making use of what is today regarded as highly critical infrastructure. This reduces cognitive load for operators while aiming to keep decision logic transparent, explainable, and auditable. In both roles, LLMs do not replace the specialized ML models running lower in the architecture. Instead, they enhance the orchestration layer by embedding reasoning and language understanding where time and resources permit. Yet it is here that one of the sharpest challenges emerges. The regulatory and policy scrutiny that inevitably follows when AI is introduced into critical infrastructure.

In the EU, the legal baseline now treats many network-embedded AI systems as high-risk by default whenever they are used as safety or operational components in the management and operation of critical digital infrastructure. This category encompasses modern telecom networks squarely. Under the EU AI Act, such systems must satisfy stringent requirements for risk management, technical documentation, transparency, logging, human oversight, robustness, and cybersecurity, and they must be prepared for conformity assessment and market surveillance. If the AI used in RAN control or orchestration cannot meet these duties, deployment can be curtailed or prohibited until compliance is demonstrated. The same regulation now also imposes obligations on general-purpose AI (foundation/LLM) providers, including additional duties when models are deemed to pose systemic risk, to enhance transparency and safety across the supply chain that may support telecom use cases. This AI-specific layer builds upon the EU’s broader critical infrastructure and cybersecurity regime. The NIS2 Directive strengthens security and incident-reporting obligations for essential entities, explicitly including digital and communications infrastructure, while promoting supply-chain due diligence. This means that operators must demonstrate how they assess and manage risks from AI components and vendors embedded in their networks. The EU’s 5G Cybersecurity Toolbox adds a risk-based, vendor-agnostic lens to supplier decisions (applied to “high-risk” vendors). Still, the logic is general: provenance alone, whether from China, the US, Israel, or any “friendly” jurisdiction, does not exempt AI/ML components from rigorous technical and governance assurances. The Cyber Resilience Act extends horizontal cybersecurity duties to “products with digital elements,” which can capture network software and AI-enabled components, linking market access to secure-by-design engineering, vulnerability handling, and update practices.

Data-protection law also bites. GDPR Article 22 places boundaries on decisions based solely on automated processing that produce legal or similarly significant effects on individuals, a genuine concern as networks increasingly mediate critical services and safety-of-life communications. Recent case law from the Court of Justice of the EU underscores a right of access to meaningful information about automated decision-making “procedures and principles,” raising the bar for explainability and auditability in any network AI that profiles or affects individuals. In short, operators must be able to show their work, not just that an AI policy improved a KPI, but how it made the call. These European guardrails are mirrored (though not identically) elsewhere. The UK Telecoms Security Act and its Code of Practice impose enforceable security measures on providers. In the US, the voluntary NIST AI Risk Management Framework has become the de facto blueprint for AI governance, emphasizing transparency, accountability, and human oversight, principles that regulators can (and do) import into sectoral supervision. None of these frameworks cares only about “who made it”. They also care about how it performs, how it fails, how it is governed, and how it can be inspected.

The AI Act’s human-oversight requirement (i.e., Article 14 in the EU Artificial Intelligence Act) exists precisely to bound such risks, ensuring operators can intervene, override, or disable when behavior diverges from safety or fundamental rights expectations. Its technical documentation and transparency obligations require traceable design choices and lifecycle records. Where these assurances cannot be demonstrated, regulators may limit or ban such deployments in critical infrastructure.

Against this backdrop, proposals to deploy autonomous AI agents deeply embedded in the RAN stack face a (very) higher bar. Autonomy risks eroding the very properties that European law demands.

  • Transparency – Reasoning steps are difficult to reconstruct: Traditional RAN algorithms are rule-based and auditable, making their logic transparent and reproducible. By contrast, modern AI models, especially deep learning and generative approaches, embed decision logic in complex weight matrices, where the precise reasoning steps cannot be reconstructed. Post-hoc explainability methods provide only approximations, not complete causal transparency. This creates tension with regulatory frameworks such as the EU AI Act, which requires technical documentation, traceability, and user-understandable logic for high-risk AI in critical infrastructure. The NIS2 Directive and GDPR Article 22 add further obligations for traceability and meaningful explanation of automated decisions. If operators cannot show why an AI system in the RAN made a given decision, compliance risks arise. The challenge is amplified with autonomous agents (i.e., Agentic AI), where decisions emerge from adaptive policies and interactions that are inherently non-deterministic. For critical infrastructure, such as telecom networks, transparency is therefore not optional but a regulatory necessity. Opaque models may face restrictions or outright bans.
  • Explainability – Decisions must be understandable: Explainability means that operators and regulators can not only observe what a model decided, but also understand why. In RAN AI, this is challenging because deep models may optimize across multiple features simultaneously, making their outputs hard to interpret. The EU AI Act requires high-risk systems to provide explanations that are “appropriate to the intended audience,” meaning engineers must be able to trace technical logic. In contrast, regulators and end-users require more accessible reasoning. Without explainability, trust in AI-driven traffic steering, slicing, or energy optimization cannot be established. A lack of clarity risks regulatory rejection and reduces operator confidence in deploying advanced AI at scale.
  • Auditability – Decisions must be verifiable: Auditability ensures that every AI-driven decision in the RAN can be logged, traced, and checked after the fact. Traditional rule-based schedulers are inherently auditable, but ML models, especially adaptive ones, require extensive logging frameworks to capture states, inputs, and outputs. The NIS2 Directive and the Cyber Resilience Act require such traceability for digital infrastructure, while the AI Act imposes additional obligations for record-keeping and post-market monitoring. Without audit trails, it becomes impossible to verify compliance or to investigate failures, outages, or discriminatory behaviors. In critical infrastructure, a lack of auditability is not just a technical gap but a regulatory showstopper, potentially leading to deployment bans.
  • Human Oversight – The challenge of real-time intervention: Both the EU AI Act and the NIS2 Directive require that high-risk AI systems remain under meaningful human oversight, with the possibility to override or disable AI-initiated actions. In the context of O-RAN, this creates a unique tension. Many RIC-driven optimizations and DU/RU control loops operate at millisecond or even microsecond timescales, where thousands or millions of inferences occur daily. Expecting a human operator to monitor, let alone intervene in real time, is technically infeasible. Instead, oversight must be implemented through policy guardrails, monitoring dashboards, fallback modes, and automated escalation procedures. The challenge is to satisfy the regulatory demand for human control without undermining the efficiency gains that AI brings. If this balance cannot be struck, regulators may judge certain autonomous functions non-compliant, slowing or blocking their deployment in critical telecom infrastructure.

The upshot for telecom is clear. Even as generative and agentic AI move into SMO/Non-RT orchestration for intent translation or semantic compression, the time-scale fundamentals do not change. RT and sub-ms loops must remain deterministic, inspectable, and controllable, with human-governed, well-documented interfaces mediating any AI influence. The regulatory risk is therefore not hypothetical. It is structural. As generative AI and LLMs move closer to the orchestration and policy layers of O-RAN, their opacity and non-deterministic reasoning raise questions about compliance. While such models may provide valuable tools for intent interpretation or telemetry summarization, their integration into live networks will only be viable if accompanied by robust frameworks for explainability, monitoring, and assurance. This places a dual burden on operators and vendors: to innovate in AI-driven automation, but also to invest in governance structures that can withstand regulatory scrutiny.

In a European context, no AI model will likely be permitted in the RAN unless it can pass the tests of explainability, auditability, and human oversight that regulators will and also should demand of functionality residing in critical infrastructures.

WRAPPING UP.

The article charts an evolution from SON-era automation to today’s AI-RAN vision, showing how O-RAN institutionalized “openness + intelligence” through a layered control stack, SMO/NRT-RIC for policy and learning, RT-RIC for fast decisions, and CU/DU/RU for deterministic execution at millisecond to microsecond timescales. It argues that LLMs belong at the top (SMO/NRT-RIC) for intent translation and semantic compression, while lightweight supervised/RL/TinyML models run the real-time loops below. “ChatGPT-like” systems (i.e., founded on human-generated context) are ill-suited to near-RT and sub-ms control. Synthetic data can stress-test rare events, but it demands statistics that are aware of extremes and validation against real holdouts to avoid misleading inference. Many low-level PHY/MAC primitives (HARQ, coding/decoding, CRC, MMSE precoding, and SINR-based link adaptation) are generally close to optimal, so AI/ML’s gains in these areas may be marginal and, at least initially, not the place to focus on.

Most importantly, pushing agentic autonomy too deep into the stack is likely to collide with both physics and law. Without reversibility, logging, and explainability, deployments risk breaching the EU AI Act’s requirements for human oversight, transparency, and lifecycle accountability. The practical stance is clear. Keep RT-RIC and DU/RU loops deterministic and inspectable, confine agency to SMO/NRT-RIC under strong policy guardrails and observability, and pair innovation with governance that can withstand regulatory scrutiny.

  • AI in RAN is evolutionary, not revolutionary, from SON and Elastic RAN-style coordination to GPU-accelerated AI-RAN and the 2024 AI-RAN Alliance.
  • O-RAN’s design incorporates AI via a hierarchical approach: SMO (governance/intent), NRT-RIC (training/policy), RT-RIC (near-real-time decisions), CU (shaping/QoS/UX, etc.), and DU/RU (deterministic PHY/MAC).
  • LLMs are well-suited for SMO/NRT-RIC for intent translation and semantic compression; however, they are ill-suited for RT-RIC or DU/RU, where millisecond–to–microsecond determinism is mandatory.
  • Lightweight supervised/RL/TinyML models, not “ChatGPT-like” systems, are the practical engines for near-real-time and real-time control loops.
  • Synthetic data for rare events, generated in the NRT-RIC and SMO, is valid but carries some risk. Approaches must be validated against real holdouts and statistics that account for extremes to avoid misleading inference.
  • Many low-level PHY/MAC primitives (HARQ, coding/decoding, CRC, classical precoding/MMSE, SINR-based link adaptation) are already near-optimal. AI may only add marginal gains at the edge.
  • Regulatory risk: Deep agentic autonomy without reversibility threatens EU AI Act Article 14 (human oversight). Operators must be able to intervene/override, which, to an extent, may defeat the more aggressive pursuits of autonomous network operations.
  • Regulatory risk: Opaque/unanalyzable models undermine transparency and record-keeping duties (Articles 12–13), especially if millions of inferences lack traceable logs and rationale.
  • Regulatory risk: For systems affecting individuals or critical services, explainability obligations (including GDPR Article 22 context) and AI Act lifecycle controls (Articles 8–17) require audit trails, documentation, and post-market monitoring, as well as curtailment of non-compliant agentic behavior risks.
  • Practical compliance stance: It may make sense to keep RT-RIC and DU/RU loops deterministic and inspectable, and constrain agency to SMO/NRT-RIC with strong policy guardrails, observability, and fallback modes.

ABBREVIATION LIST.

  • 3GPP – 3rd Generation Partnership Project.
  • A1 – O-RAN Interface between Non-RT RIC and RT-RIC.
  • AAS – Active Antenna Systems.
  • AISG – Antenna Interface Standards Group.
  • AI – Artificial Intelligence.
  • AI-RAN – Artificial Intelligence for Radio Access Networks.
  • AI-Native RAN – Radio Access Network with AI embedded into architecture, protocols, and control loops.
  • ASIC – Application-Specific Integrated Circuit.
  • CapEx – Capital Expenditure.
  • CPU – Central Processing Unit.
  • C-RAN – Cloud Radio Access Network.
  • CRC – Cyclic Redundancy Check.
  • CU – Centralized Unit.
  • DU – Distributed Unit.
  • E2 – O-RAN Interface between RT-RIC and CU/DU.
  • eCPRI – Enhanced Common Public Radio Interface.
  • EU – European Union.
  • FCAPS – Fault, Configuration, Accounting, Performance, Security.
  • FPGA – Field-Programmable Gate Array.
  • F1 – 3GPP-defined interface split between CU and DU.
  • GDPR – General Data Protection Regulation.
  • GPU – Graphics Processing Unit.
  • GRU – Gated Recurrent Unit.
  • HARQ – Hybrid Automatic Repeat Request.
  • KPI – Key Performance Indicator.
  • L1/L2 – Layer 1 / Layer 2 (in the OSI stack, PHY and MAC).
  • LLM – Large Language Model.
  • LSTM – Long Short-Term Memory.
  • MAC – Medium Access Control.
  • MANO – Management and Orchestration.
  • MIMO – Multiple Input, Multiple Output.
  • ML – Machine Learning.
  • MMSE – Minimum Mean Square Error.
  • NFVI – Network Functions Virtualization Infrastructure.
  • NIS2 – EU Directive on measures for a high standard level of cybersecurity across the Union.
  • NPU – Neural Processing Unit.
  • NRT-RIC – Non-Real-Time RAN Intelligent Controller.
  • O1 – O-RAN Operations and Management Interface to network elements.
  • O2 – O-RAN Interface to cloud infrastructure (NFVI and MANO).
  • O-RAN – Open Radio Access Network.
  • OpEx – Operating Expenditure.
  • PDCP – Packet Data Convergence Protocol.
  • PHY – Physical Layer.
  • QoS – Quality of Service.
  • RAN – Radio Access Network.
  • rApp – Non-Real-Time RIC Application.
  • RET – Remote Electrical Tilt.
  • RIC – RAN Intelligent Controller.
  • RLC – Radio Link Control.
  • R-NIB – Radio Network Information Base.
  • RT-RIC – Real-Time RAN Intelligent Controller.
  • RU – Radio Unit.
  • SDAP – Service Data Adaptation Protocol.
  • SINR – Signal-to-Interference-plus-Noise Ratio.
  • SmartNIC – Smart Network Interface Card.
  • SMO – Service Management and Orchestration.
  • SON – Self-Organizing Network.
  • T-Labs – Deutsche Telekom Laboratories.
  • TTI – Transmission Time Interval.
  • UE – User Equipment.
  • US – United States.
  • WG2 – O-RAN Working Group 2 (Non-RT RIC & A1 interface).
  • WG3 – O-RAN Working Group 3 (RT-RIC & E2 Interface).
  • xApp – Real-Time RIC Application.

ACKNOWLEDGEMENT.

I want to acknowledge my wife, Eva Varadi, for her unwavering support, patience, and understanding throughout the creative process of writing this article.

FOLLOW-UP READING.

  1. Kim Kyllesbech Larsen (May 2023), “Conversing with the Future: An interview with an AI … Thoughts on our reliance on and trust in generative AI.” An introduction to generative models and large language models.
  2. Goodfellow, I., Bengio, Y., Courville, A. (2016), Deep Learning (Adaptive Computation and Machine Learning series). The MIT Press. Kindle Edition.
  3. Collins, S. T., & Callahan, C. W. (2009). Cultural differences in systems engineering: What they are, what they aren’t, and how to measure them. 19th Annual International Symposium of the International Council on Systems Engineering, INCOSE 2009, 2.
  4. Herzog, J. (2015). Software Architecture in Practice, Third Edition, Written by Len Bass, Paul Clements, and Rick Kazman. ACM SIGSOFT Software Engineering Notes, 40(1).
  5. O-RAN Alliance (October 2018). “O-RAN: Towards an Open and Smart RAN“.
  6. TS 103 982 – V8.0.0. (2024) – Publicly Available Specification (PAS); O-RAN Architecture Description (O-RAN.WG1.OAD-R003-v08.00).
  7. Lee, H., Cha, J., Kwon, D., Jeong, M., & Park, I. (2020, December 1). “Hosting AI/ML Workflows on O-RAN RIC Platform”. 2020 IEEE Globecom Workshops, GC Wkshps 2020 – Proceedings.
  8. TS 103 983 – V3.1.0. (2024)- Publicly Available Specification (PAS); A1 interface: General Aspects and Principles (O-RAN.WG2.A1GAP-R003-v03.01).
  9. TS 104 038 – V4.1.0. (2024) – Publicly Available Specification (PAS); E2 interface: General Aspects and Principles (O-RAN.WG3.E2GAP-R003-v04.01).
  10. TS 104 039 – V4.0.0. (2024) – Publicly Available Specification (PAS); E2 interface: Application Protocol (O-RAN.WG3.E2AP-R003-v04.00).
  11. TS 104 040 – V4.0.0. (2024) – Publicly Available Specification (PAS); E2 interface: Service Model (O-RAN.WG3.E2SM-R003-v04.00).
  12. O-RAN Work Group 3. (2025). Near-Real-time RAN Intelligent Controller E2 Service Model (E2SM) KPM Technical Specification.
  13. Bao, L., Yun, S., Lee, J., & Quek, T. Q. S. (2025). LLM-hRIC: LLM-empowered Hierarchical RAN Intelligent Control for O-RAN.
  14. Tang, Y., Srinivasan, U. C., Scott, B. J., Umealor, O., Kevogo, D., & Guo, W. (2025). End-to-End Edge AI Service Provisioning Framework in 6G ORAN.
  15. Gajjar, P., & Shah, V. K. (n.d.). ORANSight-2.0: Foundational LLMs for O-RAN.
  16. Elkael, M., D’Oro, S., Bonati, L., Polese, M., Lee, Y., Furueda, K., & Melodia, T. (2025). AgentRAN: An Agentic AI Architecture for Autonomous Control of Open 6G Networks.
  17. Gu, J., Zhang, X., & Wang, G. (2025). Beyond the Norm: A Survey of Synthetic Data Generation for Rare Events.
  18. Michael Peel (July 2024), The problem of ‘model collapse’: how a lack of human data limits AI progress, Financial Times.
  19. Decruyenaere, A., Dehaene, H., Rabaey, P., Polet, C., Decruyenaere, J., Demeester, T., & Vansteelandt, S. (2025). Debiasing Synthetic Data Generated by Deep Generative Models.
  20. Decruyenaere, A., Dehaene, H., Rabaey, P., Polet, C., Decruyenaere, J., Vansteelandt, S., & Demeester, T. (2024). The Real Deal Behind the Artificial Appeal: Inferential Utility of Tabular Synthetic Data.
  21. Vishwakarma, R., Modi, S. D., & Seshagiri, V. (2025). Statistical Guarantees in Synthetic Data through Conformal Adversarial Generation.
  22. Banbury, C. R., Reddi, V. J., Lam, M., Fu, W., Fazel, A., Holleman, J., Huang, X., Hurtado, R., Kanter, D., Lokhmotov, A., Patterson, D., Pau, D., Seo, J., Sieracki, J., Thakker, U., Verhelst, M., & Yadav, P. (2021). Benchmarking TinyML Systems: Challenges and Direction.
  23. Capogrosso, L., Cunico, F., Cheng, D. S., Fummi, F., & Cristani, M. (2023). A Machine Learning-oriented Survey on Tiny Machine Learning.
  24. Hubara, I., Courbariaux, M., Soudry, D., El-Yaniv, R., & Bengio, Y. (2016). Quantized Neural Networks: Training Neural Networks with Low Precision Weights and Activations.
  25. AI Act. The AI Act is the first-ever comprehensive legal framework on AI, addressing the risks associated with AI, and is alleged to position Europe to play a leading role globally (as claimed by the European Commission).
  26. The EU Artificial Intelligence Act. For matters related explicitly to Critical Infrastructure, see in particular Annex III: High-Risk AI Systems Referred to in Article 6(2), Recital 55 and Article 6: Classification Rules for High-Risk AI Systems. I also recommend taking a look at “Article 14: Human Oversight”.
  27. European Commission (January 2020), “Cybersecurity of 5G networks – EU Toolbox of risk mitigating measures”.
  28. European Commission (June 2023), “Commission announces next steps on cybersecurity of 5G networks in complement to latest progress report by Member States”.
  29. European Commission, “NIS2 Directive: securing network and information systems”.
  30. Council of Europe (October 2024), “Cyber resilience act: Council adopts new law on security requirements for digital products.”.
  31. GDPR Article 22, “Automated individual decision-making, including profiling”. See also the following article from Crowell & Moring LLP: “Europe’s Highest Court Compels Disclosure of Automated Decision-Making “Procedures and Principles” In Data Access Request Case”.

Will LEO Satellite Direct-to-Cell Networks make Terrestrial Networks Obsolete?

Posted on by
5

THE POST-TOWER ERA – A FAIRYTAIL.

From the bustling streets of New York to the remote highlands of Mongolia, the skyline had visibly changed. Where steel towers and antennas once dominated now stood open spaces and restored natural ecosystems. Forests reclaimed their natural habitats, and birds nested in trees undisturbed by the scaring of high rural cellular towers. This transformation was not sudden but resulted from decades of progress in satellite technology, growing demand for ubiquitous connectivity, an increasingly urgent need to address the environmental footprint of traditional telecom infrastructures, and the economic need to dramatically reduce operational expenses tied up in tower infrastructure. By the time the last cell site was decommissioned, society stood at the cusp of a new age of connectivity by LEO satellites covering all of Earth.

The annual savings worldwide from making terrestrial cellular towers obsolete in total cost are estimated to amount to at least 300 billion euros, and it is expected that moving cellular access to “heaven” will avoid more than 150 million metric tons of CO2 emissions annually. The retirement of all terrestrial cellular networks worldwide has been like eliminating the entire carbon footprint of The Netherlands or Malaysia and leading to a dramatic reduction in demand for sustainable green energy sources that previously were used to power the global cellular infrastructure.

INTRODUCTION.

Recent postings and a substantial part of commentary give the impression that we are heading towards a post-tower era where Elon Musk’s Low Earth Orbit (LEO) satellite Starlink network (together with competing options, e.g., ATS Spacemobile and Lynk, and no, I do not see Amazon’s Project Kuiper in this space) will make terrestrially-based tower infrastructure and earth-bound cellular services obsolete.

T-Mobile USA is launching its Direct-to-Cell (D2C) service via SpaceX’s Starlink LEO satellite network. The T-Mobile service is designed to work with existing LTE-compatible smartphones, allowing users to connect to Starlink satellites without needing specialized hardware or smartphone applications.

Since the announcement, posts and media coverage have declared the imminent death of the terrestrial cellular network. When it is pointed out that this may be a premature death sentence to an industry, telecom operators, and their existing cellular mobile networks, it is also not uncommon to be told off as being too pessimistic and an unbeliever in Musk’s genius vision. Musk has on occasion made it clear the Starlink D2C service is aimed at texts and voice calls in remote and rural areas, and to be honest, the D2C service currently hinges on 2×5 MHz in the T-Mobile’s PCS band, adding constraints to the “broadbandedness” of the service. The fact that the service doesn’t match the best of T-Mobile US’s 5G network quality (e.g., 205+ Mbps downlink) or even get near its 4G speeds should really not bother anyone, as the value of the D2C service is that it is available in remote and rural areas with little to no terrestrial cellular coverage and that you can use your regular cellular device with no need for a costly satellite service and satphone (e.g., Iridium, Thuraya, Globalstar).

While I don’t expect to (or even want to) change people’s beliefs, I do think it would be great to contribute to more knowledge and insights based on facts about what is possible with low-earth orbiting satellites as a terrestrial substitute and what is uninformed or misguided opinion.

The rise of LEO satellites has sparked discussions about the potential obsolescence of terrestrial cellular networks. With advancements in satellite technology and increasing partnerships, such as T-Mobile’s collaboration with SpaceX’s Starlink, proponents envision a future where towers are replaced by ubiquitous connectivity from the heavens. However, the feasibility of LEO satellites achieving service parity with terrestrial networks raises significant technical, economic, and regulatory questions. This article explores the challenges and possibilities of LEO Direct-to-Cell (D2C) networks, shedding light on whether they can genuinely replace ground-based cellular infrastructure or will remain a complementary technology for specific use cases.

WHY DISTANCE MATTERS.

The distance between you (your cellular device) and the base station’s antenna determines your expected service experience in cellular and wireless networks. The longer you are away from the base station that serves you, in general, the poorer your connection quality and performance will be, with everything else being equal. As the distance increases, signal weakening (i.e., path loss) grows exponentially, reducing signal quality and making it harder for devices to maintain reliable communication. Closer proximity allows for more substantial, faster, and more stable connections, while longer distances require more power and advanced technologies like beamforming or repeaters to compensate.

Physics tells us how a signal loses its signal strength (or power) over a distance with the square of the distance from the source of the signal itself (either the base station transmitter or the consumer device). This applies universally to all electromagnetic waves traveling in free space. Free space means that there are no obstacles, reflections, or scattering. No terrain features, buildings, or atmospheric conditions interfere with the propagation signal.

So, what matters to the Free Space Path Loss (FSPL)? That is the signal strength over a given distance in free space:

  • The signal strength reduces (the path loss increases) with the square of the distance (d) from its source.
  • Path loss increases (i.e., signal strength decreases) with the (square of the) frequency (f). The higher the frequency, the higher the path loss at a given distance from the signal source.
  • A larger transmit antenna aperture reduces the path loss by focusing the transmitted signal (energy) more efficiently. An antenna aperture is an antenna’s “effective area” that captures or transmits electromagnetic waves. It depends directly on antenna gain and inverse of the square of the signal frequency (i.e., higher frequency → smaller aperture).
  • Higher receiver gain will also reduce the path loss.

$PL_{FS} \; = \; \left( \frac{4 \pi}{c} \right)^2 (d \; f)^2 \; \propto d^2 \; f^2$

$$FSPL_{dB} \; = 10 \; Log_{10} (PL_{FS}) \; = \; 20 \; Log_{10}(d) \; + \; 20 \; Log_{10}(f) \; + \; constant$$

The above equations show a strong dependency on distance; the farther away, the larger the signal loss, and the higher the frequency, the larger the signal loss. Relaxing some of the assumptions leading to the above relationship leads us to the following:

$FSPL_{dB}^{rs} \; = \; 20 \; Log_{10}(d) \; – \; 10 \; Log_{10}(A_t^{eff}) \; – \; 10 \; Log_{10}(G_{r}) \; + \; constant$

The last of the above equations introduces the transmitter’s effective antenna aperture (\(A_t^{eff}\)) and the receiver’s gain (\(G_r\)), telling us that larger apertures reduce path loss as they focus the transmitted energy more efficiently and that higher receiver gain likewise reduces the path loss (i.e., “they hear better”).

It is worth remembering that the transmitter antenna aperture is directly tied to the transmitter gain ($G_t$) when the frequency (f) has been fixed. We have

$A_t^{eff} \; = \; \frac{c^2}{4\pi} \; \frac{1}{f^2} \; G_t \; = \; 0.000585 \; m^2 \; G_t \;$ @ f = 3.5 GHz.

From the above, as an example, it is straightforward to see that the relative path loss difference between the two distances of 550 km (e.g., typical altitude of an LEO satellite) and 2.5 km (typical terrestrial cellular coverage range ) is

$\frac{PL_{FS}(550 km)}{PL_{FS}(2.5 km)} \; = \; \left( \frac {550}{2.5}\right)^2 \; = \; 220^2 \; \approx \; 50$ thousand. So if all else was equal (it isn’t, btw!), we would expect that the signal loss at a distance of 550 km would be 50 thousand times higher than at 2.5 km. Or, in the electrical engineer’s language, at a distance of 550 km, the loss would be 47 dB higher than at 2.5 km.

The figure illustrates the difference between (a) terrestrial cellular and (b) satellite coverage. A terrestrial cellular signal typically covers a radius of 0.5 to 5 km. In contrast, a LEO satellite signal travels a substantial distance to reach Earth (e.g., Starlink satellite is at an altitude of 550 km). While the terrestrial signal propagates through the many obstacles it meets on its earthly path, the satellite signal’s propagation path would typically be free-space-like (i.e., no obstacles) until it penetrates buildings or other objects to reach consumer devices. Historically, most satellite-to-Earth communication has relied on outdoor ground stations or dishes where the outdoor antenna on Earth provides LoS to the satellite and will also compensate somewhat for the signal loss due to the distance to the satellite.

Let’s compare a terrestrial 5G 3.5 GHz advanced antenna system (AAS) 2.5 km from a receiver with a LEO satellite system at an altitude of 550 km. Note I could have chosen a lower frequency, e.g., 800 MHz or the PCS 1900 band. While it would give me some advantages regarding path loss (i.e., $FSPL \; \propto \; f^2$), the available bandwidth is rather smallish and insufficient for state-or-art 5G services (imo!). From a free-space path loss perspective, independently of frequency, we need to overcome an almost 50 thousand times relative difference in distance squared (ca. 47 dB difference) in favor of the terrestrial system. In this comparison, it should be understood that the terrestrial and the satellite systems use the same carrier frequency (otherwise, one should account for the difference in frequency), and the only difference that matters (for the FSPL) is the difference in distance to the receiver.

Suppose I require that my satellite system has the same signal loss in terms of FSPL as my terrestrial system to aim at a comparable quality of service level. In that case, I have several options in terms of satellite enhancements. I could increase transmit power, although it would imply that I need a transmit power of 47 dB more than the terrestrial system, or approximately 48 kW, which is likely impractical for the satellite due to power limitations. Compare this with the current Starlink transmit power of approximately 32 W (45 dBm), ca. 1,500 times lower. Alternatively, I could (in theory!) increase my satellite antenna aperture, leading to a satellite antenna with a diameter of ca. 250 meters, which is enormous compared to current satellite antennas (e.g., Starlink’s ca. 0.05 m2 aperture for a single antenna and total area in the order of 1.6 m2 for the Ku/Ka bands). Finally, I could (super theoretically) also massively improve my consumer device (e.g., smartphone) to receive gain (with 47 dB) from today’s range of -2 dBi to +5 dBi. Achieving 46 dBi gain in a smartphone receiver seems unrealistic due to size, power, and integration constraints. As the target of LEO satellite direct-to-cell services is to support commercially available cellular devices used in terrestrial, only the satellite specifications can be optimized.

Based on a simple free-space approach, it appears unreasonable that an LEO satellite communication system can provide 5G services at parity with a terrestrial cellular network to normal (unmodified) 5G consumer devices without satellite-optimized modifications. The satellite system’s requirements for parity with a terrestrial communications system are impractical (but not impossible) and, if pursued, would significantly drive up design complexity and cost, likely making such a system highly uneconomical.

At this point, you should ask yourself if it is reasonable to assume that a terrestrial communication cellular system can be taken to propagate as its environment is “free-space” like. Thus, obstacles, reflections, and scattering are ignored. Is it really okay to presume that terrain features, buildings, or atmospheric conditions do not interfere with the propagation of the terrestrial cellular signal? Of course, the answer should be that it is not okay to assume that. When considering this, let’s see if it matters much compared to the LEO satellite path loss.

TERRESTRIAL CELLULAR PROPAGATION IS NOT HAPPENING IN FREE SPACE, AND NEITHER IS A SATELLITE’S.

The Free-Space Path Loss (FSPL) formula assumes ideal conditions where signals propagate in free space without interference, blockage, or degradation, besides what would naturally be by traveling a given distance. However, as we all experience daily, real-world environments introduce additional factors such as obstructions, multipath effects, clutter loss, and environmental conditions, necessitating corrections to the FSPL approach. Moving from one room of our house to another can easily change the cellular quality and our experience (e.g., dropped calls, poorer voice quality, lower speed, changing from using 5G to 4G or even to 2G, no coverage at all). Driving through a city may also result in ups and downs with respect to the cellular quality we experience. Some of these effects are tabulated below.

Urban environments typically introduce the highest additional losses due to dense buildings, narrow streets, and urban canyons, which significantly obstruct and scatter signals. For example, the Okumura-Hata Urban Model accounts for such obstructions and adds substantial losses to the FSPL, averaging around 30–50 dB, depending on the density and height of buildings.

Suburban environments, on the other hand, are less obstructed than urban areas but still experience moderate clutter losses from trees, houses, and other features. In these areas, corrections based on the Okumura-Hata Suburban Model add approximately 10–20 dB to the FSPL, reflecting the moderate level of signal attenuation caused by vegetation and scattered structures.

Rural environments have the least obstructions, resulting in the lowest additional loss. Corrections based on the Okumura-Hata Rural Model typically add around 5–10 dB to the FSPL. These areas benefit from open landscapes with minimal obstructions, making them ideal for long-range signal propagation.

Non-line-of-sight (NLOS) conditions increase additionally the path loss, as signals must diffract or scatter to reach the receiver. This effect adds 10–20 dB in suburban and rural areas and 20–40 dB in urban environments, where obstacles are more frequent and severe. Similarly, weather conditions such as rain and foliage contribute to signal attenuation, with rain adding up to 1–5 dB/km at higher frequencies (above 10 GHz) and dense foliage introducing an extra 5–15 dB of loss.

The corrections for these factors can be incorporated into the FSPL formula to provide a more realistic estimation of signal attenuation. By applying these corrections, the FSPL formula can reflect the conditions encountered in terrestrial communication systems across different environments.

The figure above illustrates the differences and similarities concerning the coverage environment for (a) terrestrial and (b) satellite communication systems. The terrestrial signal environment, in most instances, results in the loss of the signal as it propagates through the terrestrial environment due to vegetation, terrain variations, urban topology or infrastructure, weather, and ultimately, as the signal propagates from the outdoor environment to the indoor environment it signal reduces further as it, for example, penetrates windows with coatings, outer and inner walls. The combination of distance, obstacles, and material penetration leads to a cumulative reduction in signal strength as the signal propagates through the terrestrial environment. For the satellite, as illustrated in (b), a substantial amount of signal is reduced due to the vast distance it has to travel before reaching the consumer. If no outdoor antenna connects with the satellite signal, then the satellite signal will be further reduced as it penetrates roofs, multiple ceilings, multiple floors, and walls.

It is often assumed that a satellite system has a line of sight (LoS) without environmental obstructions in its signal propagation (besides atmospheric ones). The reasoning is not unreasonable as the satellite is on top of the consumers of its services and, of course, a correct approach when the consumer has an outdoor satellite receiver (e.g., a dish) in direct LoS with the satellite. Moreover, historically, most satellite-to-Earth communication has relied on outdoor ground stations or outdoor dishes (e.g., placed on roofs or another suitable location) where the outdoor antenna on Earth provides LoS to the satellite’s antenna also compensating somewhat for the signal loss due to the distance to the satellite.

When considering a satellite direct-to-cell device, we no longer have the luxury of a satellite-optimized advanced Earth-based outdoor antenna to facilitate the communications between the satellite and the consumer device. The satellite signal has to close the connection with a standard cellular device (e.g., smartphone, tablet, …), just like the terrestrial cellular network would have to do.

However, 80% or more of our mobile cellular traffic happens indoors, in our homes, workplaces, and public places. If a satellite system had to replace existing mobile network services, it would also have to provide a service quality similar to that of consumers from the terrestrial cellular network. As shown in the above figure, this involves urban areas where the satellite signal will likely pass through a roof and multiple floors before reaching a consumer. Depending on housing density, buildings (shadowing) may block the satellite signal, resulting in substantial service degradation for consumers suffering from such degrading effects. Even if the satellite signal would not face the same challenges as a terrestrial cellular signal, such as with vegetation, terrain variations, and the horizontal dimension of urban topology (e.g., outer& inner walls, coated windows,… ), the satellite signal would still have to overcome the vertical dimension of urban topologies (e..g, roofs, ceilings, floors, etc…) to connect to consumers cellular devices.

For terrestrial cellular services, the cellular network’s signal integrity will (always) have a considerable advantage over the satellite signal because of the proximity to the consumer’s cellular device. With respect to distance alone, an LEO satellite at an altitude of 550 km will have to overcome a 50 thousand times (or a 47 dB) path loss compared to a cellular base station antenna 2.5 km away. Overcoming that path loss penalty adds considerable challenges to the antenna design, which would seem highly challenging to meet and far from what is possible with today’s technology (and economy).

CHALLENGES SUMMARIZED.

Achieving parity between a Low Earth Orbit (LEO) satellite providing Direct-to-Cell (D2C) services and a terrestrial 5G network involves overcoming significant technical challenges. The disparity arises from fundamental differences in these systems’ environments, particularly in free-space path loss, penetration loss, and power delivery. Terrestrial networks benefit from closer proximity to the consumer, higher antenna density, and lower propagation losses. In contrast, LEO satellites must address far more significant free-space path losses due to the large distances involved and the additional challenges of transmitting signals through the atmosphere and into buildings.

The D2C challenges for LEO satellites are increasingly severe at higher frequencies, such as 3.5 GHz and above. As we have seen above, the free-space path loss increases with the square of the frequency, and penetration losses through common building materials, such as walls and floors, are significantly higher. For an LEO satellite system to achieve indoor parity with terrestrial 5G services at this frequency, it would need to achieve extraordinary levels of effective isotropic radiated power (EIRP), around 65 dB, and narrow beamwidths of approximately 0.5° to concentrate power on specific service areas. This would require very high onboard power outputs, exceeding 1 kW, and large antenna apertures, around 2 m in diameter, to achieve gains near 55 dBi. These requirements place considerable demands on satellite design, increasing mass, complexity, and cost. Despite these optimizations, indoor service parity at 3.5 GHz remains challenging due to persistent penetration losses of around 20 dB, making this frequency better suited for outdoor or line-of-sight applications.

Achieving a stable beam with the small widths required for a LEO satellite to provide high-performance Direct-to-Cell (D2C) services presents significant challenges. Narrow beam widths, on the order of 0.5° to 1°, are essential to effectively focus the satellite’s power and overcome the high free-space path loss. However, maintaining such precise beams demands advanced satellite antenna technologies, such as high-gain phased arrays or large deployable apertures, which introduce design, manufacturing, and deployment complexities. Moreover, the satellite must continuously track rapidly moving targets on Earth as it orbits around 7.8 km/s. This requires highly accurate and fast beam-steering systems, often using phased arrays with electronic beamforming, to compensate for the relative motion between the satellite and the consumer. Any misalignment in the beam can result in significant signal degradation or complete loss of service. Additionally, ensuring stable beams under variable conditions, such as atmospheric distortion, satellite vibrations, and thermal expansion in space, adds further layers of technical complexity. These requirements increase the system’s power consumption and cost and impose stringent constraints on satellite design, making it a critical challenge to achieve reliable and efficient D2C connectivity.

As the operating frequency decreases, the specifications for achieving parity become less stringent. At 1.8 GHz, the free-space path loss and penetration losses are lower, reducing the signal deficit. For a LEO satellite operating at this frequency, a 2.5 m² aperture (1.8 m diameter) antenna and an onboard power output of around 800 W would suffice to deliver EIRP near 60 dBW, bringing outdoor performance close to terrestrial equivalency. Indoor parity, while more achievable than 3.5 GHz, would still face challenges due to penetration losses of approximately 15 dB. However, the balance between the reduced propagation losses and achievable satellite optimizations makes 1.8 GHz a more practical compromise for mixed indoor and outdoor coverage.

At 800 MHz, the frequency-dependent losses are significantly reduced, making it the most feasible option for LEO satellite systems to achieve parity with terrestrial 5G networks. The free-space path loss decreases further, and penetration losses into buildings are reduced to approximately 10 dB, comparable to what terrestrial systems experience. These characteristics mean that the required specifications for the satellite system are notably relaxed. A 1.5 m² aperture (1.4 m diameter) antenna, combined with a power output of 400 W, would achieve sufficient gain and EIRP (~55 dBW) to deliver robust outdoor coverage and acceptable indoor service quality. Lower frequencies also mitigate the need for extreme beamwidth narrowing, allowing for more flexible service deployment.

Most consumers’ cellular consumption happens indoors. These consumers are compared to an LEO satellite solution typically better served by existing 5G cellular broadband networks. When considering a direct-to-normal-cellular device, it would not be practical to have an LEO satellite network, even an extensive one, to replace existing 5G terrestrial-based cellular networks and the services these support today.

This does not mean that LEO satellite cannot be of great utility when connecting to an outdoor Earth-based consumer dish, as is already evident in many remote, rural, and suburban places. The summary table above also shows that LEO satellite D2C services are feasible, without too challenging modifications, at the lower cellular frequency ranges between 600 MHz to 1800 MHz at service levels close to the terrestrial systems, at least in rural areas and for outdoor services in general. In indoor situations, the LEO Satellite D2C signal is more likely to be compromised due to roof and multiple floor penetration scenarios to which a terrestrial signal may be less exposed.

WHAT GOES DOWN MUST COME UP.

LEO satellite services that provide direct to unmodified mobile cellular device services are getting us all too focused on the downlink path from the satellite directly to the device. It seems easy to forget that unless you deliver a broadcast service, we also need the unmodified cellular device to directly communicate meaningfully with the LEO satellite. The challenge for an unmodified cellular device (e.g., smartphone, tablet, etc.) to receive the satellite D2C signal has been explained extensively in the previous section. In the satellite downlink-to-device scenario, we can optimize the design specifications of the LEO satellite to overcome some (or most, depending on the frequency) of the challenges posed by the satellite’s high altitude (compared to a terrestrial base station’s distance to the consumer device). In the device direct-uplink-to-satellite, we have very little to no flexibility unless we start changing the specifications of the terrestrial device portfolio. Suppose we change the specifications for consumer devices to communicate better with satellites. In that case, we also change the premise and economics of the (wrong) idea that LEO satellites should be able to completely replace terrestrial cellular networks at service parity with those terrestrial cellular networks.

Achieving uplink communication from a standard cellular device to an LEO satellite poses significant challenges, especially when attempting to match the performance of a terrestrial 5G network. Cellular devices are designed with limited transmission power, typically in the range of 23–30 dBm (0.2–1 watt), sufficient for short-range communication with terrestrial base stations. However, when the receiving station is a satellite orbiting between 550 and 1,200 kilometers, the transmitted signal encounters substantial free-space path loss. The satellite must, therefore, be capable of detecting and processing extremely weak signals, often below -120 dBm, to maintain a reliable connection.

The free-space path loss in the uplink direction is comparable to that in the downlink, but the challenges are compounded by the cellular device’s limitations. At higher frequencies, such as 3.5 GHz, path loss can exceed 155 dB, while at 1.8 GHz and 800 MHz, it reduces to approximately 149.6 dB and 143.6 dB, respectively. Lower frequencies favor uplink communication because they experience less path loss, enabling better signal propagation over large distances. However, cellular devices typically use omnidirectional antennas with very low gain (0–2 dBi), poorly suited for long-distance communication, placing even greater demands on the satellite’s receiving capabilities.

The satellite must compensate for these limitations with highly sensitive receivers and high-gain antennas. Achieving sufficient antenna gain requires large apertures, often exceeding 4 meters in diameter for 800 MHz or 2 meters for 3.5 GHz, increasing the satellite’s size, weight, and complexity. Phased-array antennas or deployable reflectors are often used to achieve the required gain. Still, their implementation is constrained by the physical limitations and costs of launching such systems into orbit. Additionally, the satellite’s receiver must have an exceptionally low noise figure, typically in the range of 1–3 dB, to minimize internal noise and allow the detection of weak uplink signals.

Interference is another critical challenge in the uplink path. Unlike terrestrial networks, where signals from individual devices are isolated into small sectors, satellites receive signals over larger geographic areas. This broad coverage makes it difficult to separate and process individual transmissions, particularly in densely populated areas where numerous devices transmit simultaneously. Managing this interference requires sophisticated signal processing capabilities on the satellite, increasing its complexity and power demands.

The motion of LEO satellites introduces additional complications due to the Doppler effect, which causes a shift in the uplink signal frequency. At higher frequencies like 3.5 GHz, these shifts are more pronounced, requiring real-time adjustments to the receiver to compensate. This dynamic frequency management adds another layer of complexity to the satellite’s design and operation.

Among the frequencies considered, 3.5 GHz is the most challenging for uplink communication due to high path loss, pronounced Doppler effects, and poor building penetration. Satellites operating at this frequency must achieve extraordinary sensitivity and gain, which is difficult to implement at scale. At 1.8 GHz, the challenges are somewhat reduced as the path loss and Doppler effects are less severe. However, the uplink requires advanced receiver sensitivity and high-gain antennas to approach terrestrial network performance. The most favorable scenario is at 800 MHz, where the lower path loss and better penetration characteristics make uplink communication significantly more feasible. Satellites operating at this frequency require less extreme sensitivity and gain, making it a practical choice for achieving parity with terrestrial 5G networks, especially for outdoor and light indoor coverage.

Uplink, the consumer device to satellite signal direction, poses additional limitations to the frequency range. Such systems may be interesting to 600 MHz to a maximum of 1.8 GHz, which is already challenging for uplink and downlink in indoor usage. Service in the lower cellular frequency range is feasible for outdoor usage scenarios in rural and remote areas and for non-challenging indoor environments (e.g., “simple” building topologies).

The premise that LEO satellite D2C services would make terrestrial cellular networks redundant everywhere by offering service parity appears very unlikely, and certainly not with the current generation of LEO satellites being launched. The altitude range of the LEO satellites (300 – 1200 km) and frequency ranges used for most terrestrial cellular services (600 MHz to 5 GHz) make it very challenging and even impractical (for higher cellular frequency ranges) to achieve quality and capacity parity with existing terrestrial cellular networks.

LEO SATELLITE D2C ARCHITECTURE.

A subscriber would realize they have LEO satellite Direct-to-Cell coverage through network signaling and notifications provided by their mobile device and network operator. Using this coverage depends on the integration between the LEO satellite system and the terrestrial cellular network, as well as the subscriber’s device and network settings. Here’s how this process typically works:

When a subscriber moves into an area where traditional terrestrial coverage is unavailable or weak, their mobile device will periodically search for available networks, as it does when trying to maintain connectivity. If the device detects a signal from a LEO satellite providing D2C services, it may indicate “Satellite Coverage” or a similar notification on the device’s screen.

This recognition is possible because the LEO satellite extends the subscriber’s mobile network. The satellite broadcasts system information on the same frequency bands licensed to the subscriber’s terrestrial network operator. The device identifies the network using the Public Land Mobile Network (PLMN) ID, which matches the subscriber’s home network or a partner network in a roaming scenario. The PLMN is a fundamental component of terrestrial and LEO satellite D2C networks, which is the identifier that links a mobile consumer to a specific mobile network operator. It enables communication, access rights management, network interoperability, and supporting services such as voice, text, and data.

The PLMN is also directly connected to the frequency bands used by an operator and any satellite service provider, acting as an extension of the operator’s network. It ensures that devices access the appropriately licensed bands through terrestrial or satellite systems and governs spectrum usage to maintain compliance with regulatory frameworks. Thus, the PLMN links the network identification and frequency allocation, ensuring seamless and lawful operation in terrestrial and satellite contexts.

In an LEO satellite D2C network, the PLMN plays a similar but more complex role, as it must bridge the satellite system with terrestrial mobile networks. The satellite effectively operates as an extension of the terrestrial PLMN, using the same MCC and MNC codes as the consumer’s home network or a roaming partner. This ensures that consumer devices perceive the satellite network as part of their existing subscription, avoiding the need for additional configuration or specialized hardware. When the satellite provides coverage, the PLMN enables the device to authenticate and access services through the operator’s core network, ensuring consistency with terrestrial operations. It ensures that consumer authentication, billing, and service provisioning remain consistent across the terrestrial and satellite domains. In cases where multiple terrestrial operators share access to a satellite system, the PLMN facilitates the correct routing of consumer sessions to their respective home networks. This coordination is particularly important in roaming scenarios, where a consumer connected to a satellite in one region may need to access services through their home network located in another region.

For a subscriber to make use of LEO satellite coverage, the following conditions must be met:

  • Device Compatibility: The subscriber’s mobile device must support satellite connectivity. While many standard devices are compatible with satellite D2C services using terrestrial frequencies, certain features may be required, such as enhanced signal processing or firmware updates. Modern smartphones are increasingly being designed to support these capabilities.
  • Network Integration: The LEO satellite must be integrated with the subscriber’s mobile operator’s core network. This ensures the satellite extends the terrestrial network, maintaining seamless authentication, billing, and service delivery. Consumers can make and receive calls, send texts, or access data services through the satellite link without changing their settings or SIM card.
  • Service Availability: The type of services available over the satellite link depends on the network and satellite capabilities. Initially, services may be limited to text messaging and voice calls, as these require less bandwidth and are easier to support in shared satellite coverage zones. High-speed data services, while possible, may require further advancements in satellite capacity and network integration.
  • Subscription or Permissions: Subscribers must have access to satellite services through their mobile plan. This could be included in their existing plan or offered as an add-on service. In some cases, roaming agreements between the subscriber’s home network and the satellite operator may apply.
  • Emergency Use: In specific scenarios, satellite connectivity may be automatically enabled for emergencies, such as SOS messages, even if the subscriber does not actively use the service for regular communication. This is particularly useful in remote or disaster-affected areas with unavailable terrestrial networks.

Once connected to the satellite, the consumer experience is designed to be seamless. The subscriber can initiate calls, send messages, or access other supported services just as they would under terrestrial coverage. The main differences may include longer latency due to the satellite link and, potentially, lower data speeds or limitations on high-bandwidth activities, depending on the satellite network’s capacity and the number of consumers sharing the satellite beam.

Managing a call on a Direct-to-Cell (D2C) satellite network requires specific mobile network elements in the core network, alongside seamless integration between the satellite provider and the subscriber’s terrestrial network provider. The service’s success depends on how well the satellite system integrates into the terrestrial operator’s architecture, ensuring that standard cellular functions like authentication, session management, and billing are preserved.

In a 5G network, the core network plays a central role in managing calls and data sessions. For a D2C satellite service, key components of the operator’s core network include the Access and Mobility Management Function (AMF), which handles consumer authentication and signaling. The AMF establishes and maintains connectivity for subscribers connecting via the satellite. Additionally, the Session Management Function (SMF) oversees the session context for data services. It ensures compatibility with the IP Multimedia Subsystem (IMS), which manages call control, routing, and handoffs for voice-over-IP communications. The Unified Data Management (UDM) system, another critical core component, stores subscriber profiles, detailing permissions for satellite use, roaming policies, and Quality of Service (QoS) settings.

To enforce network policies and billing, the Policy Control Function (PCF) applies service-level agreements and ensures appropriate charges for satellite usage. For data routing, elements such as the User Plane Function (UPF) direct traffic between the satellite ground stations and the operator’s core network. Additionally, interconnect gateways manage traffic beyond the operator’s network, such as the Internet or another carrier’s network.

The role of the satellite provider in this architecture depends on the integration model. If the satellite system is fully integrated with the terrestrial operator, the satellite primarily acts as an extension of the operator’s radio access network (RAN). In this case, the satellite provider requires ground stations to downlink traffic from the satellites and forward it to the operator’s core network via secure, high-speed connections. The satellite provider handles radio gateway functionality, translating satellite-specific protocols into formats compatible with terrestrial systems. In this scenario, the satellite provider does not need its own core network because the operator’s core handles all call processing, authentication, billing, and session management.

In a standalone model, where the LEO satellite provider operates independently, the satellite system must include its own complete core network. This requires implementing AMF, SMF, UDM, IMS, and UPF, allowing the satellite provider to directly manage subscriber sessions and calls. In this case, interconnect agreements with terrestrial operators would be needed to enable roaming and off-network communication.

Most current D2C solutions, including those proposed by Starlink with T-Mobile or AST SpaceMobile, follow the integrated model. In these cases, the satellite provider relies on the terrestrial operator’s core network, reducing complexity and leveraging existing subscriber management systems. The LEO satellites are primarily responsible for providing RAN functionality and ensuring reliable connectivity to the terrestrial core.

REGULATORY CHALLENGES.

LEO satellite networks offering Direct-to-Cell (D2C) services face substantial regulatory challenges in their efforts to operate within frequency bands already allocated to terrestrial cellular services. These challenges are particularly significant in regions like Europe and the United States, where cellular frequency ranges are tightly regulated and managed by national and regional authorities to ensure interference-free operations and equitable access among service providers.

The cellular frequency spectrum in Europe and the USA is allocated through licensing frameworks that grant exclusive usage rights to mobile network operators (MNOs) for specific frequency bands, often through competitive auctions. For example, in the United States, the Federal Communications Commission (FCC) regulates spectrum usage, while in Europe, national regulatory authorities manage spectrum allocations under the guidelines set by the European Union and CEPT (European Conference of Postal and Telecommunications Administrations). The spectrum currently allocated for cellular services, including low-band (e.g., 600 MHz, 800 MHz), mid-band (e.g., 1.8 GHz, 2.1 GHz), and high-band (e.g., 3.5 GHz), is heavily utilized by terrestrial operators for 4G LTE and 5G networks.

In March 2024, the Federal Communications Commission (FCC) adopted a groundbreaking regulatory framework to facilitate collaborations between satellite operators and terrestrial mobile service providers. This initiative, termed “Supplemental Coverage from Space,” allows satellite operators to use the terrestrial mobile spectrum to offer connectivity directly to consumer handsets and is an essential component of FCC’s “Single Network Future.” The framework aims to enhance coverage, especially in remote and underserved areas, by integrating satellite and terrestrial networks. The FCC granted SpaceX (November 2024) approval to provide direct-to-cell services via its Starlink satellites. This authorization enables SpaceX to partner with mobile carriers, such as T-Mobile, to extend mobile coverage using satellite technology. The approval includes specific conditions to prevent interference with existing services and to ensure compliance with established regulations. Notably, the FCC also granted SpaceX’s request to provide service to cell phones outside the United States. For non-US operations, Starlink must obtain authorization from the relevant governments. Non-US operations are authorized in various sub-bands between 1429 MHz and 2690 MHz.

In Europe, the regulatory framework for D2C services is under active development. The European Conference of Postal and Telecommunications Administrations (CEPT) is exploring the regulatory and technical aspects of satellite-based D2C communications. This includes understanding connectivity requirements and addressing national licensing issues to facilitate the integration of satellite services with existing mobile networks. Additionally, the European Space Agency (ESA) has initiated feasibility studies on Direct-to-Cell connectivity, collaborating with industry partners to assess the potential and challenges of implementing such services across Europe. These studies aim to inform future regulatory decisions and promote innovation in satellite communications.

For LEO satellite operators to offer D2C services in these regulated bands, they would need to reach agreements with the licensed MNOs with the rights to these frequencies. This could take the form of spectrum-sharing agreements or leasing arrangements, wherein the satellite operator obtains permission to use the spectrum for specific purposes, often under strict conditions to avoid interference with terrestrial networks. For example, SpaceX’s collaboration with T-Mobile in the USA involves utilizing T-Mobile’s existing mid-band spectrum (i.e., PCS1900) under a partnership model, enabling satellite-based connectivity without requiring additional spectrum licensing.

In Europe, the situation is more complex due to the fragmented nature of the regulatory environment. Each country manages its spectrum independently, meaning LEO operators must negotiate agreements with individual national MNOs and regulators. This creates significant administrative and logistical hurdles, as the operator must align with diverse licensing conditions, technical requirements, and interference mitigation measures across multiple jurisdictions. Furthermore, any satellite use of the terrestrial spectrum in Europe must comply with European Union directives and ITU (International Telecommunication Union) regulations, prioritizing terrestrial services in these bands.

Interference management is a critical regulatory concern. LEO satellites operating in the same frequency bands as terrestrial networks must implement sophisticated coordination mechanisms to ensure their signals do not disrupt terrestrial operations. This includes dynamic spectrum management, geographic beam shaping, and power control techniques to minimize interference in densely populated areas where terrestrial networks are most active. Regulators in the USA and Europe will likely require detailed technical demonstrations and compliance testing before approving such operations.

Another significant challenge is ensuring equitable access to spectrum resources. MNOs have invested heavily in acquiring and deploying their licensed spectrum, and many may view satellite D2C services as a competitive threat. Regulators would need to establish clear frameworks to balance the rights of terrestrial operators with the potential societal benefits of extending connectivity through satellites, particularly in underserved rural or remote areas.

Beyond regulatory hurdles, LEO satellite operators must collaborate extensively with MNOs to integrate their services effectively. This includes interoperability agreements to ensure seamless handoffs between terrestrial and satellite networks and the development of business models that align incentives for both parties.

TAKEAWAYS.

Ditect-to-cell LEO satellite networks face considerable technology hurdles in providing services comparable to terrestrial cellular networks.

  • Overcoming free-space path loss and ensuring uplink connectivity from low-power mobile devices with omnidirectional antennas.
  • Cellular devices transmit at low power (typically 23–30 dBm), making it difficult for uplink signals to reach satellites in LEO at 500–1,200 km altitudes.
  • Uplink signals from multiple devices within a satellite beam area can overlap, creating interference that challenges the satellite’s ability to separate and process individual uplink signals.
  • Developing advanced phased-array antennas for satellites, dynamic beam management, and low-latency signal processing to maintain service quality.
  • Managing mobility challenges, including seamless handovers between satellites and beams and mitigating Doppler effects due to the high relative velocity of LEO satellites.
  • The high relative velocity of LEO satellites introduces frequency shifts (i.e., Doppler Effect) that the satellite must compensate for dynamically to maintain signal integrity.
  • Address bandwidth limitations and efficiently reuse spectrum while minimizing interference with terrestrial and other satellite networks.
  • Scaling globally may require satellites to carry varied payload configurations to accommodate regional spectrum requirements, increasing technical complexity and deployment expenses.
  • Operating on terrestrial frequencies necessitates dynamic spectrum sharing and interference mitigation strategies, especially in densely populated areas, limiting coverage efficiency and capacity.
  • Ensuring the frequent replacement of LEO satellites due to shorter lifespans increases operational complexity and cost.

On the regulatory front, integrating D2C satellite services into existing mobile ecosystems is complex. Spectrum licensing is a key issue, as satellite operators must either share frequencies already allocated to terrestrial mobile operators or secure dedicated satellite spectrum.

  • Securing access to shared or dedicated spectrum, particularly negotiating with terrestrial operators to use licensed frequencies.
  • Avoiding interference between satellite and terrestrial networks requires detailed agreements and advanced spectrum management techniques.
  • Navigating fragmented regulatory frameworks in Europe, where national licensing requirements vary significantly.
  • Spectrum Fragmentation: With frequency allocations varying significantly across countries and regions, scaling globally requires navigating diverse and complex spectrum licensing agreements, slowing deployment and increasing administrative costs.
  • Complying with evolving international regulations, including those to be defined at the ITU’s WRC-27 conference.
  • Developing clear standards and agreements for roaming and service integration between satellite operators and terrestrial mobile network providers.
  • The high administrative and operational burden of scaling globally diminishes economic benefits, particularly in regions where terrestrial networks already dominate.
  • While satellites excel in rural or remote areas, they might not meet high traffic demands in urban areas, restricting their ability to scale as a comprehensive alternative to terrestrial networks.

The idea of D2C satellite networks making terrestrial cellular networks obsolete is ambitious but fraught with practical limitations. While LEO satellites offer unparalleled reach in remote and underserved areas, they struggle to match terrestrial networks’ capacity, reliability, and low latency in urban and suburban environments. The high density of base stations in terrestrial networks enables them to handle far greater traffic volumes, especially for data-intensive applications.

  • Coverage advantage: Satellites provide global reach, particularly in remote or underserved regions, where terrestrial networks are cost-prohibitive and often of poor quality or altogether lacking.
  • Capacity limitations: Satellites struggle to match the high-density traffic capacity of terrestrial networks, especially in urban areas.
  • Latency challenges: Satellite latency, though improving, cannot yet compete with the ultra-low latency of terrestrial 5G for time-critical applications.
  • Cost concerns: Deploying and maintaining satellite constellations is expensive, and they still depend on terrestrial core infrastructure (although the savings if all terrestrial RAN infrastructure could be avoided is also very substantial).
  • Complementary role: D2C networks are better suited as an extension to terrestrial networks, filling coverage gaps rather than replacing them entirely.

The regulatory and operational constraints surrounding using terrestrial mobile frequencies for D2C services severely limit scalability. This fragmentation makes it difficult to achieve global coverage seamlessly and increases operational and economic inefficiencies. While D2C services hold promise for addressing connectivity gaps in remote areas, their ability to scale as a comprehensive alternative to terrestrial networks is hampered by these challenges. Unless global regulatory harmonization or innovative technical solutions emerge, D2C networks will likely remain a complementary, sub-scale solution rather than a standalone replacement for terrestrial mobile networks.

FURTHER READING.

  1. Kim K. Larsen, “The Next Frontier: LEO Satellites for Internet Services.” Techneconomyblog, (March 2024).
  2. Kim K. Larsen, “Stratospheric Drones & Low Earth Satellites: Revolutionizing Terrestrial Rural Broadband from the Skies?” Techneconomyblog, (January 2024).
  3. Kim K. Larsen, “A Single Network Future“, Techneconomyblog, (March 2024).
  4. T.S. Rappaport, “Wireless Communications – Principles & Practice,” Prentice Hall (1996). In my opinion, it is one of the best graduate textbooks on communications systems. I bought it back in 1999 as a regular hardcover. I have not found it as a Kindle version, but I believe there are sites where a PDF version may be available (e.g., Scribd).

ACKNOWLEDGEMENT.

I greatly acknowledge my wife, Eva Varadi, for her support, patience, and understanding during the creative process of writing this article.

A Single Network Future.

Posted on by
2

How to think about a single network future? What does it entail, and what is it good for?

Well, imagine a world where your mobile device, unchanged and unmodified, connects to the nearest cell tower and satellites orbiting Earth, ensuring customers will always be best connected, getting the best service, irrespective of where they are. Satellite-based supplementary coverage (from space) seeks to deliver on this vision by leveraging superior economic coverage in terms of larger footprint (than feasible with terrestrial networks) and better latency (compared to geostationary satellite solutions) to bring connectivity directly to unmodified consumer handsets (e.g., smartphone, tablet, IoT devices), enhance emergency communication, and foster advancements in space-based technologies. The single network future does not only require certain technological developments, such as 3GPP Non-Terrestrial Network standardization efforts (e.g., Release 17 and forward). We also need the regulatory spectrum policy to change, allowing today’s terrestrially- and regulatory-bounded cellular frequency spectra to be re-used by satellite operators providing the same mobile service under satellite coverage in areas without terrestrial communications infrastructure, as mobile customers enjoy within the normal terrestrial cellular network.

It is estimated that less than 40% of the world’s population, or roughly 2.9 billion people, have never used the internet (as of 2023). That 60% of the world population have access to internet and 40% have not, is the digital divide. A massive gap most pronounced in developing countries, rural & remote areas, and among older populations and economically disadvantaged groups. Most of the 2.9 billion on the wrong side of the divide live in areas lacking terrestrial-based technology infrastructure that would readily facilitate access to the internet. It lacks the communications infrastructure because it may either be impractical or (and) un-economical to deploy, including difficulty in monetizing and yielding a positive return on investment over a relatively short period. Satellites that are allowed by regulatory means to re-use terrestrially-based cellular spectrum for supplementary (to terrestrial) coverage can largely solve the digital divide challenges (as long as affordable mobile devices and services are available to the unconnected).

This blog explores some of the details of the, in my opinion, forward-thinking FCC’s Supplementary Coverage from Space (SCS) framework and vision of a Single Network in which mobile cellular communication is not limited to tera firma but supplemented and enhanced by satellites, ensuring connectivity everywhere.

SUPPLEMENTARY COVERAGE FROM SPACE.

Federal Communications Commission (FCC) recently published a new regulatory framework (“Report & Order and further notice of proposed rulemaking“) designed to facilitate the integration of satellite and terrestrial networks to provide Supplemental Coverage from Space (SCS), marking a significant development toward achieving ubiquitous connectivity. In the following, I will use the terms “SCS framework” and ” SCS initiative” to cover the reference to the FCC’s regulatory framework. The SCS initiative, which, to my knowledge, is the first of its kind globally, aims to allow satellite operators and terrestrial service providers to collaborate, leveraging the spectrum previously allocated exclusively for terrestrial services to extend connectivity directly to consumer handsets, what is called satellite direct-to-device (D2D), especially in remote, unserved, and underserved areas. The proposal is expected to enhance emergency communication availability, foster advancements in space-based technologies, and promote the innovative and efficient use of spectrum resources.

The “Report and Order” formalizes a spectrum-use framework, adopting a secondary mobile-satellite service (MSS) allocation in specific frequency bands devoid of primary non-flexible-use legacy incumbents, both federal and non-federal. Let us break this down in a bit more informal language. So, the FCC proposes to designate certain parts of the radio frequency spectrum (see below) for mobile-satellite services on a “secondary” basis. In spectrum management, an allocation is deemed “secondary” when it allows for the operation of a service without causing interference to the “primary” services in the same band. This means that the supplementary satellite service, deemed secondary, must accept interference from primary services without claiming protection. Moreover, this only applies to locations that lack (i.e., devoid of) the use of a given frequency band by existing ” primary” spectrum users (i.e., incumbents), non-federal as well as federal primary uses.

The setup encourages collaboration and permits supplemental coverage from space (SCS) in designated bands where terrestrial licensees, holding all licenses for a channel throughout a geographically independent area (GIA), lease access to their terrestrial spectrum rights to a satellite operator. Furthermore, the framework establishes entry criteria for satellite operators to apply for or modify an existing “part 25” space station license for SCS operations, that is the regulatory requirements established by the FCC governing the licensing and operation of satellite communications in the United States. The framework also outlines a licensing-by-rule approach for terrestrial devices acting as SCS earth stations, referring to a regulatory and technological framework where conventional consumer devices, such as smartphones or tablets, are equipped to communicate directly with satellites (after all we do talk about Direct-2-Device).

The above picture showcases a moment in the remote Arizona desert where an individual receives a direct signal to the device from a Low-Earth Orbit (LEO) satellite to his or her smartphone. The remote area has no terrestrial cellular coverage, and supplementary coverage from space is the only way for individuals with a subscription to access their cellular services or make a distress call apart from using a costly satellite phone service. It should be remembered that the SCS service is likely to be capacity-limited due to the typical large satellite coverage area and possible limited available SCS spectrum bandwidth.

Additionally, the Further Notice of Proposed Rulemaking seeks further commentary on aspects such as 911 service provision and the protection of radio astronomy, indicating the FCC’s consistent commitment to refining and expanding the SCS framework responsibly. This commitment ensures that the framework will continue to evolve, adapting to new challenges and opportunities and providing a solid foundation for future developments.

BALANCING THE AIRWAVES IN THE USA.

Two agencies in the US manage the frequency spectrum, the Federal Communications Commission (FCC) and the National Telecommunications and Information Administration (NTIA) . They collaboratively manage and coordinate frequency spectrum use and reuse for satellites, among other applications, within the United States. This partnership is important for maintaining a balanced approach to spectrum management that supports federal and non-federal needs, ensuring that satellite communications and other services can operate effectively without causing harmful interference to each other.

The Federal Communications Commission, the FCC for short, is an independent agency that exclusively regulates all non-Federal spectrum use across the United States. FCC allocates spectrum licenses for commercial use, typically through spectrum auctions. A new or re-purposed commercialized spectrum has been reclaimed from other uses, both from federal uses and existing commercial uses. Spectrum can be re-purposed either because newer, more spectrally efficient technologies become available (e.g., the transition from analog to digital broadcasting) or it becomes viable to shift operation to other spectrum bands with less commercial value (and, of course, without jeopardizing existing operational excellence). It is also possible that spectrum, previously having been for exclusive federal use (e.g., military applications, fixed satellite uses, etc.), can be shared, such as the case with Citizens Broadband Radio Service (CBRS), which allows non-federal parties access to 150 MHz in the 3.5 GHz band (i.e., band 48). However, it has recently been concluded that (centralized) dynamic spectrum sharing only works in certain use cases and is associated with considerable implementation complexities. Multiple parties with possible vastly different requirements co-exist within a given band, which is a work in progress and may not be consistent with the commercialized spectrum operation required for high-quality broadband cellular operation.

Alongside the FCC, the National Telecommunications and Information Administration (NTIA) plays a crucial role in US spectrum management. The NTIA is the sole authority responsible for authorizing Federal spectrum use. It also serves as the principal adviser on telecommunications policies to the President of the United States, coordinating the views of the Executive Branch. The NTIA manages a significant portion of the spectrum, approximately 2,398 MHz (69%), within the range of 225 MHz to 3.7 GHz, known as the ‘beachfront spectrum’. Of the total 3,475 MHz, 591 MHz (17%) is exclusively for Federal use, and 1,807 MHz (52%) is shared or coordinated between Federal and non-Federal entities. This leaves 1,077 MHz (31%) for exclusive commercial use, which falls under the management of the FCC.

NTIA, in collaboration with the FCC, has been instrumental in the past in freeing up substantial C-band spectrum, 480 MHz in total, of which 100 MHz is conditioned on prioritized sharing (i.e., Auction 105), for commercial and shared use that subsequently has been auctioned off over the last three years raising USD 109 billion. In US Dollar (USD) per MHz per population count (pop), we have, on average, ca. USD 0.68 per MHz-pop from the C-band auctions in the US, compared to USD 0.13 per MHz-pop in Europe C-band auctions and USD 0.23 per MHz-pop in APAC auctions. It should be remembered that the United States exclusive-use spectrum licenses can be regarded as an indefinite-lived intangible asset, while European spectrum rights expire between 10 and 20 years. This may explain a big part of the difference between US-based spectrum pricing and Europe and Asia.

The FCC and the NTIA jointly manage all the radio spectrum in the United States, licensed (e.g., cellular mobile frequencies, TV signals) and unlicensed (e.g., WiFi, MW Owens). The NTIA oversees spectrum use for Federal purposes, while the FCC is responsible for non-Federal use. In addition to its role in auctioning spectrum licenses, the FCC is also authorized to redistribute licenses. This authority allows the FCC to play a vital role in ensuring efficient spectrum use and adapting to changing needs.

THE SINGLE NETWORK.

The Supplementary Coverage from Space (SCS) framework creates an enabling regulatory framework for satellite operators to provide mobile broadband services to unmodified mobile devices (i.e., D2D services), such as smartphones and other terrestrial cellular devices, in rural and remote areas without such services, where no or only scarce terrestrial infrastructure exists. By leveraging SCS, terrestrial cellular broadband services will be enhanced, and the combination may result in a unified network. This network will ensure continuous and ubiquitous access to communication services, overcoming geographical and environmental challenges. Thus, this led to the inception of the Single Network that can provide seamless connectivity across diverse environments, including remote, unserved, and underserved areas.

The above picture illustrates the idea behind the FCC’s SCS framework and “Single Network” on a high level. In this example, an LEO satellite provides direct-to-device (D2D) supplementary coverage in rural and remote areas, using an advanced phase-array antenna, to unmodified user equipment (e.g., smartphone, tablet, cellular-IoT, …) in the same frequency band (i.e., f1,sat) owned and used by a terrestrial operator operating a cellular network (f1). The LEO satellite operator must partner with the terrestrial spectrum owner to manage and coordinate the frequency re-use in areas where the frequency owner (i.e., mobile/cellular operator) does not have the terrestrial-based infrastructure to deliver a service to its customers (i.e., typically remote, rural areas where terrestrial infrastructure is impractical and uneconomic to deploy). The satellite operator has to avoid geographical regions where the frequency (e.g., f1) is used by the spectrum owner, typically in urban, suburban, and rural areas (where terrestrial cellular infrastructure has already been deployed and service offered).

How does the “Single Network” of FCC differ from the 3GPP Non-Terrestrial Network (NTN) standardization? Simply put, the “Single Network” is a regulatory framework that paves the way for satellite operators to re-use the terrestrial cellular spectrum on their non-terrestrial (satellite-based) network. The 3GPP NTN standardization initiatives, e.g., Release 16, 17 and 18+, are a technical effort to incorporate satellite communication systems within the 5G network architecture. Shortly, the following 3GPP releases are it relates to how NTN should function with terrestrial 5G networks;

  • Release 15 laid the groundwork for 5G New Radio (NR) and started to consider the broader picture of integrating non-terrestrial networks with terrestrial 5G networks. It marks the beginning of discussions on how to accommodate NTNs within the 5G framework, focusing on study items rather than specific NTN standards.
  • Release 16 took significant steps toward defining NTN by including study items and work items specifically aimed at understanding and specifying the adjustments needed for NR to support communication with devices served by NTNs. Release 16 focuses on identifying modifications to the NR protocol and architecture to accommodate the unique characteristics of satellite communication, such as higher latency and different mobility characteristics compared to terrestrial networks.
  • Release 17 further advancements in NTN specifications aiming to integrate specific technical solutions and standards for NTNs within the 5G architecture. This effort includes detailed specifications for supporting direct connectivity between 5G devices and satellites, covering aspects like signal timing, frequency bands, and protocol adaptations to handle the distinct challenges posed by satellite communication, such as the Doppler effect and signal delay.
  • Release 18 and beyond will continue to evolve its standards to enhance NTN support, addressing emerging requirements and incorporating feedback from early implementations. These efforts include refining and expanding NTN capabilities to support a broader range of applications and services, improving integration with terrestrial networks, and enhancing performance and reliability.

The NTN architecture ensures (should ensure) that satellite communications systems can seamlessly integrate into 5G networks, supporting direct communication between satellites and standard mobile devices. This integration idea includes adapting 5G protocols and technologies to accommodate the unique characteristics of satellite communication, such as higher latency and different signal propagation conditions. The NTN standardization aims to expand the reach of 5G services to global scales, including maritime, aerial, and sparsely populated land areas, thereby aligning with the broader goal of universal service coverage.

The FCC’s vision of a “single network” and the 3GPP NTN standardization aims to integrate satellite and terrestrial networks to extend connectivity, albeit from slightly different angles. The FCC’s concept provides a regulatory and policy framework to enable such integration across different network types and service providers, focusing on the broad goal of universal connectivity. In contrast, 3GPP’s NTN standardization provides the technical specifications and protocols to make this integration possible, particularly within next-generation (5G) networks. At the same time, 3GPP’s NTN efforts address the technical underpinnings required to realize that vision in practice, especially for 5G technologies. The FCC’s “single network” concept lays the regulatory foundation for enabling satellite and terrestrial cellular network service integration to the same unmodified device portfolio. Together, they are highly synergistic, addressing the regulatory and technical challenges of creating a seamlessly connected world.

Depicting a moment in the Colorado mountains, a hiker receives a direct signal from a Low Earth Orbit (LEO) satellite supplementary coverage to their (unmodified) smartphone. The remote area has no terrestrial cellular coverage. It should be remembered that the SCS service is likely to be capacity-limited due to the typical large satellite coverage area and possible limited available SCS spectrum bandwidth.

SINGLE NETWORK VS SATELLITE ATC

The FCC’s Single Network vision and the Supplemental Coverage from Space (SCS) concept, akin to the Satellite Ancillary Terrestrial Component (ATC) architectural concept (an area that I spend a significant portion of my career working on operationalizing and then defending … a different story though), share a common goal of merging satellite and terrestrial networks to fortify connectivity. These strategies, driven by the desire to enhance the reach and reliability of communication services, particularly in underserved regions, hold the promise of expanded service coverage.

The Single Network and SCS initiatives broadly focus on comprehensively integrating satellite services with terrestrial infrastructures, aiming to directly connect satellite systems with standard consumer devices across various services and frequency bands. This expansive approach seeks to ensure ubiquitous connectivity, significantly closing the coverage gaps in current network deployments. Conversely, the Satellite ATC concept is more narrowly tailored, concentrating on using terrestrial base stations to complement and enhance satellite mobile services. This method explicitly addresses the need for improved signal availability and service reliability in urban or obstructed areas by integrating terrestrial components within the satellite network framework.

Although the Single Network and Satellite ATC shared goals, the paths to achieving them diverge significantly in the application, regulatory considerations, and technical execution. The SCS concept, for instance, involves navigating regulatory challenges associated with direct-to-device satellite communications, including the complexities of spectrum sharing and ensuring the harmonious coexistence of satellite and terrestrial services. This highlights the intricate nature of network integration, making your audience more aware of the regulatory and technical hurdles in this field.

The distinction between the two concepts lies in their technological and implementation specifics, regulatory backdrop, and focus areas. While both aim to weave together the strengths of satellite and terrestrial technologies, the Single Network and SCS framework envisions a more holistic integration of connectivity solutions, contrasting with the ATC’s targeted approach to augmenting satellite services with terrestrial network support. This illustrates the evolving landscape of communication networks, where the convergence of diverse technologies opens new avenues for achieving seamless and widespread connectivity.

THE RELATED SCS FREQUENCIES & SPECTRUM.

The following frequency bands and the total bandwidth associated with the frequency have by the FCC been designated for Supplemental Coverage from Space (SCS):

  • 70MHz @ 600 MHz Band
  • 96 MHz @ 700 MHz Band
  • 50 MHz @ 800 MHz Band
  • 130 MHz @ Broadband PCS
  • 10 MHz @ AWS-H Block

The above comprises a total frequency bandwidth of 350+ MHz, currently used for terrestrial cellular services across the USA. According to the FCC, the above frequency bands and spectrum can also be used for satellite direct-to-device SCS services to normal mobile devices without built-in satellite transceiver functionality. Of course, this is subject to spectrum owners’ approval and contractual and commercial arrangements.

Moreover, the 758-769/788-799 MHz band, licensed to the First Responder Network Authority (FirstNet), is also eligible for SCS under the established framework. This frequency band has been selected to enhance connectivity in remote, unserved, and underserved areas by facilitating collaborations between satellite and terrestrial networks within these specific frequency ranges.

SpaceX recently reported a peak download speed of 17 Mb/s from a satellite direct to an unmodified Samsung Android Phone using 2×5 MHz of T-Mobile USA’s PCS (i.e., the G-block). The speed corresponds to a downlink spectral efficiency of ~3.4 Mbps/MHz/beam, which is pretty impressive. Using this as rough guidance for the ~350 MHz, we should expect this to be equivalent to an approximate download speed of ca. 600 Mbps (@ 175 MHz) per satellite beam. As the satellite antenna technology improves, we should expect that spectral efficiency will also increase, resulting in increasing downlink throughput.

SCS INFANCY, BUT ALIVE AND KICKING.

In the FCC’s framework on the Supplemental Coverage from Space (SCS), the partnership between SpaceX and T-Mobile is described as a collaborative effort where SpaceX would utilize a block of T-Mobile’s mid-band Personal Communications Services (PCS G-Block) spectrum across a nationwide footprint. This initiative aims to provide service to T-Mobile’s subscribers in rural and remote locations, thereby addressing coverage gaps in T-Mobile’s terrestrial network. The FCC has facilitated this collaboration by allowing SpaceX and T-Mobile to deploy and test their proposed SCS system while their pending applications and the FCC’s proceedings continue.

Specifically, SpaceX has been authorized (by FCC’s Space Bureau) to deploy a modified version of its second-generation (2nd generation) Starlink satellites with SCS-capable antennas that can operate in specific frequencies. FCC authorized experimental testing on terrestrial locations for SpaceX and T-Mobile to progress with their SCS system, although SpaceX’s requests for broader authority remain under consideration by the FCC.

Lynk Global has partnered with mobile network operators (MNOs) outside the United States to allow the MNOs’ customers to send texts using Lynk’s satellite network. In 2022, the FCC authorized Lynk’s request to operate a non-geostationary satellite orbit (NGSO) satellite system (e.g., Low-Earth Orbit, Medium Earth Orbit, or Highly-Elliptical Orbit) intended for text message communications in locations outside the United States and in countries where Lynk has obtained agreements with MNOs and the required local regulatory approval. Lynk aims to deploy ten mobile-satellite service (MSS) satellites as part of a “cellular-based satellite communications network” operating on cellular frequencies globally in the 617-960 MHz band (i.e., within the UHF band), targeting international markets only.

Lynk has announced contracts with more than 30 MNOs (full list not published) covering over 50 countries for Lynk’s “satellite-direct-to-standard-mobile-phone-system,” which provides emergency alerts and two-way Short Message Service (SMS) messaging. Lynk currently has three LEO satellites in orbit as of March 2023, and they plan to expand their constellation to include up to 5,000 satellites with 50 additional satellites planned for end of 2024, and with that substantially broadening its geographic coverage and service capabilities​​. Lynk recently claimed that they had in Hawaii achieved repeated successful downlink speeds above 10 Mbps with several mass market unmodified smartphones (10+ Mbps indicates a spectral efficiency of 2+ Mbps/MHz/beam). Lynk Mobile has also, recently (July 2023) demonstrated (as a proof of concept) phone calls via their LEO satellite between two unmodified smartphones (see the YouTube link).

AST SpaceMobile is also mentioned for its partnerships with several MNOs, including AT&T and Vodafone, to develop its direct-to-device or satellite-to-smartphone service. Overall AST SpaceMobile has announced it has entered into “more than 40 agreements and understandings with mobile network operators globally” (e.g., AT&T, Vodafone, Rakuten, Orange, Telefonica, TIM, MTN, Ooredoo, …). In 2020, AST filed applications with the FCC seeking U.S. market access for gateway links in the V-band for its SpaceMobile satellite system, which is planned to consist of 243 LEO satellites. AST clarified that its operation in the United States would collaborate with terrestrial licensee partners without seeking to operate independently on terrestrial frequencies​​.

AST SpaceMobile BlueWalker 3 (BW3) LEO satellite 64 square-meter phased array. Source: AST SpaceMobile.

AST SpaceMobile’s satellite antenna design marks a pioneering step in satellite communications. AST recently deployed the largest commercial phased array antenna into Low Earth Orbit (LEO). On September 10, 2022, AST SpaceMobile launched its prototype direct-to-device testbed BlueWalker 3 (BW3) satellite. This mission marked a significant step forward in the company’s efforts to test and validate its technology for providing direct-to-cellphone communication via a Low Earth Orbit (LEO) satellite network. The launch of BW3 aimed to demonstrate the capabilities of its large phased array antenna, a critical component for the AST’s targeted global broadband service.

The BW3’s phased array antenna with a surface area of 64 square meters is technologically quite advanced (actually, I find it very beautiful and can’t wait to see the real thing for their commercial constellation) and designed for dynamic beamforming as one would expect for a state-of-art direct-to-device satellite. The BlueWalker 3, a proof of concept design, supports a frequency range of 100 MHz in the UHF band, with 5 MHz channels and a spectral efficiency expected to be 3 Mbps/MHz/channel. This capability is crucial for establishing direct-to-device communications, as it allows the satellite to concentrate its signals on specific geographic areas or directly on mobile devices, enhancing the quality of coverage and minimizing potential interference with terrestrial networks. AST SpaceMobile is expected to launch the first 5 of 243 LEO satellites, BlueBirds, on SpaceX’s Falcon 9 in the 2nd quarter of 2024. The first 5 will be similar to BW3 design including the phased array antenna. Subsequent AST satellites are expected to be larger with substantially up-scaled phased array antenna supporting an even larger frequency span covering the most of the UHF band and supporting 40 MHz channels with peak download speeds of 120 Mbps (using their estimated 3 Mbps/MHz/channel).

These above examples underscore the the ongoing efforts and potential of satellite service providers like Starlink/SpaceX, Lynk Global, and AST SpaceMobile within the evolving SCS framework. The examples highlight the collaborative approach between satellite operators and terrestrial service providers to achieve ubiquitous connectivity directly to unmodified cellular consumer handsets.

PRACTICAL PREREQUISITES.

In general, the satellite operator would need a terrestrial frequency license owner willing to lease out its spectrum for services in areas where that spectrum has not been deployed on its network infrastructure or where the license holder has no infrastructure deployed. And, of course, a terrestrial communication service provider owning spectrum and interested in extending services to remote areas would need a satellite operator to provide direct-to-device services to its customers. Eventually, terrestrial operators might see an economic benefit in decommissioning uneconomical rural terrestrial infrastructure and providing satellite broadband cellular services instead. This may be particularly interesting in low-density rural and remote areas supported today by a terrestrial communications infrastructure.

Under the SCS framework, terrestrial spectrum owners can make leasing arrangements with satellite operators. These agreements would allow satellite services to utilize the terrestrial cellular spectrum for direct satellite communication with devices, effectively filling coverage gaps with satellite signals. This kind of arrangement could be similar to the one between T-Mobile USA and StarLink to offer cellular services in the absence of T-Mobile cellular infrastructure, e.g., mainly remote and rural areas.

As the regulatory body for non-federal frequencies, the FCC delineates a regulatory environment that specifies the conditions under which the spectrum can be shared or used by terrestrial and satellite services, minimizing the risk of harmful interference (which both parties should be interested in anyway). This includes setting technical standards and identifying suitable frequency bands supporting dual use. The overarching goal is to bolster the reach and reliability of cellular networks in remote areas, enhancing service availability.

For terrestrial cellular networks and spectrum owners, this means adhering to FCC regulations that govern these new leasing arrangements and the technical criteria designed to protect incumbent services from interference. The process involves meticulous planning and, if necessary, implementing measures to mitigate interference, ensuring that the integration of satellite and terrestrial networks proceeds smoothly.

Moreover, the SCS framework should leapfrog innovation and allow network operators to broaden their service offerings into areas where they are not present today. This could include new applications, from emergency communications facilitated by satellite connectivity to IoT deployments and broadband access in underserved locations.

Depicting a moment somewhere in the Arctic (e.g., Greenland), an eco-tourist receives a direct signal from a Low Earth Orbit (LEO) satellite supplementary coverage to their (unmodified) smartphone. The remote area has no terrestrial cellular coverage. It should be remembered that the SCS service is likely to be capacity-limited due to the typical large satellite coverage area and possible limited available SCS spectrum bandwidth. Several regulatory, business, and operational details must be in place for the above service to work.

TECHNICAL PREREQUISITES FOR DELIVERING SATELLITE SCS SERVICES.

Satellite constellations providing D2D services are naturally targeting supplementary coverage of geographical areas where no terrestrial cellular services are present at the target frequency bands used by the satellite operator.

As the satellite operator has gotten access to the terrestrial cellular spectrum for its supplementary coverage direct-to-device service, it has a range of satellite technical requirements that either need to be in place of an existing constellation (though that might require some degree of foresight) or a new satellite would need to be designed consistent with frequency band and range, the targeted radio access technology such as LTE or 5G (assuming the ambition eventually is beyond messaging), and the device portfolio that the service aims to support (e.g., smartphone, tablet, IoTs, …). In general, I would assume that existing satellite constellations would not automatically support SCS services they have not been designed for upfront. It would make sense (economically) if a spectrum arrangement already exists between the satellite and terrestrial cellular spectrum owner and operator.

Direct-to-device LEO satellites directly connect to unmodified mobile devices such as smartphones, tablets, or other personal devices. This necessitates a design that can accommodate low-power signals and small antennas typically found on consumer devices. Therefore, these satellites often incorporate advanced beamforming capabilities through phased array antennas to focus signals precisely on specific geographic locations, enhancing signal strength and reliability for individual users. Moreover, the transceiver electronics must be highly sensitive and capable of handling simultaneous connections, each potentially requiring different levels of service quality. As the satellite provides services over remote and scarcely populated areas, at least initially, there is no need for high-capacity designs, e.g., typically requiring terrestrial cellular-like coverage areas and large frequency bandwidths. The satellites are designed to operate in frequency bands compatible with terrestrial consumer devices, necessitating coordination and compliance with various regulatory standards compared to traditional satellite services.

Implementing satellite-based SCS successfully hinges on complying with many fairly sophisticated technical requirements, such as phased array antenna design and transceiver electronics, enabling direct communication with consumer devices terrestrially. The phased array antenna, a cornerstone of this architecture, must possess advanced beamforming capabilities, allowing it to dynamically focus and steer its signal beams towards specific geographic areas or even moving targets on the Earth’s surface. This flexibility is super important for maximizing the coverage and quality of the communication link with individual devices, which might be spread across diverse and often challenging terrains. The antenna design needs to be wideband and highly efficient to handle the broad spectrum of frequencies designated for SCS operations, ensuring compatibility with the communication standards used by consumer devices (e.g., 4G LTE, 5G).

An illustration of a LEO satellite with a phased array antenna providing direct to smartphone connectivity at a 850 MHz carrier frequency. All practical purposes the antenna beamforming at a LEO altitude can be considered far-field. Thus the electromagnetic fields behave as planar waves and the antenna array becomes more straightforward to design and to manage performance (e.g., beam steering at very high accuracy).

Designing phased array antennas for satellite-based direct-to-device services, envisioned by the SCS framework, requires considering various technical design parameters to ensure the system’s optimal performance and efficiency. These antennas are crucial for effective direct-to-device communication, encompassing multiple technical and practical considerations.

The SCS frequency band not only determines the operational range of the antenna but also its ability to communicate effectively with ground-based devices through the Earth’s atmosphere; in this respect, lower frequencies are better than higher frequencies. The frequency, or frequencies, significantly influences the overall design of the antenna, affecting everything from its physical dimensions to the materials used in its construction. The spacing and configuration of the antenna elements are carefully planned to prevent interference while maximizing coverage and connectivity efficiency. Typically, element spacing is kept around half the operating frequency wavelength, and the configuration involves choosing linear, planar, or circular arrays.

Beamforming capabilities are at the heart of the phased array design, allowing for the precise direction of communication beams toward targeted areas on the ground. This necessitates advanced signal processing to adjust signal phases dynamically and amplitudes, enabling the system to focus its beams, compensate for the satellite’s movement, and handle numerous connections.

The antenna’s polarization strategy is chosen to enhance signal reception and minimize interference. Dual (e.g., horizontal & vertical) or circular (e.g., right or left hand) polarization ensures compatibility with a wide range of devices and as well as more efficient spectrum use. Polarization refers to the orientation of the electromagnetic waves transmitted or received by an antenna. In satellite communications, polarization is used to differentiate between signals and increase the capacity of the communication link without requiring additional frequency bandwidth.

Physical constraints of size, weight, and form factor are also critical, dictated by the satellite’s design and launch parameters, including the launch cost. The antenna must be compact and lightweight to fit within the satellite’s structure and comply with launch weight limitations, impacting the satellite’s overall design and deployment mechanisms.

Beyond the antenna, the transceiver electronics within the satellite play an important role. These must be capable of handling high-throughput data to accommodate simultaneous connections, each demanding reliable and quality service. Sensitivity is another critical factor, as the electronics need to detect and process the relatively weak signals sent by consumer-grade devices, which possess much less power than traditional ground stations. Moreover, given the energy constraints inherent in satellite platforms, these transceiver systems must efficiently manage the power to maintain optimal operation over long durations as it directly relates to the satellite’s life span.

Operational success also depends on the satellite’s compliance with regulatory standards, particularly frequency use and signal interference. Achieving this requires a deep integration of technology and regulatory strategy, ensuring that the satellite’s operations do not disrupt existing services and align with global communication protocols.

CONCERNS.

The FCC’s Supplemental Coverage from Space (SCS) framework has been met with both anticipation and critique, reflecting diverse stakeholder interests and concerns. While the framework aims to enhance connectivity by integrating satellite and terrestrial networks, several critiques and concerns have been raised:

Interference concerns: A primary critique revolves around potential interference with existing terrestrial services. Stakeholders worry that SCS operations might disrupt the current users, including terrestrial mobile networks and other satellite services. A significant challenge is ensuring that SCS services coexist harmoniously with these incumbent services without causing harmful interference.

Certification of terrestrial mobile devices: FCC requires that terrestrial mobile devices has to be certified SCS. The expressed concerns have been multifaceted, reflecting the complexities of integrating satellite communication capabilities into standard consumer mobile devices. These concerns, as in particular highlighted in the FCC’s SCS framework, revolving around technical, regulatory, and practical aspects. As 3GPP NTN standardization are considering changes to mobile devices that would enhance the direct connectivity between device and satellite, it may at least for devices developed for NTN communication make sense to certify those.

Spectrum allocation and management: Spectrum allocation for SCS poses another concern, particularly the repurposing of spectrum bands previously dedicated to other uses. Critics argue that spectrum reallocation must be carefully managed to avoid disadvantaging existing services or limiting future innovation in those bands.

Regulatory and licensing framework: The complexity of the regulatory and licensing framework for SCS services has also been a point of contention. Critics suggest that the framework could be burdensome for new entrants or more minor players, potentially stifling innovation and competition in the satellite and telecommunications industries.

Technical and operational challenges: The technical requirements for SCS, including the need for advanced phased array antennas and the integration of satellite systems with terrestrial networks, pose significant challenges. Concerns about the feasibility and cost of developing and deploying the necessary technology at scale have been raised.

Market and economic impacts: There are concerns about the SCS framework’s economic implications, particularly its impact on existing market dynamics. Critics worry that the framework might favor certain players or technologies, potentially leading to market consolidation or barriers to entry for innovative solutions.

Environmental and space traffic management: The increased deployment of satellites for SCS services raises concerns about space debris and the sustainability of space activities. Critics emphasize the need for robust space traffic management and debris mitigation strategies to ensure the long-term viability of space operations.

Global coordination and equity: The global nature of satellite communications underscores the need for international coordination and equitable access to SCS services. Critics point out the importance of ensuring that the benefits of SCS extend to all regions, particularly those currently underserved by telecommunications infrastructure.

FURTHER READING.

  1. FCC-CIRC2403-03, Report and Order and further notice of proposed rulemaking, related to the following context: “Single Network Future: Supplemental Coverage from Space” (February 2024).
  2. A. Vanelli-Coralli, N. Chuberre, G. Masini, A. Guidotti, M. El Jaafari, “5G Non-Terrestrial Networks.”, Wiley (2024). A recommended reading for deep diving into NTN networks of satellites, typically the LEO kind, and High-Altitude Platform Systems (HAPS) such as stratospheric drones.
  3. Kim Kyllesbech Larsen, The Next Frontier: LEO Satellites for Internet Services. | techneconomyblog, (March 2024).
  4. Kim Kyllesbech Larsen, Stratospheric Drones: Revolutionizing Terrestrial Rural Broadband from the Skies? | techneconomyblog, (January 2024).
  5. Kim Kyllesbech Larsen, Spectrum in the USA – An overview of Today and a new Tomorrow. | techneconomyblog, (May 2023).
  6. Starlink, “Starlink specifications” (Starlink.com page). The following Wikipedia resource is also quite good: Starlink.
  7. R.K. Mailloux, “Phased Array Antenna Handbook, 3rd Edition”, Artech House, (September 2017).
  8. Professor Emil Björnson, “Basics of Antennas and Beamforming”, (2019). Provides a high-level understand of what beamforming is in relative simple terms.
  9. Professor Emil Björnson, “Physically Large Antenna Arrays: When the Near-Field Becomes Far-Reaching”, (2022). Provides a high-level understand of what phased array and their working in relative simple terms with lots of simply illustrations. I also recommend to check Prof. Björnson’s “Reconfigurable intelligent surfaces: Myths and realities” (2020).
  10. AST SpaceMobile website: https://ast-science.com/ Constellation Areas: Internet, Direct-to-Cell, Space-Based Cellular Broadband, Satellite-to-Cellphone. 243 LEO satellites planned. 2 launched.
  11. Jon Brodkin, “Google and AT&T invest in Starlink rival for satellite-to-smartphone service”, Ars Technica (January 2024). There is a very nice picture of AST’s 64 square meter large BlueWalker 3 phased array antenna (i.e., with a total supporting bandwidth of 100 MHz with a channels of 5 MHz and a theoretical spectral efficiency of 3 Mbps/MHz/channel).
  12. Lynk Global website: https://lynk.world/ (see also FCC Order and Authorization). It should be noted that Lynk can operate within 617 to 960 MHz (Space-to-Earth) and 663 to 915 MHz (Earth-to-Space). However, only outside the USA. Constellation Area: IoT / M2M, Satellite-to-Cellphone, Internet, Direct-to-Cell. 8 LEO satellites out of 10 planned.
  13. NewSpace Index: https://www.newspace.im/ I find this resource to have excellent and up-to-date information on commercial satellite constellations.
  14. Up-to-date rocket launch schedule and launch details can be found here: https://www.rocketlaunch.live/

ACKNOWLEDGEMENT.

I greatly acknowledge my wife, Eva Varadi, for her support, patience, and understanding during the creative process of writing this article.

The Next Frontier: LEO Satellites for Internet Services.

Posted on by
5

THE SPACE RACE IS ON.

If all current commercial satellite plans were to be realized within the next decade, we would have more, possibly substantially more, than 65 thousand satellites circling Earth. Today, that number is less than 10 thousand, with more than half that number realized by StarLink’s Low Earth Orbit (LEO) constellation over the last couple of years (i.e., since 2018).

While the “Arms Race” during the Cold War was “a thing” mainly between The USA and the former Soviet Union, the Space Race will, in my opinion, be “battled out” between the commercial interests of the West against the political interest of China (as illustrated in Figure 1 below). The current numbers strongly indicate that Europe, Canada, the Middle East, Africa, and APAC (minus China) will likely and largely be left on the sideline to watch the US and China impose, in theory, a “duopoly” in LEO satellite-based services. However, in practice, it will be a near-monopoly when considering security concerns between the West and the (re-defined) East block.

Figure 1 Illustrates my thesis that we will see a Space Race over the next 10 years between a (or very few) commercial LEO constellation, represented by a Falcon-9 like design (for maybe too obvious reasons), and a Chinese-state owned satellite constellation. (Courtesy: DALL-E).

As of end of 2023, more than 50% of launched and planned commercial LEO satellites are USA-based. Of those, the largest fraction is accounted for by the US-based StarLink constellation (~75%). More than 30% are launched or planned by Chinese companies headed by the state-owned Guo Wang constellation rivaling Elon Musk’s Starlink in ambition and scale. Europe comes in at a distant number 3 with about 8% of the total of fixed internet satellites. Apart from being disappointed, alas, not surprised by the European track record, it is somewhat more baffling that there are so few Indian and African satellite (there are none) constellations given the obvious benefits such satellites could bring to India and the African continent.

India is a leading satellite nation with a proud tradition of innovative satellite designs and manufacturing and a solid track record of satellite launches. However, regarding commercial LEO constellations, India still needs to catch up on some opportunities here. Having previously worked on the economics and operationalizing a satellite ATC (i.e., a satellite service with an ancillary terrestrial component) internet service across India, it is mind-blowing (imo) how much economic opportunity there is to replace by satellite the vast terrestrial cellular infrastructure in rural India. Not to mention a quantum leap in communication broadband services resilience and availability that could be provided. According to the StarLink coverage map, the regulatory approval in India for allowing StarLink (US) services is still pending. In the meantime, Eutelsat’s OneWeb (EU) received regulatory approval in late 2023 for its satellite internet service over India in collaboration with Barthi Enterprises (India), that is also the largest shareholder in the recently formed Eutelsat Group with 21.2%. Moreover, Jio’s JioSpaceFiber satellite internet services were launched in several Indian states at the end of 2023, using the SES (EU) MEO O3b mPower satellite constellation. Despite the clear satellite know-how and capital available, it appears there is little activity for Indian-based LEO satellite development, taking up the competition with international operators.

The African continent is attracting all the major LEO satellite constellations such as StarLink (US), OneWeb (EU), Amazon Kuipers (US), and Telesat Lightspeed (CAN). However, getting regulatory approval for their satellite-based internet services is a complex, time-consuming, and challenging process with Africa’s 54 recognized sovereign countries. I would expect that we will see the Chinese-based satellite constellations (e.g., Guo Wang) taking up here as well due to the strong ties between China and several of the African nations.

This article is not about SpaceX’s StarLink satellite constellation. Although StarLink is mentioned a lot and used as an example. Recently, at the Mobile World Congress 2024 in Barcelona, talking to satellite operators (but not StarLink) providing fixed broadband satellite services, we joked about how long into a meeting we could go before SpaceX and StarLink would be mentioned (~ 5 minutes where the record, I think).

This article is about the key enablers (frequencies, frequency bandwidth, antenna design, …) that make up an LEO satellite service, the LEO satellite itself, the kind of services one should expect from it, and its limitations.

There is no doubt that LEO satellites of today have an essential mission: delivering broadband internet to rural and remote areas with little or no terrestrial cellular or fixed infrastructure to provide internet services. Satellites can offer broadband internet to remote areas with little population density and a population spread out reasonably uniformly over a large area. A LEO satellite constellation is not (in general) a substitute for an existing terrestrial communications infrastructure. Still, it can enhance it by increasing service availability and being an important remedy for business continuity in remote rural areas. Satellite systems are capacity-limited as they serve vast areas, typically with limited spectral resources and capacity per unit area.

In comparison, we have much smaller coverage areas with demand-matched spectral resources in a terrestrial cellular network. It is also easier to increase capacity in a terrestrial cellular system by adding more sectors or increasing the number of sites in an area that requires such investments. Adding more cells, and thus increasing the system capacity, to satellite coverage requires a new generation of satellites with more advanced antenna designs, typically by increasing the number of phased-array beams and more complex modulation and coding mechanisms that boost the spectral efficiency, leading to increased capacity and quality for the services rendered to the ground. Increasing the system capacity of a cellular communications system by increasing the number of cells (i.e., cell splitting) works the same in satellite systems as it does for a terrestrial cellular system.

So, on average, LEO satellite internet services to individual customers (or households), such as those offered by StarLink, are excellent for remote, lowly populated areas with a nicely spread-out population. If we de-average this statement. Clearly, within the satellite coverage area, we may have towns and settlements where, locally, the population density can be fairly large despite being very small over the larger footprint covered by the satellite. As the capacity and quality of the satellite is a shared resource, serving towns and settlements of a certain size may not be the best approach to providing a sustainable and good customer experience as the satellite resources exhaust rapidly in such scenarios. In such scenarios, a hybrid architecture is of much better use as well as providing all customers in a town or settlement with the best service possible leveraging the existing terrestrial communications infrastructure, cellular as well as fixed, with that of a satellite backhaul broadband connection between a satellite ground gateway and the broadband internet satellite. This is offered by several satellite broadband providers (both from GEO, MEO and LEO orbits) and has the beauty of not only being limited to one provider. Unfortunately, this particular finesse, is often overlooked by the awe of massive scale of the StarLink constellation.

AND SO IT STARTS.

When I compared the economics of stratospheric drone-based cellular coverage with that of LEO satellites and terrestrial-based cellular networks in my previous article, “Stratospheric Drones: Revolutionizing Terrestrial Rural Broadband from the Skies?”, it was clear that even if LEO satellites are costly to establish, they provide a substantial cost advantage over cellular coverage in rural and remote areas that are either scarcely covered or not at all. Although the existing LEO satellite constellations have limited capacity compared to a terrestrial cellular network and would perform rather poorly over densely populated areas (e.g., urban and suburban areas), they can offer very decent fixed-wireless-access-like broadband services in rural and remote areas at speeds exceeding even 100 Mbps, such as shown by the Starlink constellation. Even if the provided speed and capacity is likely be substantially lower than what a terrestrial cellular network could offer, it often provides the missing (internet) link. Anything larger than nothing remains infinitely better.

Low Earth Orbit (LEO) satellites represent the next frontier in (novel) communication network architectures, what we in modern lingo would call non-terrestrial networks (NTN), with the ability to combine both mobile and fixed broadband services, enhancing and substituting terrestrial networks. The LEO satellites orbit significantly closer to Earth than their Geostationary Orbit (GEO) counterparts at 36 thousand kilometers, typically at altitudes between 300 to 2,000 kilometers, LEO satellites offer substantially reduced latency, higher bandwidth capabilities, and a more direct line of sight to receivers on the ground. It makes LEO satellites an obvious and integral component of non-terrestrial networks, which aim to extend the reach of existing fixed and mobile broadband services, particularly in rural, un-and under-served, or inaccessible regions as a high-availability element of terrestrial communications networks in the event of natural disasters (flooding, earthquake, …), or military conflict, in which the terrestrial networks are taken out of operation.

Another key advantage of LEO satellite is that the likelihood of a line-of-sight (LoS) to a point on the ground is very high compared to establishing a LoS for terrestrial cellular coverage that, in general, would be very low. In other words, the signal propagation from a LEO satellite closely approximates that of free space. Thus, all the various environmental signal loss factors we must consider for a standard terrestrial-based cellular mobile network do not apply to our satellite with signal propagation largely being determined by the distance between the satellite and the ground (see Figure 2).

Figure 2 illustrates the difference between terrestrial cellular coverage from a cell tower and that of a Low Earth Orbit (LEO) Satellite. The benefit of seeing the world from above is that environmental and physical factors have substantially less impact on signal propagation and quality primarily being impacted by distance as it approximates free space propagation with signal attenuation mainly determined by the Line-of-Sight (LoS) distance from antenna to Earth. This situation is very different for a terrestrial-based cellular tower with its radiated signal being substantially compromised by environmental factors.

Low Earth Orbit (LEO) satellites, compared to GEO and MEO-based higher-altitude satellite systems, in general, have simpler designs and smaller sizes, weights, and volumes. Their design and architecture are not just a function of technological trends but also a manifestation of their operational environment. The (relative) simplicity of LEO satellites also allows for more standardized production, allowing for off-the-shelf components and modular designs that can be manufactured in larger quantities, such as the case with CubeSats standard and SmallSats in general. The lower altitude of LEO satellites translates to a reduced distance from the launch site to the operational orbit, which inherently affects the economics of satellite launches. This proximity to Earth means that the energy required to propel a satellite into LEO is significantly less than needed to reach Geostationary Earth Orbit (GEO), resulting in lower launch costs.

The advent of LEO satellite constellations marks an important shift in how we approach global connectivity. With the potential to provide ubiquitous internet coverage in rural and remote places with little or no terrestrial communications infrastructure, satellites are increasingly being positioned as vital elements in global communication. The LEO satellites, as well as stratospheric drones, have the ability to provide economical internet access, as addressed in my previous article, in remote areas and play a significant role in disaster relief efforts. For example, when terrestrial communication networks may be disrupted after a natural disaster, LEO satellites can quickly re-establish communication links to normal cellular devices or ad-how earth-based satellite systems, enabling efficient coordination of rescue and relief operations. Furthermore, they offer a resilient network backbone that complements terrestrial infrastructure.

The Internet of Things (IoT) benefits from the capabilities of LEO satellites. Particular in areas where there is little or no existing terrestrial communications networks. IoT devices often operate in remote or mobile environments, from sensors in agricultural fields to trackers across shipping routes. LEO satellites provide reliable connectivity to IoT networks, facilitating many applications, such as non- and near real-time monitoring of environmental data, seamless asset tracking over transcontinental journeys, and rapid deployment of smart devices in smart city infrastructures. As an example, let us look at the minimum requirements for establishing a LEO satellite constellation that can gather IoT measurements. At an altitude of 550 km the satellite would take ca. 1.5 hour to return to a given point on its orbit. Earth rotates (see also below) which require us to deploy several orbital planes to ensure that we have continuous coverage throughout the 24 hours of a day (assuming this is required). Depending on the satellite antenna design, the target coverage area, and how often a measurement is required, a satellite constellation to support an IoT business may not require much more than 20 (lower measurement frequency) to 60 (higher measurement frequency, but far from real real-time data collection) LEO satellites (@ 550 km).

For defense purposes, LEO satellite systems present unique advantages. Their lower orbits allow for high-resolution imagery and rapid data collection, which are crucial for surveillance, reconnaissance, and operational awareness. As typically more LEO satellites will be required, compared to a GEO satellite, such systems also offer a higher degree of redundancy in case of anti-satellite (ASAT) warfare scenarios. When integrated with civilian applications, military use cases can leverage the robust commercial infrastructure for communication and geolocation services, enhancing capabilities while distributing the system’s visibility and potential targets.

Standalone military LEO satellites are engineered for specific defense needs. These may include hardened systems for secure communication, resistance to jamming, and interception. For instance, they can be equipped with advanced encryption algorithms to ensure secure transmission of sensitive military data. They also carry tailored payloads for electronic warfare, signal intelligence, and tactical communications. For example, they can host sensors for detecting and locating enemy radar and communication systems, providing a significant advantage in electronic warfare. As the line between civilian and military space applications blurs, dual-use LEO satellite systems are emerging, capable of serving civilian broadband and specialized military requirements. It should be pointed out that there also military applications, such as signal gathering, that may not be compatible with civil communications use cases.

In a military conflict, the distributed architecture and lower altitude of LEO constellations may offer some advantages regarding resilience and targetability compared to GEO and MEO-based satellites. Their more significant numbers (i.e., 10s to 1000s) compared to GEO, and the potential for quicker orbital resupply can make them less susceptible to complete system takedown. However, their lower altitudes could make them accessible to various ASAT technologies, including ground-based missiles or space-based kinetic interceptors.

It is not uncommon to encounter academic researchers and commentators who give the impression that LEO satellites could replace existing terrestrial-based infrastructures and solve all terrestrial communications issues known to man. That is (of course) not the case. Often, such statements appears to be based an incomplete understanding of the capacity limitation of satellite systems. Due to satellites’ excellent coverage with very large terrestrial footprints, the satellite capacity is shared over very large areas. For example, consider an LEO satellite at 550 km altitude. The satellite footprint, or coverage area (aka ground swath), is the area on the Earth’s surface over which the satellite can establish a direct line of sight. The satellite footprint in our example diameter would be ca. five thousand five hundred kilometers. An equivalent area of ca. 23 million square kilometers is more than twice that of the USA (or China or Canada). Before you get too excited, the satellite antenna will typically restrict the surface area the satellite will cover. The extent of the observable world that is seen at any given moment by the satellite antenna is defined as the Field of View (FoV) and can vary from a few degrees (narrow beams, small coverage area) to 40 degrees or higher (wide beams, large coverage areas). At a FoV of 20 degrees, the antenna footprint would be ca. 2 thousand 400 kilometers, equivalent to a coverage area of ca. 5 million square kilometers.

In comparison, for a FoV of 0.8 degrees, the antenna footprint would only be 100 kilometers. If our satellite has a 16-satellite beam capability, it would translate into a coverage diameter of 24 km per beam. For the StarLink system based on the Ku-band (13 GHz) and a cell downlink (Satellite-to-Earth) capacity of ca. 680 Mbps (in 250 MHz) we would have ca. 2 Mbps per km2 unit coverage area. Compared to a terrestrial rural cellular site with 85 MHz (Downlink, Base station antenna to customer terminal), it would deliver 10+ Mbps per km2 unit coverage area.

It is always good to keep in mind that “Satellites mission is not to replace terrestrial communications infrastructures but supplement and enhance them”, and furthermore, “Satellites offer the missing (internet) link in areas where there is no terrestrial communications infrastructure present”. Satellites offer superior coverage to any terrestrial communications infrastructure. Satellites limitations are in providing capacity, and quality, at population scale as well as supporting applications and access technologies requiring very short latencies (e.g., smaller than 10 ms).

In the following, I will focus on terrestrial cellular coverage and services that LEO satellites can provide. At the end of my blog, I hope I have given you (the reader) a reasonable understanding of how terrestrial coverage, capacity, and quality work in a (LEO) satellite system and have given you an impression of key parameters we can add to the satellite to improve those.

EARTH ROTATES, AND SO DO SATELLITES.

Before getting into the details of low earth orbit satellites, let us briefly get a couple of basic topics off the table. Skipping this part may be a good option if you are already into and in the know satellites. Or maybe carry on an get a good laugh of those terra firma cellular folks that forgot about the rotation of Earth 😉

From an altitude and orbit (around Earth) perspective, you may have heard of two types of satellites: The GEO and the LEO satellites. Geostationary (GEO) satellites are positioned in a geostationary orbit at ~36 thousand kilometers above Earth. That the satellite is geostationary means it rotates with the Earth and appears stationary from the ground, requiring only one satellite to maintain constant coverage over an area that can be up to one-third of Earth’s surface. Low Earth Orbit (LEO) satellites are positioned at an altitude between 300 to 2000 kilometers above Earth and move relative to the Earth’s surface at high speeds, requiring a network or constellation to ensure continuous coverage of a particular area.

I have experienced that terrestrial cellular folks (like myself) when first thinking about satellite coverage are having some intuitive issues with satellite coverage. “We are not used to our antennas moving away from the targeted coverage area, and our targeted coverage area, too, is moving away from our antenna“. The geometry and dynamics of terrestrial cellular coverage are simpler than they are for satellite-based coverage. For LEO satellite network planners, it is not rocket science (pun intended) that the satellites move around in their designated orbit over Earth at orbital speeds of ca. 70 to 80 km per second. Thus, at an altitude of 500 km, a LEO satellite orbits Earth approximately every 1.5 hours. Earth, thankfully, rotates. Compared to its GEO satellite “cousin,” the LEO satellite ” is not “stationary” from the perspective of the ground. Thus, as Earth rotates, the targeted coverage area moves away from the coverage provided by the orbital satellite.

We need several satellites in the same orbit and several orbits (i.e., orbital planes) to provide continuous satellite coverage of a target area. This is very different from terrestrial cellular coverage of a given area (needles to say).

WHAT LEO SATELLITES BRING TO THE GROUND.

Anything is infinitely more than nothing. The Low Earth Orbit satellite brings the possibility of internet connectivity where there previously was nothing, either because too few potential customers spread out over a large area made terrestrial-based services hugely uneconomical or the environment is too hostile to build normal terrestrial networks within reasonable economics.

Figure 3 illustrates a low Earth satellite constellation providing internet to rural and remote areas as a way to solve part of the digital divide challenge in terms of availability. Obviously, the affordability is likely to remain a challenge unless subsidized by customers who can afford satellite services in other places where availability is more of a convenience question. (Courtesy: DALL-E)

The LEO satellites represent a transformative shift in internet connectivity, providing advantages over traditional cellular and fixed broadband networks, particularly for global access, speed, and deployment capabilities. As described in “Stratospheric Drones: Revolutionizing Terrestrial Rural Broadband from the Skies?”, LEO satellite constellations, or networks, may also be significantly more economical than equivalent cellular networks in rural and remote areas where the economics of coverage by satellite, as depicted in the above Figure 3, is by far better than by traditional terrestrial cellular means.

One of the foremost benefits of LEO satellites is their ability to offer global coverage as well as reasonable broadband and latency performance that is difficult to match with GEO and MEO satellites. The GEO stationary satellite obviously also offers global broadband coverage, the unit coverage being much more extensive than for a LEO satellite, but it is not possible to offer very low latency services, and it is more difficult to provide high data rates (in comparison to a LEO satellite). LEO satellites can reach the most remote and rural areas of the world, places where laying cables or setting up cell towers is impractical. This is a crucial step in delivering communications services where none exist today, ensuring that underserved populations and regions gain access to internet connectivity.

Another significant advantage is the reduction in latency that LEO satellites provide. Since they orbit much closer to Earth, typically at an altitude between 350 to 700 km, compared to their geostationary counterparts that are at 36 thousand kilometers altitude, the time it takes for a communications signal to travel between the user and the satellite is significantly reduced. This lower latency is crucial for enhancing the user experience in real-time applications such as video calls and online gaming, making these activities more enjoyable and responsive.

An inherent benefit of satellite constellations is their ability for quick deployment. They can be deployed rapidly in space, offering a quicker solution to achieving widespread internet coverage than the time-consuming and often challenging process of laying cables or erecting terrestrial infrastructure. Moreover, the network can easily be expanded by adding more satellites, allowing it to dynamically meet changing demand without extensive modifications on the ground.

LEO satellite networks are inherently scalable. By launching additional satellites, they can accommodate growing internet usage demands, ensuring that the network remains efficient and capable of serving more users over time without significant changes to ground infrastructure.

Furthermore, these satellite networks offer resilience and reliability. With multiple satellites in orbit, the network can maintain connectivity even if one satellite fails or is obstructed, providing a level of redundancy that makes the network less susceptible to outages. This ensures consistent performance across different geographical areas, unlike terrestrial networks that may suffer from physical damage or maintenance issues.

Another critical advantage is (relative) cost-effectiveness compared to a terrestrial-based cellular network. In remote or hard-to-reach areas, deploying satellites can be more economical than the high expenses associated with extending terrestrial broadband infrastructure. As satellite production and launch costs continue to decrease, the economics of LEO satellite internet become increasingly competitive, potentially reducing the cost for end-users.

LEO satellites offer a promising solution to some of the limitations of traditional connectivity methods. By overcoming geographical, infrastructural, and economic barriers, LEO satellite technology has the potential to not just complement but effectively substitute terrestrial-based cellular and fixed broadband services, especially in areas where such services are inadequate or non-existent.

Figure 4 below provides an overview of LEO satellite coverage with fixed broadband services offered to customers in the Ku band with a Ka backhaul link to ground station GWs that connect to, for example, the internet. Having inter-satellite communications (e.g., via laser links such as those used by Starlink satellites as per satellite version 1.5) allows for substantially less ground-station gateways. Inter-satellite laser links between intra-plane satellites are a distinct advantage in ensuring coverage for rural and remote areas where it might be difficult, very costly, and impractical to have a satellite ground station GW to connect to due to the lack of global internet infrastructure.

Figure 4 In general, a satellite is required to have LoS to its ground station gateway (GW); in other words, the GW needs to be within the coverage footprint of the satellite. For LEO satellites, which are at low altitudes, between 300 and 2000 km, and thus have a much lower footprint than MEO and GEO satellites, this would result in a need for a substantial amount of ground stations. This is depicted in (a) above. With inter-satellite laser links (SLL), e.g., those implemented by Starlink, it is possible to reduce the ground station gateways significantly, which is particularly helpful in rural and very remote areas. These laser links enable direct communication between satellites in orbit, which enhances the network’s performance, reliability, and global reach.

Inter-satellite laser links (ISLLs), or, as it is also called Optical Inter-satellite Links (OISK), are an advanced communication technology utilized by satellite constellations, such as for example Starlink, to facilitate high-speed secure data transmission directly between satellites. Inter-satellite laser links are today (primarily) designed for intra-plane communication within satellite constellations, enabling data transfer between satellites that share the same orbital plane. This is due to the relatively stable geometries and predictable distances between satellites in the same orbit, which facilitate maintaining the line-of-sight connections necessary for laser communications. ISLLs mark a significant departure from traditional reliance on ground stations for inter-satellite communication, and as such the ISL offers many benefits, including the ability to transmit data at speeds comparable to fiber-optic cables. Additionally, ISLLs enable satellite constellations to deliver seamless coverage across the entire planet, including over oceans and polar regions where ground station infrastructure is limited or non-existent. The technology also inherently enhances the security of data transmissions, thanks to the focused nature of laser beams, which are difficult to intercept.

However, the deployment of ISLLs is not without challenges. The technology requires a clear line of sight between satellites, which can be affected by their orbital positions, necessitating precise control mechanisms. Moreover, the theoretical limit to the number of satellites linked in a daisy chain is influenced by several factors, including the satellite’s power capabilities, the network architecture, and the need to maintain clear lines of sight. High-power laser systems also demand considerable energy, impacting the satellite’s power budget and requiring efficient management to balance operational needs. The complexity and cost of developing such sophisticated laser communication systems, combined with very precise pointing mechanisms and sensitive detectors, can be quite challenging and need to be carefully weighted against building satellite ground stations.

Cross-plane ISLL transmission, or the ability to communicate between satellites in different orbital planes, presents additional technical challenges, as it is technically highly challenging to maintain a stable line of sight between satellites moving in different orbital planes. However, the potential for ISLLs to support cross-plane links is recognized as a valuable capability for creating a fully interconnected satellite constellation. The development and incorporation of cross-plane ISLL capabilities into satellites are an area of active research and development. Such capabilities would reduce the reliance on ground stations and significantly increase the resilience of satellite constellations. I see the development as a next-generation topic together with many other important developments as described in the end of this blog. However, the power consumption of the ISLL is a point of concern that needs careful attention as it will impact many other aspects of the satellite operation.

THE DIGITAL DIVIDE.

The digital divide refers to the “internet haves and haves not” or “the gap between individuals who have access to modern information and communication technology (ICT),” such as the internet, computers, and smartphones, and those who do not have access. This divide can be due to various factors, including economic, geographic, age, and educational barriers. Essentially, as illustrated in Figure 5, it’s the difference between the “digitally connected” and the “digitally disconnected.”.

The significance of the digital divide is considerable, impacting billions of people worldwide. It is estimated that a little less than 40% of the world’s population, or roughly 2.9 billion people, had never used the internet (as of 2023). This gap is most pronounced in developing countries, rural areas, and among older populations and economically disadvantaged groups.

The digital divide affects individuals’ ability to access information, education, and job opportunities and impacts their ability to participate in digital economies and the modern social life that the rest of us (i.e., the other side of the divide or the privileged 60%) have become used to. Bridging this divide is crucial for ensuring equitable access to technology and its benefits, fostering social and economic inclusion, and supporting global development goals.

Figure 5 illustrates the digital divide, that is, the gap between individuals with access to modern information and communication technology (ICT), such as the internet, computers, and smartphones, and those who do not have access. (Courtesy: DALL-E)

CHALLENGES WITH LEO SATELLITE SOLUTIONS.

Low-Earth-orbit satellites offer compelling advantages for global internet connectivity, yet they are not without challenges and disadvantages when considered substitutes for cellular and fixed broadband services. These drawbacks underscore the complexities and limitations of deploying LEO satellite technology globally.

The capital investment required and the ongoing costs associated with designing, manufacturing, launching, and maintaining a constellation of LEO satellites are substantial. Despite technological advancements and increased competition driving costs down, the financial barrier to entry remains high. Compared to their geostationary counterparts, the relatively short lifespan of LEO satellites necessitates frequent replacements, further adding to operational expenses.

While LEO satellites offer significantly reduced latency (round trip times, RTT ~ 4 ms) compared to geostationary satellites (RTT ~ 240 ms), they may still face latency and bandwidth limitations, especially as the number of users on the satellite network increases. This can lead to reduced service quality during peak usage times, highlighting the potential for congestion and bandwidth constraints. This is also the reason why the main business model of LEO satellite constellations is primarily to address coverage and needs in rural and remote locations. Alternatively, the LEO satellite business model focuses on low-bandwidth needs such as texting, voice messaging, and low-bandwidth Internet of Things (IoT) services.

Navigating the regulatory and spectrum management landscape presents another challenge for LEO satellite operators. Securing spectrum rights and preventing signal interference requires coordination across multiple jurisdictions, which can complicate deployment efforts and increase the complexity of operations.

The environmental and space traffic concerns associated with deploying large numbers of satellites are significant. The potential for space debris and the sustainability of low Earth orbits are critical issues, with collisions posing risks to other satellites and space missions. Additionally, the environmental impact of frequent rocket launches raises further concerns.

FIXED-WIRELESS ACCESS (FWA) BASED LEO SATELLITE SOLUTIONS.

Using the NewSpace Index database, updated December 2023, there are currently more than 6,463 internet satellites launched, of which 5,650 (~87%) from StarLink, and 40,000+ satellites planned for launch, with SpaceX’s Starlink satellites having 11,908 planned (~30%). More than 45% of the satellites launched and planned support multi-application use cases. Thus internet, together with, for example, IoT (~4%) and/or Direct-2-Device (D2D, ~39%). The D2D share is due to StarLink’s plans to provide services to mobile terminals with their latest satellite constellation. The first six StarLink v2 satellites with direct-to-cellular capability were successfully launched on January 2nd, 2024. Some care should be taken in the share of D2D satellites in the StarLink number as it does not consider the different form factors of the version 2 satellite that do not all include D2D capabilities.

Most LEO satellites, helped by StarLink satellite quantum, operational and planned, support satellite fixed broadband internet services. It is worth noting that the Chinese Guo Wang constellation ranks second in terms of planned LEO satellites, with almost 13,000 planned, rivaling the StarLink constellation. After StarLink and Guo Wang are counted there is only 34% or ca. 16,000 internet satellites left in the planning pool across 30+ satellite companies. While StarLink is privately owned (by Elon Musk), the Guo Wang (國網 ~ “The state network”) constellation is led by China SatNet and created by the SASAC (China’s State-Owned Assets Supervision and Administration Commission). SASAC oversees China’s biggest state-owned enterprises. I expect that such an LEO satellite constellation, which would be the second biggest LEO constellation, as planned by Guo Wang and controlled by the Chinese State, would be of considerable concern to the West due to the possibility of dual-use (i.e., civil & military) of such a constellation.

StarLink coverage as of March 2024 (see StarLink’s availability map) does not provide services in Russia, China, Iran, Iraq, Afghanistan, Venezuela, and Cuba (20% of Earth’s total land base surface area). There are still quite a few countries in Africa and South-East Asia, including India, where regulatory approval remains pending.

Figure 6 NewSpace Index data of commercial satellite constellations in terms of total number of launched and planned (top) per company (or constellation name) and (bottom) per country.

While the term FWA, fixed wireless access, is not traditionally used to describe satellite internet services, the broadband services offered by LEO satellites can be considered a form of “wireless access” since they also provide connectivity without cables or fiber. In essence, LEO satellite broadband is a complementary service to traditional FWA, extending wireless broadband access to locations beyond the reach of terrestrial networks. In the following, I will continue to use the term FWA for the fixed broadband LEO satellite services provided to individual customers, including SMEs. As some of the LEO satellite businesses eventually also might provide direct-to-device (D2D) services to normal terrestrial mobile devices, either on their own acquired cellular spectrum or in partnership with terrestrial cellular operators, the LEO satellite operation (or business architecture) becomes much closer to terrestrial cellular operations.

Figure 7 Illustrating a Non-Terrestrial Network consisting of a Low Earth Orbit (LEO) satellite constellation providing fixed broadband services, such as Fixed Wireless Access, to individual terrestrial users (e.g., Starlink, Kuiper, OneWeb,…). Each hexagon represents a satellite beam inside the larger satellite coverage area. Note that, in general, there will be some coverage overlap between individual satellites, ensuring a continuous service. The operating altitude of an LEO satellite constellation is between 300 and 2,000 km, with most aiming to be at 450 to 550 km altitude. It is assumed that the satellites are interconnected, e.g., laser links. The User Terminal antenna (UT) is dynamically orienting itself after the best line-of-sight (in terms of signal quality) to a satellite within UT’s field-of-view (FoV). The FoV has not been shown in the picture above so as not to overcomplicate the illustration.

Low Earth Orbit (LEO) satellite services like Starlink have emerged to provide fixed broadband internet to individual consumers and small to medium-sized enterprises (SMEs) targeting rural and remote areas often where no other broadband solutions are available or with poor legacy copper- or coax-based infrastructure. These services deploy constellations of satellites orbiting close to Earth to offer high-speed internet with the significant advantage of reaching rural and remote areas where traditional ground-based infrastructure is absent or economically unfeasible.

One of the most significant benefits of LEO satellite broadband is the ability to deliver connectivity with lower latency compared to traditional satellite internet delivered by geosynchronous satellites, enhancing the user experience for real-time applications. The rapid deployment capability of these services also means that areas in dire need of internet access can be connected much quicker than waiting for ground infrastructure development. Additionally, satellite broadband’s reliability is less affected by terrestrial challenges, such as natural disasters that can disrupt other forms of connectivity.

The satellite service comes with its challenges. The cost of user equipment, such as satellite dishes, can be a barrier for some users. So, can the installation process be of the terrestrial satellite dish required to establish the connection to the satellite. Moreover, services might be limited by data caps or experience slower speeds after reaching certain usage thresholds, which can be a drawback for users with high data demands. Weather conditions can also impact the signal quality, particularly at the higher frequencies used by the satellite, albeit to a lesser extent than geostationary satellite services. However, the target areas where the fixed broadband satellite service is most suited are rural and remote areas that either have no terrestrial broadband infrastructure (terrestrial cellular broadband or wired broadband such as coax or fiber)

Beyond Starlink, other providers are venturing into the LEO satellite broadband market. OneWeb is actively developing a constellation to offer internet services worldwide, focusing on communities that are currently underserved by broadband. Telesat Lightspeed is also gearing up to provide broadband services, emphasizing the delivery of high-quality internet to the enterprise and government sectors.

Other LEO satellite businesses, such as AST SpaceMobile and Lynk Mobile, are taking a unique approach by aiming to connect standard mobile phones directly to their satellite network, extending cellular coverage beyond the reach of traditional cell towers. More about that in the section below (see “New Kids on the Block – Direct-to-Devices LEO satellites”).

I have been asked why I appear somewhat dismissive of the Amazon’s Project Kuiper in a previous version of article particular compared to StarLink (I guess). The expressed mission is to “provide broadband services to unserved and underserved consumers, businesses in the United States, …” (FCC 20-102). Project Kuiper plans for a broadband constellation of 3,226 microsatellites at 3 altitudes (i.e., orbital shells) around 600 km providing fixed broadband services in the Ka-band (i.e.,~ 17-30 GHz). In its US-based FCC (Federal Communications Commission) filling and in the subsequent FCC authorization it is clear that the Kuiper constellation primarily targets contiguous coverage of the USA (but mentions that services cannot be provided in the majority of Alaska, … funny I thought that was a good definition of a underserved remote and scarcely populated area?). Amazon has committed to launch 50% (1,618 satellites) of their committed satellites constellation before July 2026 (until now 2+ has been launched) and the remaining 50% before July 2029. There is however far less details on the Kuiper satellite design, than for example is available for the various versions of the StarLink satellites. Given the Kuiper will operate in the Ka-band there may be more frequency bandwidth allocated per beam than possible in the StarLink satellites using the Ku-band for customer device connectivity. However, Ka-band is at a higher frequency which may result in a more compromised signal propagation. In my opinion based on the information from the FCC submissions and correspondence, the Kuiper constellation appear less ambitious compared to StarLink vision, mission and tangible commitment in terms of aggressive launches, very high level of innovation and iterative development on their platform and capabilities in general. This may of course change over time and as more information becomes available on the Amazon’s Project Kuiper.

FWA-based LEO satellite solutions – takeaway:

  • LoS-based and free-space-like signal propagation allows high-frequency signals (i.e., high throughput, capacity, and quality) to provide near-ideal performance only impacted by the distance between the antenna and the ground terminal. Something that is, in general, not possible for a terrestrial-based cellular infrastructure.
  • Provides satellite fixed broadband internet connectivity typically using the Ku-band in geographically isolated locations where terrestrial broadband infrastructure is limited or non-existent.
  • Lower latency (and round trip time) compared to MEO and GEO satellite internet solutions.
  • Current systems are designed to provide broadband internet services in scarcely populated areas and underserved (or unserved) regions where traditional terrestrial-based communications infrastructures are highly uneconomical and/or impractical to deploy.
  • As shown in my previous article (i.e., “Stratospheric Drones: Revolutionizing Terrestrial Rural Broadband from the Skies?”), LEO satellite networks may be an economical interesting alternative to terrestrial rural cellular networks in countries with large scarcely populated rural areas requiring tens of thousands of cellular sites to cover. Hybrid models with LEO satellite FWA-like coverage to individuals in rural areas and with satellite backhaul to major settlements and towns should be considered in large geographies.
  • Resilience to terrestrial disruptions is a key advantage. It ensures functionality even when ground-based infrastructure is disrupted, which is an essential element for maintaining the Business Continuity of an operator’s telecommunications services. Particular hierarchical architectures with for example GEO-satellite, LEO satellite and Earth-based transport infrastructure will result in very high reliability network operations (possibly approaching ultra-high availability, although not with service parity).
  • Current systems are inherently capacity-limited due to their vast coverage areas (i.e., lower performance per unit coverage area). In the peak demand period, they will typically perform worse than terrestrial-based cellular networks (e.g., LTE or 5G).
  • In regions where modern terrestrial cellular and fixed broadband services are already established, satellite broadband may face challenges competing with these potentially cheaper, faster, and more reliable services, which are underpinned by the terrestrial communications infrastructure.
  • It is susceptible to weather conditions, such as heavy rain or snow, which can degrade signal quality. This may impact system capacity and quality, resulting in inconsistent customer experience throughout the year.
  • Must navigate complex regulatory environments in each country, which can affect service availability and lead to delays in service rollout.
  • Depending on the altitude, LEO satellites are typically replaced on a 5—to 7-year cycle due to atmospheric drag (which increases as altitude decreases; thus, the lower the altitude, the shorter a satellite’s life). This ultimately means that any improvements in system capacity and quality will take time to be thoroughly enjoyed by all customers.

SATELLITE BACKHAUL SOLUTIONS.

Figure 8 illustrates the architecture of a Low Earth Orbit (LEO) satellite backhaul system used by providers like OneWeb as well as StarLink with their so-called “Community Gateway”. It showcases the connectivity between terrestrial internet infrastructure (i.e., Satellite Gateways) and satellites in orbit, enabling high-speed data transmission. The network consists of LEO satellites that communicate with each other (inter-satellite Comms) using the Ku and Ka frequency bands. These satellites connect to ground-based satellite gateways (GW), which interface with Points of Presence (PoP) and Internet Exchange Points (IXP), integrating the space-based network with the terrestrial internet (WWW). Note: The indicated speeds and frequency bands (e.g., Ku: 12–18 GHz, Ka: 28–40 GHz) and data speeds illustrate the network’s capabilities.

LEO satellites providing backhaul connectivity, such as shown in Figure 8 above, are extending internet services to the farthest reaches of the globe. These satellites offer many benefits, as already discussed above, in connecting remote, rural, and previously un- and under-served areas with reliable internet services. Many remote regions lack foundational telecom infrastructure, particularly long-haul transport networks needed for carrying traffic away from remote populated areas. Satellite backhauls do not only offer a substantially better financial solution for enhancing internet connectivity to remote areas but are often the only viable solution for connectivity.

Take, for example, Greenland. The world’s largest non-continental island, the size of Western Europe, is characterized by its sparse population and distinct unconnected by road settlement patterns mainly along the West Coast (as well as a couple of settlements on the East Coast), influenced mainly by its vast ice sheets and rugged terrain. With a population of around 56+ thousand, primarily concentrated on the west coast, Greenland’s demographic distribution is spread out over ca. 50+ settlements and about 20 towns. Nuuk, the capital, is the island’s most populous city, housing over 18+ thousand residents and serving as the administrative, economic, and cultural hub. Terrestrial cellular networks serve settlements’ and towns’ communication and internet services needs, with the traffic carried back to the central switching centers by long-haul microwave links, sea cables, and satellite broadband connectivity. Several settlements connectivity needs can only be served by satellite backhaul, e.g., settlements on the East Coast (e.g., Tasiilaq with ca. 2,000 inhabitants and Ittoqqotooormiit (an awesome name!) with around 400+ inhabitants). LEO satellite backhaul solutions serving Satellite-only communities, such as those operated and offered by OneWeb (Eutelsat), could provide a backhaul transport solution that would match FWA latency specifications due to better (round trip time) performance than that of a GEO satellite backhaul solution.

It should also be clear that remote satellite-only settlements and towns may have communications service needs and demand that a localized 4G (or 5G) terrestrial cellular network with a satellite backhaul can serve much better than, for example, relying on individual ad-hoc connectivity solution from for example Starlink. When the area’s total bandwidth demand exceeds the capacity of an FWA satellite service, a localized terrestrial network solution with a satellite backhaul is, in general, better.

The LEO satellites should offer significantly reduced latency compared to their geostationary counterparts due to their closer proximity to the Earth. This reduction in delay is essential for a wide range of real-time applications and services, from adhering to modern radio access (e.g., 4G and 5G) requirements, VoIP, and online gaming to critical financial transactions, enhancing the user experience and broadening the scope of possible services and business.

Among the leading LEO satellite constellations providing backhaul solutions today are SpaceX’s Starlink (via their community gateway), aiming to deliver high-speed internet globally with a preference of direct to consumer connectivity; OneWeb, focusing on internet services for businesses and communities in remote areas; Telesat’s Lightspeed, designed to offer secure and reliable connectivity; and Amazon’s Project Kuiper, which plans to deploy thousands of satellites to provide broadband to unserved and underserved communities worldwide.

Satellite backhaul solutions – takeaway:

  • Satellite-backhaul solutions are excellent, cost-effective solution for providing an existing isolated cellular (and fixed access) network with high-bandwidth connectivity to the Internet (such as in remote and deep rural areas).
  • LEO satellites can reduce the need for extensive and very costly ground-based infrastructure by serving as a backhaul solution. For some areas, such as Greenland, the Sahara, or the Brazilian rainforest, it may not be practical or economical to connect by terrestrial-based transmission (e.g., long-haul microwave links or backbone & backhaul fiber) to remote settlements or towns.
  • An LEO-based backhaul solution supports applications and radio access technologies requiring a very low round trip time scale (RTT<50 ms) than is possible with a GEO-based satellite backhaul. However, the optimum RTT will depend on where the LEO satellite ground gateway connects to the internet service provider and how low the RTT can be.
  • The collaborative nature of a satellite-backhaul solution allows the terrestrial operator to focus on and have full control of all its customers’ network experiences, as well as optimize the traffic within its own network infrastructure.
  • LEO satellite backhaul solutions can significantly boost network resilience and availability, providing a secure and reliable connectivity solution.
  • Satellite-backhaul solutions require local ground-based satellite transmission capabilities (e.g., a satellite ground station).
  • The operator should consider that at a certain threshold of low population density, direct-to-consumer satellite services like Starlink might be more economical than constructing a local telecom network that relies on satellite backhaul (see above section on “Fixed Wireless Access (FWA) based LEO satellite solutions”).
  • Satellite backhaul providers require regulatory permits to offer backhaul services. These permits are necessary for several reasons, including the use of radio frequency spectrum, operation of satellite ground stations, and provision of telecommunications services within various jurisdictions.
  • The Satellite life-time in orbit is between 5 to 7 years depending on the LEO altitude. A MEO satellite (2 to 36 thousand km altitude) last between 10 to 20 years (GEO). This also dictates the modernization and upgrade cycle as well as timing of your ROI investment case and refinancing needs.

NEW KIDS ON THE BLOCK – DIRECT-TO-DEVICE LEO SATELLITES.

A recent X-exchange (from March 2nd):

Elon Musk: “SpaceX just achieved peak download speed of 17 Mb/s from a satellite direct to unmodified Samsung Android Phone.” (note: the speed correspond to a spectral efficiency of ~3.4 Mbps/MHz/beam).

Reply from user: “That’s incredible … Fixed wireless networks need to be looking over their shoulders?”

Elon Musk: “No, because this is the current peak speed per beam and the beams are large, so this system is only effective where there is no existing cellular service. This services works in partnership with wireless providers, like what @SpaceX and @TMobile announced.”

Figure 9 illustrating a LEO satellite direct-to-device communication in a remote areas without any terrestrially-based communications infrastructure. Satellite being the only means of communications either by a normal mobile device or by classical satphone. (Courtesy: DALL-E).

Low Earth Orbit (LEO) Satellite Direct-to-Device technology enables direct communication between satellites in orbit and standard mobile devices, such as smartphones and tablets, without requiring additional specialized hardware. This technology promises to extend connectivity to remote, rural, and underserved areas globally, where traditional cellular network infrastructure is absent or economically unfeasible to deploy. The system can offer lower latency communication by leveraging LEO satellites, which orbit closer to Earth than geostationary satellites, making it more practical for everyday use. The round trip time (RTT), the time it takes the for the signal to travel from the satellite to the mobile device and back, is ca. 4 milliseconds for a LEO satellite at 550 km compared to ca. 240 milliseconds for a geosynchronous satellite (at 36 thousand kilometers altitude).

The key advantage of a satellite in low Earth orbit is that the likelihood of a line-of-sight to a point on the ground is very high compared to establishing a line-of-sight for terrestrial cellular coverage that, in general, would be very low. In other words, the cellular signal propagation from a LEO satellite closely approximates that of free space. Thus, all the various environmental signal loss factors we must consider for a standard terrestrial-based mobile network do not apply to our satellite. In other, more simplistic words, the signal propagation directly from the satellite to the mobile device is less compromised than it typically would be from a terrestrial cellular tower to the same mobile device. The difference between free-space propagation, which considers only distance and frequency, and the terrestrial signal propagation models, which quantifies all the gains and losses experienced by a terrestrial cellular signal, is very substantial and in favor of free-space propagation.  As our Earth-bound cellular intuition of signal propagation often gets in the way of understanding the signal propagation from a satellite (or antenna in the sky in general), I recommend writing down the math using the formula of free space propagation loss and comparing this with terrestrial cellular link budget models, such as for example the COST 231-Hata Model (relatively simple) or the more recent 3GPP TR 38.901 Model (complex). In rural and sub-urban areas, depending on the environment, in-door coverage may be marginally worse, fairly similar, or even better than from terrestrial cell tower at a distance. This applies to both the uplink and downlink communications channel between the mobile device and the LEO satellite, and is also the reason why higher frequency (with higher frequency bandwidths available) use on LEO satellites can work better than in a terrestrial cellular network.

However, despite its potential to dramatically expand coverage, after all that is what satellites do, LEO Satellite Direct-to-Device technology is not a replacement for terrestrial cellular services and terrestrial communications infrastructures for several reasons: (a) Although the spectral efficiency can be excellent, the frequency bandwidth (in MHz) and data speeds (in Mbps) available through satellite connections are typically lower than those provided by ground-based cellular networks, limiting its use for high-bandwidth applications. (b) The satellite-based D2D services are, in general, capacity-limited and might not be able to handle higher user density typical for urban areas as efficiently as terrestrial networks, which are designed to accommodate large numbers of users through dense deployment of cell towers. (c) Environmental factors like buildings or bad weather can more significantly impact satellite communications’ reliability and quality than terrestrial services. (d) A satellite D2D service requires regulatory approval (per country), as the D2D frequency typically will be limited to terrestrial cellular services and will have to be coordinated and managed with any terrestrial use to avoid service degradation (or disruption) for customers using terrestrial cellular services also using the frequency. The satellites will have to be able to switch off their D2D service when the satellite covers jurisdictions that have not provided approval or where the relevant frequency/frequencies are in use terrestrially.

Using the NewSpace Index database, updated December 2023, there are current more than 8,000 Direct-to Device (D2D), or Direct-2-Cell (D2C), satellites planned for launch, with SpaceX’s Starlink v2 having 7,500 planned. The rest, 795 satellites, are distributed on 6 other satellite operators (e.g. AST Mobile, Sateliot (Spain), Inmarsat (HEO-orbit), Lynk,…). If we look at satellites designed for IoT connectivity we get in total 5,302, with 4,739 (not including StarLink) still planned, distributed out over 50+ satellite operators. The average IoT satellite constellation including what is currently planned is ~95 satellites with the majority targeted for LEO. The the satellite operators included in the 50+ count have confirmed funding with a minimum amount of US$2 billion (half of the operators have only funding confirmed without an amount). About 2,937 (435 launched) satellites are being planned to only serve IoT markets (note: I think this seems a bit excessive). With Swarm Technologies, a SpaceX subsidiary rank number 1 in terms of both launched and planned satellites. Swarm Technologies having launched at least 189 CubeSats (e.g., both 0.25U and 1U types) and have planned an addition 150. The second ranked IoT-only operator is Orbcomm with 51 satellites launched and an additional 52 planned. The average launched of the remaining IoT specific satellites operators are 5 with on average planning to launch 55 (over 42 constellations).

There are also 3 satellite operators (i.e., Chinese-based Galaxy Space: 1,000 LEO-sats; US-based Mangata Networks: 791 MEO/HEO-sats, and US-based Omnispace: 200 LEO?-sats) that have planned a total of 2,000 satellites to support 5G applications with their satellite solutions and one operator (i.e., Hanwha Systems) has planned 2,000 LEO satellites for 6G.

The emergence of LEO satellite direct-to-device (D2D) services, as depicted in the Figure 10 below, is at the forefront of satellite communication innovations, offering a direct line of connectivity between devices that bypasses the need for traditional cellular-based ground-based network infrastructure (e.g., cell towers). This approach benefits from the relatively short distance of hundreds of kilometers between LEO satellites and the Earth, reducing communication latency and broadening bandwidth capabilities compared to their geostationary counterparts. One of the key advantages of LEO D2D services is their ability to provide global coverage with an extensive number of satellites, i.e., in their 100s to 1000s depending the targeted quality of service, to support the services, ensuring that even the most remote and underserved areas have access to reliable communication channels. They are also critical in disaster resilience, maintaining communications when terrestrial networks fail due to emergencies or natural disasters.

Figure 10 This schematic presents the network architecture for satellite-based direct-to-device (D2D) communication facilitated by Low Earth Orbit (LEO) satellites, exemplified by collaborations like Starlink and T-Mobile US, Lynk Mobile, and AST Space Mobile. It illustrates how satellites in LEO enable direct connectivity between user equipment (UE), such as standard mobile devices and IoT (Internet of Things) devices, using terrestrial cellular frequencies and VHF/UHF bands. The system also shows inter-satellite links operating in the Ka-band for seamless network integration, with satellite gateways (GW) linking the space-based network to ground infrastructure, including Points of Presence (PoP) and Internet Exchange Points (IXP), which connect to the wider internet (WWW). This architecture supports innovative services like Omnispace and Astrocast, offering LEO satellite IoT connectivity. The network could be particularly crucial for defense and special operations in remote and challenging environments, such as the deserts or the Arctic regions of Greenland, where terrestrial networks are unavailable. As an example shown here, using regular terrestrial cellular frequencies in both downlink (~300 MHz to 7 GHz) and uplinks (900 MHz or lower to 2.1 GHz) ensures robust and versatile communication capabilities in diverse operational contexts.

While the majority of the 5,000+ Starlink constellation is 13 GHz (Ku-band), at the beginning of 2024, SpaceX launched a few 2nd generation Starlink satellites that support direct connections from the satellite to a normal cellular device (e.g., smartphone), using 5 MHz of T-Mobile USA’s PCS band (1900 MHz). The targeted consumer service, as expressed by T-Mobile USA, provides texting capabilities across the USA for areas with no or poor existing cellular coverage. This is fairly similar to services at similar cellular coverage areas presently offered by, for example, AST SpaceMobile, OmniSpace, and Lynk Global LEO satellite services with reported maximum downlink speed approaching 20 Mbps. The so-called Direct-2-Device, where the device is a normal smartphone without satellite connectivity functionality, is expected to develop rapidly over the next 10 years and continue to increase the supported user speeds (i.e., utilized terrestrial cellular spectrum) and system capacity in terms of smaller coverage areas and higher number of satellite beams.

Table 1 below provides an overview of the top 13 LEO satellite constellations targeting (fixed) internet services (e.g., Ku band), IoT and M2M services, and Direct-to-Device (or Direct-to-Cell, D2C) services. The data has been compiled from the NewSpace Index website, which should be with data as of 31st of December 2023. The Top-satellite constellation rank has been based on the number of launched satellites until the end of 2023. Two additional Direct-2-Cell (D2C or Direct-to-Device, D2D) LEO satellite constellations are planned for 2024-2025. One is SpaceX Starlink 2nd generation, which launched at the beginning of 2024, using T-Mobile USA’s PCS Band to connect (D2D) to normal terrestrial cellular handsets. The other D2D (D2C) service is Inmarsat’s Orchestra satellite constellation based on L-band (for mobile terrestrial services) and Ka for fixed broadband services. One new constellation (Mangata Networks, see also the NewSpace constellation information) targeting 5G services. With two 5G constellations already launched, i.e., Galaxy Space (Yinhe) launched 8 LEO satellites, 1,000 planned using Q- and V-bands (i.e., not a D2D cellular 5G service), and OmniSpace launched two satellites and appear to have planned a total of 200 satellites. Moreover, currently, there is one planned constellation targeting 6G by the South Korean Hanwha Group (a bit premature, but interesting to follow nevertheless) with 2,000 6G (LEO) satellites planned.

Most currently launched and planned satellite constellations offering (or plan to provide) Direct-2-Cell services, including IoT and M2M, are designed for low-frequency bandwidth services that are unlikely to compete with terrestrial cellular networks’ quality of service where reasonable good coverage (or better) exists.

Table 1 An overview of the Top-14 LEO satellite constellations targeting (fixed) internet services (e.g., Ku band), IoT and M2M services, and Direct-to-Device (or direct-to-cell) services. The data has been compiled from the NewSpace Index website, which should be with data as of 31st of December 2023.

The deployment of LEO D2D services also navigates a complicated regulatory landscape, with the need for harmonized spectrum allocation across different regions. Managing interference with terrestrial cellular networks and other satellite operations is another interesting challenge albeit complex aspect, requiring sophisticated solutions to ensure signal integrity. Moreover, despite the cost-effectiveness of LEO satellites in terms of launch and operation, establishing a full-fledged network for D2D services demands substantial initial investment, covering satellite development, launch, and the setup of supporting ground infrastructure.

LEO satellites with D2D-based capabilities – takeaway:

  • Provides lower-bandwidth services (e.g., GPRS/EDGE/HSDPA-like) where no existing terrestrial cellular service is present.
  • (Re-)use on Satellite of the terrestrial cellular spectrum.
  • D2D-based satellite services may become crucial in business continuity scenarios, providing redundancy and increased service availability to existing terrestrial cellular networks. This is particularly essential as a remedy for emergency response personnel in case terrestrial networks are not functional. Limited capacity (due to little assigned frequency bandwidth) over a large coverage area serving rural and remote areas with little or no cellular infrastructure.
  • Securing regulatory approval for satellite services over independent jurisdictions is a complex and critical task for any operator looking to provide global or regional satellite-based communications. The satellite operator may have to switch off transmission over jurisdictions where no permission has been granted.
  • If the spectrum is also deployed on the ground, satellite use of it must be managed and coordinated (due to interference) with the terrestrial cellular networks.
  • Require lowly or non-utilized cellular spectrum in the terrestrial operator’s spectrum portfolio.
  • D2D-based communications require a more complex and sophisticated satellite design, including the satellite antenna resulting in higher manufacturing and launch cost.
  • The IoT-only commercial satellite constellation “space” is crowded with a total of 44 constellations (note: a few operators have several constellations). I assume that many of those plans will eventually not be realized. Note that SpaceX Swarm Technology is leading and in terms of total numbers (available in the NewSpace Index) database will remain a leader from the shear amount of satellites once their plan has been realized. I expect we will see a Chinese constellation in this space as well unless the capability will be built into the Guo Wang constellation.
  • The Satellite life-time in orbit is between 5 to 7 years depending on the altitude. This timeline also dictates the modernization and upgrade cycle as well as timing of your ROI investment and refinancing needs.
  • Today’s D2D satellite systems are frequency-bandwidth limited. However, if so designed, satellites could provide a frequency asymmetric satellite-to-device connection. For instance, the downlink from the satellite to the device could utilize a high frequency (not used in the targeted rural or remote area) and a larger bandwidth, while the uplink communication between the terrestrial device and the LEO satellite could use a sufficiently lower frequency and smaller frequency bandwidth.

MAKERS OF SATELLITES.

In the rapidly evolving space industry, a diverse array of companies specializes in manufacturing satellites for Low Earth Orbit (LEO), ranging from small CubeSats to larger satellites for constellations similar to those used by OneWeb (UK) and Starlink (USA). Among these, smaller companies like NanoAvionics (Lithuania) and Tyvak Nano-Satellite Systems (USA) have carved out niches by focusing on modular and cost-efficient small satellite platforms typically below 25 kg. NanoAvionics is renowned for its flexible mission support, offering everything from design to operation services for CubeSats (e.g., 1U, 3U, 6U) and larger small satellites (100+ kg). Similarly, Tyvak excels in providing custom-made solutions for nano-satellites and CubeSats, catering to specific mission needs with a comprehensive suite of services, including design, manufacturing, and testing.

UK-based Surrey Satellite Technology Limited (SSTL) stands out for its innovative approach to small, cost-effective satellites for various applications, with cost-effectiveness in achieving the desired system’s performance, reliability, and mission objectives at a lower cost than traditional satellite projects that easily runs into USD 100s of million. SSTL’s commitment to delivering satellites that balance performance and budget has made it a popular satellite manufacturer globally.

On the larger end of the spectrum, companies like SpaceX (USA) and Thales Alenia Space (France-Italy) are making significant strides in satellite manufacturing at scale. SpaceX has ventured beyond its foundational launch services to produce thousands of small satellites (250+ kg) for its Starlink broadband constellation, which comprises 5,700+ LEO satellites, showcasing mass satellite production. Thales Alenia Space offers reliable satellite platforms and payload integration services for LEO constellation projects.

With their extensive expertise in aerospace and defense, Lockheed Martin Space (USA) and Northrop Grumman (USA) produce various satellite systems suitable for commercial, military, and scientific missions. Their ability to support large-scale satellite constellation projects from design to launch demonstrates high expertise and reliability. Similarly, aerospace giants Airbus Defense and Space (EU) and Boeing Defense, Space & Security (USA) offer comprehensive satellite solutions, including designing and manufacturing small satellites for LEO. Their involvement in high-profile projects highlights their capacity to deliver advanced satellite systems for a wide range of use cases.

Together, these companies, from smaller specialized firms to global aerospace leaders, play crucial roles in the satellite manufacturing industry. They enable a wide array of LEO missions, catering to the burgeoning demand for satellite services across telecommunications, Earth observation, and beyond, thus facilitating access to space for diverse clients and applications.

ECONOMICS.

Before going into details, let’s spend some time on an example illustrating the basic components required for building a satellite and getting it to launch. Here, I point at a super cool alternative to the above-mentioned companies, the USA-based startup Apex, co-founded by CTO Max Benassi (ex-SpaceX and Astra) and CEO Ian Cinnamon. To get an impression of the macro-components of a satellite system, I recommend checking out the Apex webpage and “playing” with their satellite configurator. The basic package comes at a price tag of USD 3.2 million and a 9-month delivery window. It includes a 100 kg satellite bus platform, a power system, a communication system based on X-band (8 – 12 GHz), and a guidance, navigation, and control package. The basic package does not include a solar array drive assembly (SADA), which plays a critical role in the operation of satellites by ensuring that the satellite’s solar panels are optimally oriented toward the Sun. Adding the SADA brings with it an additional USD 500 thousand. Also, the propulsion mechanism (e.g., chemical or electric; in general, there are more possibilities) is not provided (+ USD 450 thousand), nor are any services included (e.g., payload & launch vehicle integration and testing, USD 575 thousand), including SADAs, propulsion, and services, Apex will have a satellite launch ready for an amount of close to USD 4.8 million.

However, we are not done. The above solution still needs to include the so-called payload, which relates to the equipment or instruments required to perform the LEO satellite mission (e.g., broadband communications services), the actual satellite launch itself, and the operational aspects of a successful post-launch (i.e., ground infrastructure and operation center(s)).

Let’s take SpaceX’s Starlink satellite as an example illustrating mission and payload more clearly. The Starlink satellite’s primary mission is to provide fixed-wireless access broadband internet to an Earth-based fixed antenna using. The Starlink payload primarily consists of advanced broadband internet transmission equipment designed to provide high-speed internet access across the globe. This includes phased-array antennas for communication with user terminals on the ground, high-frequency radio transceivers to facilitate data transmission, and inter-satellite links allowing satellites to communicate in orbit, enhancing network coverage and data throughput.

The economical aspects of launching a Low Earth Orbit (LEO) satellite project span a broad spectrum of costs from the initial concept phase to deployment and operational management. These projects commence with research and development, where significant investments are made in design, engineering, and the iterative process of prototyping and testing to ensure the satellite meets its intended performance and reliability standards in harsh space conditions (e.g., vacuum, extreme temperature variations, radiation, solar flares, high-velocity impacts with micrometeoroids and man-made space debris, erosion, …).

Manufacturing the satellite involves additional expenses, including procuring high-quality components that can withstand space conditions and assembling and integrating the satellite bus with its mission-specific payload. Ensuring the highest quality standards throughout this process is crucial to minimizing the risk of in-orbit failure, which can substantially increase project costs. The payload should be seen as the heart of the satellite’s mission. It could be a set of scientific instruments for measuring atmospheric data, optical sensors for imaging, transponders for communication, or any other equipment designed to fulfill the satellite’s specific objectives. The payload will vary greatly depending on the mission, whether for Earth observation, scientific research, navigation, or telecommunications.

Of course, there are many other types and more affordable options for LEO satellites than a Starlink-like one (although we should also not ignore achievements of SpaceX and learn from them as much as possible). As seen from Table 1, we have a range of substantially smaller satellite types or form factors. The 1U (i.e., one unit) CubeSat is a satellite with a form factor of 10x10x11.35 cm3 and weighs no more than 1.33 kilograms. A rough cost range for manufacturing a 1U CubeSat could be from USD 50 to 100+ thousand depending on mission complexity and payload components (e.g., commercial-off-the-shelf or application or mission-specific design). The range includes considering the costs associated with the satellite’s design, components, assembly, testing, and initial integration efforts. The cost range, however, does not include other significant costs associated with satellite missions, such as launch services, ground station operations, mission control, and insurance, which is likely to (significantly) increase the total project cost. Furthermore, we have additional form factors, such as 3U CubeSat (10x10x34.05 cm3, <4 kg), manufacturing cost in the range of USD 100 to 500+ thousand, 6U CubeSat (20x10x34 cm3, <12 kg), that can carry more complex payload solutions than the smaller 1U and 3U, with the manufacturing cost in the range of USD 200 thousand to USD 1+ million and 12U satellites (20x20x34 cm3, <24 kg) that again support complex payload solutions and in general will be significantly more expensive to manufacture.

Securing a launch vehicle is one of the most significant expenditures in a satellite project. This cost not only includes the price of the rocket and launch itself but also encompasses integration, pre-launch services, and satellite transportation to the launch site. Beyond the launch, establishing and maintaining the ground segment infrastructure, such as ground stations and a mission control center, is essential for successful satellite communication and operation. These facilities enable ongoing tracking, telemetry, and command operations, as well as the processing and management of the data collected by the satellite.

The SpaceX Falcon rocket is used extensively by other satellite businesses (see above Table 1) as well as by SpaceX for their own Starlink constellation network. The rocket has a payload capability of ca. 23 thousand kg and a volume handling capacity of approximately 300 cubic meters. SpaceX has launched around 60 Starlink satellites per Falcon 9 mission for the first-generation satellites. The launch cost per 1st generation satellite would then be around USD 1 million per satellite using the previously quoted USD 62 million (2018 figure) for a Falcon 9 launch. The second-generation Starlink satellites are substantially more advanced compared to the 1st generation. They are also heavier, weighing around a thousand kilograms. A Falcon 9 would only be able to launch around 20 generation 2 satellites (only considering the weight limitation), while a Falcon Heavy could lift ca. 60 2nd gen. satellites but also at a higher price point of USD 90 million (2018 figure). Thus the launch cost per satellite would be between USD 1.5 million using Falcon Heavy and USD 3.1 million using Falcon 9. Although the launch cost is based on price figures from 2018, the expected efficiency gained from re-use may have either kept the cost level or reduced it further as expected, particularly with Falcon Heavy.

Satellite businesses looking to launch small volumes of satellites, such as CubeSats, have a variety of strategies at their disposal to manage launch costs effectively. One widely adopted approach is participating in rideshare missions, where the expenses of a single launch vehicle are shared among multiple payloads, substantially reducing the cost for each operator. This method is particularly attractive due to its cost efficiency and the regularity of missions offered by, for example, SpaceX. Prices for rideshare missions can start from as low as a few thousand dollars for very small payloads (like CubeSats) to several hundred thousand dollars for larger small satellites. For example, SpaceX advertises rideshare prices starting at $1 million for payloads up to 200 kg. Alternatively, dedicated small launcher services cater specifically to the needs of small satellite operators, offering more tailored launch options in terms of timing and desired orbit. Companies such as Rocket Lab (USA) and Astra (USA) launch services have emerged, providing flexibility that rideshare missions might not, although at a slightly higher cost. However, these costs remain significantly lower than arranging a dedicated launch on a larger vehicle. For example, Rocket Lab’s Electron rocket, specializing in launching small satellites, offers dedicated launches with prices starting around USD 7 million for the entire launch vehicle carrying up to 300 kg. Astra has reported prices in the range of USD 2.5 million for a dedicated LEO launch with their (discontinued) Rocket 3 with payloads of up to 150 kg. The cost for individual small satellites will depend on their share of the payload mass and the specific mission requirements.

Satellite ground stations, which consist of arrays of phased-array antennas, are critical for managing the satellite constellation, routing internet traffic, and providing users with access to the satellite network. These stations are strategically located to maximize coverage and minimize latency, ensuring that at least one ground station is within the line of sight of satellites as they orbit the Earth. As of mid-2023, Starlink operated around 150 ground stations worldwide (also called Starlink Gateways), with 64 live and an additional 33 planned in the USA. The cost of constructing a ground station would be between USD 300 thousand to half a million not including the physical access point, also called the point-of-presence (PoP), and transport infrastructure connecting the PoP (and gateway) to the internet exchange where we find the internet service providers (ISPs) and the content delivery networks (CDNs). The Pop may add another USD 100 to 200 thousand to the ground infrastructure unit cost. The transport cost from the gateway to the Internet exchange can vary a lot depending on the gateway’s location.

Insurance is a critical component of the financial planning for a satellite project, covering risks associated with both the launch phase and the satellite’s operational period in orbit. These insurances are, in general, running at between 5% to 20% of the total project cost depending on the satellite value, the track record of the launch vehicle, mission complexity, and duration (i.e., typically 5 – 7 years for a LEO satellite at 500 km) and so forth. Insurance could be broken up into launch insurance and insurance covering the satellite once it is in orbit.

Operational costs, the Opex, include the day-to-day expenses of running the satellite, from staffing and technical support to ground station usage fees.

Regulatory and licensing fees, including frequency allocation and orbital slot registration, ensure the satellite operates without interfering with other space assets. Finally, at the end of the satellite’s operational life, costs associated with safely deorbiting the satellite are incurred to comply with space debris mitigation guidelines and ensure a responsible conclusion to the mission.

The total cost of an LEO satellite project can vary widely, influenced by the satellite’s complexity, mission goals, and lifespan. Effective project management and strategic decision-making are crucial to navigating these expenses, optimizing the project’s budget, and achieving mission success.

Figure 11 illustrates an LEO CubeSat orbiting above the Earth, capturing the satellite’s compact design and its role in modern space exploration and technology demonstration. Note that the CubeSat design comes in several standardized dimensions, with the reference design, also called 1U, being almost 1 thousandth of a cubic meter and weighing less than 1.33 kg. More advanced CubeSat satellites would typically be 6U or higher.

CubeSats (e.g., 1U, 3U, 6U, 12U):

  • Manufacturing Cost: Ranges from USD 50,000 for a simple 1U CubeSat to over USD 1 million for a more complex missions supported by 6U (or higher) CubeSat with advanced payloads (and 12U may again amount to several million US dollars).
  • Launch Cost: This can vary significantly depending on the launch provider and the rideshare opportunities, ranging from a few thousand dollars for a 1U CubeSat on a rideshare mission to several million dollars for a dedicated launch of larger CubeSats or small satellites.
  • Operational Costs: Ground station services, mission control, and data handling can add tens to hundreds of thousands of dollars annually, depending on the mission’s complexity and duration.

Small Satellites (25 kg up to 500 kg):

  • Manufacturing Cost: Ranges from USD 500,000 to over 10 million, depending on the satellite’s size, complexity, and payload requirements.
  • Launch Cost: While rideshare missions can reduce costs, dedicated launches for small satellites can range from USD 10 million to 62 million (e.g., Falcon 9) and beyond (e.g., USD 90 million for Falcon Heavy).
  • Operational Costs: These are similar to CubeSats, but potentially higher due to the satellite’s larger size and more complex mission requirements, reaching several hundred thousand to over a million dollars annually.

The range for the total project cost of a LEO satellite:

Given these considerations, the total cost range for a LEO satellite project can vary from as low as a few hundred thousand dollars for a simple CubeSat project utilizing rideshare opportunities and minimal operational requirements to hundreds of millions of dollars for more complex small satellite missions requiring dedicated launches and extensive operational support.

It is important to note that these are rough estimates, and the actual cost can vary based on specific mission requirements, technological advancements, and market conditions.

CAPACITY AND QUALITY

Figure 12 Satellite-based cellular capacity, or quality measured, by the unit or total throughput in Mbps is approximately driven by the amount of spectrum (in MHz) times the effective spectral efficiency (in Mbps/MHz/units) times the number of satellite beams resulting in cells on the ground.

The overall capacity and quality of satellite communication systems, given in Mbps, is on a high level, the product of three key factors: (i) the amount of frequency bandwidth in MHz allocated to the satellite operations multiplied by (ii) the effective spectral efficiency in Mbps per MHz over a unit satellite-beam coverage area multiplied by (iii) the number of satellite beams that provide the resulting terrestrial cell coverage. Thus, in other words:

Satellite Capacity (in Mbps) =
Frequency Bandwidth in MHz ×
Effective Spectral Efficiency in Mbps/MHz/Beam ×
Number of Beams (or Cells)

Consider a satellite system supporting 8 beams (and thus an equivalent of terrestrial coverage cells), each with 250 MHz allocated within the same spectral frequency range, can efficiently support ca. 680 Mbps per beam. This is achieved with an antenna setup that effectively provides a spectral efficiency of ~2.7 Mbps/MHz/cell (or beam) in the downlink (i.e., from the satellite to the ground). Moreover, the satellite typically will have another frequency and antenna configuration that establishes a robust connection to the ground station that connects to the internet via, for example, third-party internet service providers. The 680 Mbps is then shared among users that are within the satellite beam coverage, e.g., if you have 100 customers demanding a service, the speed each would experience on average would be around 7 Mbps. This may not seem very impressive compared to the cellular speeds we are used to getting with an LTE or 5G terrestrial cellular service. However, such speeds are, of course, much better than having no means of connecting to the internet.

Higher frequencies (i.e., in the GHz range) used to provide terrestrial cellular broadband services are in general quiet sensitive to the terrestrial environment and non-LoS propagation. It is a basic principle of physics that signal propagation characteristics, including the range and penetration capabilities of an electromagnetic waves, is inversely related to their frequency. Vegetation and terrain becomes an increasingly critical factor to consider in higher frequency propagation and the resulting quality of coverage. For example trees, forests, and other dense foliage can absorb and scatter radio waves, attenuating signals. The type and density of vegetation, along with seasonal changes like foliage density in summer versus winter, can significantly impact signal strength. Terrains often include varied topographies such as housing, hills, valleys, and flat plains, each affecting signal reach differently. For instance, housing, hilly or mountainous areas may cause signal shadowing and reflection, while flat terrains might offer less obstruction, enabling signals to travel further. Cellular mobile operators tend to like high frequencies (GHz) for cellular broadband services as it is possible to get substantially more system throughput in bits per second available to deliver to our demanding customers than at frequencies in the MHz range. As can be observed in Figure 12 above, we see that the frequency bandwidth is a multiplier for the satellite capacity and quality. Cellular mobile operators tend to “dislike” higher frequencies because of their poorer propagation conditions in their terrestrially based cellular networks resulting in the need for increased site densification at a significant incremental capital expense.

The key advantage of a LEO satellite is that the likelihood of a line-of-sight to a point on the ground is very high compared to establishing a line-of-sight for terrestrial cellular coverage that, in general, would be very low. In other words, the cellular signal propagation from an satellite closely approximates that of free space. Thus, all the various environmental signal loss factors we must consider for a standard terrestrial-based mobile network do not apply to our satellite having only to overcome the distance from the satellite antenna to the ground.

Let us first look at the satellite frequency component of the above satellite capacity, and quality, formula:

FREQUENCY SPECTRUM FOR SATELLITES.

The satellite frequency spectrum encompasses a range of electromagnetic frequencies allocated specifically for satellite communication. These frequencies are divided into bands, commonly known as L-band (e.g., mobile broadband), S-band (e.g., mobile broadband), C-band, X-band (e.g., mainly used by military), Ku-band (e.g., fixed broadband), Ka-band (e.g., fixed broadband), and V-band. Each serves different satellite applications due to its distinct propagation characteristics and capabilities. The spectrum bandwidth used by satellites refers to the width of the frequency range that a satellite system is licensed to use for transmitting and receiving signals.

Careful management of satellite spectrum bandwidth is critical to prevent interference with terrestrial communications systems. Since both satellite and terrestrial systems can operate on similar frequency ranges, there is a potential for crossover interference, which can degrade the performance of both systems. This is particularly important for bands like C-band and Ku-band, which are also used for terrestrial cellular networks and other applications like broadcasting.

Using the same spectrum for both satellite and terrestrial cellular coverage within the same geographical area is challenging due to the risk of interference. Satellites transmit signals over vast areas, and if those signals are on the same frequency as terrestrial cellular systems, they can overpower the local ground-based signals, causing reception issues for users on the ground. Conversely, the uplink signals from terrestrial sources can interfere with the satellite’s ability to receive communications from its service area.

Regulatory bodies such as the International Telecommunication Union (ITU) are crucial in mitigating these interference issues. They coordinate the allocation of frequency bands and establish regulations that govern their use. This includes defining geographical zones where certain frequencies may be used exclusively for either terrestrial or satellite services, as well as setting limits on signal power levels to minimize the chance of interference. Additionally, technology solutions like advanced filtering, beam shaping, and polarization techniques are employed to further isolate satellite communications from terrestrial systems, ensuring that both may coexist and operate effectively without mutual disruption.

The International Telecommunication Union (ITU) has designated several frequency bands for Fixed Satellite Services (FSS) and Mobile Satellite Services (MSS) that can be used by satellites operating in Low Earth Orbit (LEO). The specific bands allocated for satellite services, FSS and MSS, are determined by the ITU’s Radio Regulations, which are periodically updated to reflect global telecommunication’s evolving needs and address emerging technologies. Here are some of the key frequency bands commonly considered for FSS and MSS with LEO satellites:

V-Band 40 GHz to 75 GHz (microwave frequency range).
The V-band is appealing for Low Earth Orbit (LEO) satellite constellations designed to provide global broadband internet access. LEO satellites can benefit from the V-band’s capacity to support high data rates, which is essential for serving densely populated areas and delivering competitive internet speeds. The reduced path loss at lower altitudes, compared to GEO, also makes the V-band a viable option for LEO satellites. Due to the higher frequencies offered by V-band it also is significant more sensitive to atmospheric attenuation, (e.g., oxygen absorption around 60 GHz), including rain fade, which is likely to affect signal integrity. This necessitates the development of advanced technologies for adaptive coding and modulation, power amplification, and beamforming to ensure reliable communication under various weather conditions. Several LEO satellite operators have expressed an interest in operationalizing the V-band in their satellite constellations (e.g., StarLink, OneWeb, Kuiper, Lightspeed). This band should be regarded as an emergent LEO frequency band.

Ka-Band 17.7 GHz to 20.2 GHz (Downlink) & 27.5 GHz to 30.0 GHz (Uplink).
The Ka-band offers higher bandwidths, enabling greater data throughput than lower bands. Not surprising this band is favored by high-throughput satellite solutions. It is widely used by fixed satellite services (FSS). This makes it ideal for high-speed internet services. However, it is more susceptible to absorption and scattering by atmospheric particles, including raindrops and snowflakes. This absorption and scattering effect weakens the signal strength when it reaches the receiver. To mitigate rain fade effects in the Ka-band, satellite, and ground equipment must be designed with higher link margins, incorporating more powerful transmitters and more sensitive receivers. Additionally, adaptive modulation and coding techniques can be employed to adjust the signal dynamically in response to changing weather conditions. Overall, the system is more costly and, therefore, primarily used for satellite-to-earth ground station communications and high-performance satellite backhaul solutions.

For example, Starlink and OneWeb use the Ka-band to connect to satellite Earth gateways and point-of-presence, which connect to Internet Exchange and the wider internet. It is worth noticing that the terrestrial 5 G band n256 (26.5 to 29.5 GHz) falls within the Ka-band’s uplink frequency band. Furthermore, SES’s mPower satellites, operating at Middle Earth Orbit (MEO), operate exclusively in this band, providing internet backhaul services.

Ku-Band 12.75 GHz to 13.25 GHz (Downlink) & 14.0 GHz to 14.5 GHz (Uplink).
The Ku-band is widely used for FSS satellite communications, including fixed satellite services, due to its balance between bandwidth availability and susceptibility to rain fade. It is suitable for broadband services, TV broadcasting, and backhaul connections. For example, Starlink and OneWeb satellites are using this band to provide broadband services to earth-based customer terminals.

X-Band 7.25 GHz to 7.75 GHz (Downlink) & 7.9 GHz to 8.4 GHz (Uplink).
The X-band in satellite applications is governed by international agreements and national regulations to prevent interference between different services and to ensure the efficient use of the spectrum. The X-band is extensively used for secure military satellite communications, offering advantages like high data rates and relative resilience to jamming and eavesdropping. It supports a wide range of military applications, including mobile command, control, communications, computer, intelligence, surveillance, and reconnaissance (i.e., C4ISR) operations. Most defense-oriented satellites operate at geostationary orbit, ensuring constant coverage of specific geographic areas (e.g., Airbus Skynet constellations, Spain’s XTAR-EUR, and France’s Syracuse satellites). Most European LEO defense satellites, used primarily for reconnaissance, are fairly old, with more than 15 years since the first launch, and are limited in numbers (i.e., <10). The most recent European LEO satellite system is the French-based Multinational Space-based Imaging System (MUSIS) and Composante Spatiale Optique (CSO), where the first CSO components were launched in 2018. There are few commercial satellites utilizing the X-band.

C-Band 3.7 GHz to 4.2 GHz (Downlink) & 5.925 GHz to 6.425 GHz (Uplink)
C-band is less susceptible to rain fade and is traditionally used for satellite TV broadcasting, maritime, and aviation communications. However, parts of the C-band are also being repurposed for terrestrial 5G networks in some regions, leading to potential conflicts and the need for careful coordination. The C-band is primarily used in geostationary orbit (GEO) rather than Low Earth Orbit (LEO), due to the historical allocation of C-band for fixed satellite services (FSS) and its favorable propagation characteristics. I haven’t really come across any LEO constellation using the C-band. GEO FSS satellite operators using this band extensively are SES (Luxembourg), Intelsat (Luxembourg/USA), Eutelsat (France), Inmarsat (UK), etc..

S-Band 2.0 GHz to 4.0 GHz
S-band is used for various applications, including mobile communications, weather radar, and some types of broadband services. It offers a good compromise between bandwidth and resistance to atmospheric absorption. Both Omnispace (USA) and Globalstar (USA) LEO satellites operate in this band. Omnispace is also interesting as they have expressed intent to have LEO satellites supporting the 5G services in the band n256 (26.5 to 29.5 GHz), which falls within the uplink of the Ka-band.

L-Band 1.0 GHz to 2.0 GHz
L-band is less commonly used for fixed satellite services but is notable for its use in mobile satellite services (MSS), satellite phone communications, and GPS. It provides good coverage and penetration characteristics. Both Lynk Mobile (USA), offering Direct-2-Device, IoT, and M2M services, and Astrocast (Switzerland), with their IoT/M2M services, are examples of LEO satellite businesses operating in this band.

UHF 300 MHz to 3.0 GHz
The UHF band is more widely used for satellite communications, including mobile satellite services (MSS), satellite radio, and some types of broadband data services. It is favored for its relatively good propagation characteristics, including the ability to penetrate buildings and foliage. For example, Fossa Systems LEO pico-satellites (i.e., 1p form-factor) use this frequency for their IoT and M2M communications services.

VHF 30 MHz to 300 MHz

The VHF band is less commonly used in satellite communications for commercial broadband services. Still, it is important for applications such as satellite telemetry, tracking, and control (TT&C) operations and amateur satellite communications. Its use is often limited due to the lower bandwidth available and the higher susceptibility to interference from terrestrial sources. Swarm Technologies (USA and a SpaceX subsidiary) using 137-138 MHz (Downlink) and 148-150 MHz (Uplink). However, it appears that they have stopped taking new devices on their network. Orbcomm (USA) is another example of a satellite service provider using the VHF band for IoT and M2M communications. There is very limited capacity in this band due to many other existing use cases, and LEO satellite companies appear to plan to upgrade to the UHF band or to piggyback on direct-2-cell (or direct-2-device) satellite solutions, enabling LEO satellite communications with 3GPP compatible IoT and M2M devices.

SATELLITE ANTENNAS.

Satellites operating in Geostationary Earth Orbit (GEO), Medium Earth Orbit (MEO), and Low Earth Orbit (LEO) utilize a variety of antenna types tailored to their specific missions, which range from communication and navigation to observation (e.g., signal intelligence). The satellite’s applications influence the selection of an antenna, the characteristics of its orbit, and the coverage area required.

Antenna technology is intrinsically linked to spectral efficiency in satellite communications systems and of course any other wireless systems. Antenna designs influence how effectively a communication system can transmit and receive signals within a given frequency band, which is the essence of spectral efficiency (i.e., how much information per unit time in bits per second can I squeeze through my communications channel).

Thus, advancements in antenna technology are fundamental to improving spectral efficiency, making it a key area of research and development in the quest for more capable and efficient communication systems.

Parabolic dish antennas are prevalent for GEO satellites due to their high gain and narrow beam width, making them ideal for broadcasting and fixed satellite services. These antennas focus a tight beam on specific areas on Earth, enabling strong and direct signals essential for television, internet, and communication services. Horn antennas, while simpler, are sometimes used as feeds for larger parabolic antennas or for telemetry, tracking, and command functions due to their reliability. Additionally, phased array antennas are becoming more common in GEO satellites for their ability to steer beams electronically, offering flexibility in coverage and the capability to handle multiple beams and frequencies simultaneously.

Phased-array antennas are indispensable in for MEO satellites, such as those used in navigation systems like GPS (USA), BeiDou (China), Galileo (European), or GLONASS (Russian). These satellite constellations cover large areas of the Earth’s surface and can adjust beam directions dynamically, a critical feature given the satellites’ movement relative to the Earth. Patch antennas are also widely used in MEO satellites, especially for mobile communication constellations, due to their compact and low-profile design, making them suitable for mobile voice and data communications.

Phased-array antennas are very important for LEO satellites use cases as well, which include broadband communication constellations like Starlink and OneWeb. Their (fast) beam-steering capabilities are essential for maintaining continuous communication with ground stations and user terminals as the satellites quickly traverse the sky. The phased-array antenna also allow for optimizing coverage with both narrow as well as wider field of view (from the perspective of the satellite antenna) that allow the satellite operator to trade-off cell capacity and cell coverage.

Simpler Dipole antennas are employed for more straightforward data relay and telemetry purposes in smaller satellites and CubeSats, where space and power constraints are significant factors. Reflect array antennas, which offer a mix of high gain and beam steering capabilities, are used in specific LEO satellites for communication and observation applications (e.g., for signal intelligence gathering), combining features of both parabolic and phased array antennas.

Mission specific requirements drive the choice of antenna for a satellite. For example, GEO satellites often use high-gain, narrowly focused antennas due to their fixed position relative to the Earth, while MEO and LEO satellites, which move relatively closer to the Earth’s surface, require antennas capable of maintaining stable connections with moving ground terminals or covering large geographical areas.

Advanced antenna technologies such as beamforming, phased-arrays, and Multiple In Multiple Out (MMO) antenna configurations are crucial in managing and utilizing the spectrum more efficiently. They enable precise targeting of radio waves, minimizing interference, and optimizing bandwidth usage. This direct control over the transmission path and signal shape allows more data (bits) to be sent and received within the same spectral space, effectively increasing the communication channel’s capacity. In particular, MIMO antenna configurations and advanced antenna beamforming have enabled terrestrial mobile cellular access technologies (e.g., LTE and 5G) to quantum leap the effective spectral efficiency, broadband speed and capacity orders of magnitude above and beyond older technologies of 2G and 3G. Similar principles are being deployed today in modern advanced communications satellite antennas, providing increased capacity and quality within the satellite cellular coverage area provided by the satellite beam.

Moreover, antenna technology developments like polarization and frequency reuse directly impact a satellite system’s ability to maximize spectral resources. Allowing simultaneous transmissions on the same frequency through different polarizations or spatial separations effectively double the capacity without needing additional spectrum.

WHERE DO WE END UP.

If all current commercial satellite plans were realized, within the next decade, we would have more, possibly substantially more than 65 thousand satellites circling Earth. Today, that number is less than 10 thousand, with more than half that number realized by StarLink’s LEO constellation. Imagine the increase in, and the amount of, space debris circling Earth within the next 10 years. This will likely pose a substantial increase in operational risk for new space missions and will have to be addressed urgently.

Over the next decade, we may have at least 2 major LEO satellite constellations. One from Starlink with an excess of 12 thousand satellites, and one from China, the Guo Wang, the state network, likewise with 12 thousand LEO satellites. One global satellite constellation is from an American-based commercial company; the other is a worldwide satellite constellation representing the Chinese state. It would not be too surprising to see that by 2034, the two satellite constellations will divide Earth in part, being serviced by a commercial satellite constellation (e.g., North America, Europe, parts of the Middle East, some of APAC including India, possibly some parts of Africa). Another part will likely be served by a Chinese-controlled LEO constellation providing satellite broadband service to China, Russia, significant parts of Africa, and parts of APAC.

Over the next decade, satellite services will undergo transformative advancements, reshaping the architecture of global communication infrastructures and significantly impacting various sectors, including broadband internet, global navigation, Earth observation, and beyond. As these services evolve, we should anticipate major leaps in satellite technologies, driven by innovation in propulsion systems, miniaturization of technology, advancements in onboard processing capabilities, increasing use of AI and machine learning leapfrogging satellites operational efficiency and performance, breakthrough in material science reducing weight and increasing packing density, leapfrogs in antenna technology, and last but not least much more efficient use of the radio frequency spectrum. Moreover, we will see the breakthrough innovation that will allow better co-existence and autonomous collaboration of frequency spectrum utilization between non-terrestrial and terrestrial networks reducing the need for much regulatory bureaucracy that might anyway be replaced by decentralized autonomous organizations (DAOs) and smart contracts. This development will be essential as satellite constellations are being integrated into 5G and 6G network architectures as the non-terrestrial network cellular access component. This particular topic, like many in this article, is worth a whole new article on its own.

I expect that over the next 10 years we will see electronically steerable phased-array antennas, as a notable advancement. These would offer increased agility and efficiency in beamforming and signal direction. Their ability to swiftly adjust beams for optimal coverage and connectivity without physical movement makes them perfect for the dynamic nature of Low Earth Orbit (LEO) satellite constellations. This technology will becomes increasingly cost-effective and energy-efficient, enabling widespread deployment across various satellite platforms (not only LEO designs). The advance in phased-array antenna technology will facilitate substantial increase in the satellite system capacity by increasing the number of beams, the variation on beam size (possibly down to a customer ground station level), and support multi-band operations within the same antenna.

Another promising development is the integration of metamaterials in antenna design, which will lead to more compact, flexible, and lightweight antennas. The science of metamaterials is super interesting and relates to manufacturing artificial materials to have properties not found in naturally occurring materials with unique electromagnetic behaviors arising from their internal structure. Metamaterial antennas is going to offer superior performance, including better signal control and reduced interference, which is crucial for maintaining high-quality broadband connections. These materials are also important for substantially reducing the weight of the satellite antenna, while boosting its performance. Thus, the technology will also support bringing the satellite launch cost down dramatically.

Although primarily associated MIMO antennas with terrestrial networks, I would also expect that massive MIMO technology will find applications in satellite broadband systems. Satellite systems, just like ground based cellular networks, can significantly increase their capacity and efficiency by utilizing many antenna elements to simultaneously communicate with multiple ground terminals. This could be particularly transformative for next-generation satellite networks, supporting higher data rates and accommodating more users. The technology will increase the capacity and quality of the satellite system dramatically as it has done on terrestrial cellular networks.

Furthermore, advancements in onboard processing capabilities will allow satellites to perform more complex signal processing tasks directly in space, reducing latency and improving the efficiency of data transmission. Coupled with AI and machine learning algorithms, future satellite antennas could dynamically optimize their operational parameters in real-time, adapting to changes in the network environment and user demand.

Additionally, research into quantum antenna technology may offer breakthroughs in satellite communication, providing unprecedented levels of sensitivity and bandwidth efficiency. Although still early, quantum antennas could revolutionize signal reception and transmission in satellite broadband systems. In the context of LEO satellite systems, I am particularly excited about utilizing the Rydberg Effect to enhance system sensitivity could lead to massive improvements. The heightened sensitivity of Rydberg atoms to electromagnetic fields could be harnessed to develop ultra-sensitive detectors for radio frequency (RF) signals. Such detectors could surpass the performance of traditional semiconductor-based devices in terms of sensitivity and selectivity, enabling satellite systems to detect weaker signals, improve signal-to-noise ratios, and even operate effectively over greater distances or with less power. Furthermore, space could potentially be the near-ideal environment for operationalizing Rydberg antenna and communications systems as space had near-perfect vacuum, very low-temperatures (in Earth shadow at least or with proper thermal management), relatively free of electromagnetic radiation (compared to Earth), as well as its micro-gravity environment that may facilitate long-range “communications” between Rydberg atoms. This particular topic may be further out in the future than “just” a decade from now, although it may also be with satellites we will see the first promising results of this technology.

One key area of development will be the integration of LEO satellite networks with terrestrial 5G and emerging 6G networks, marking a significant step in the evolution of Non-Terrestrial Network (NTN) architectures. This integration promises to deliver seamless, high-speed connectivity across the globe, including in remote and rural areas previously underserved by traditional broadband infrastructure. By complementing terrestrial networks, LEO satellites will help achieve ubiquitous wireless coverage, facilitating a wide range of applications and use cases from high-definition video streaming to real-time IoT data collection.

The convergence of LEO satellite services with 5G and 6G will also spur network management and orchestration innovation. Advanced techniques for managing interference, optimizing handovers between terrestrial and non-terrestrial networks, and efficiently allocating spectral resources will be crucial. It would be odd not to mention it here, so artificial intelligence and machine learning algorithms will, of course, support these efforts, enabling dynamic network adaptation to changing conditions and demands.

Moreover, the next decade will likely see significant improvements in the environmental sustainability of LEO satellite operations. Innovations in satellite design and materials, along with more efficient launch vehicles and end-of-life deorbiting strategies, will help mitigate the challenges of space debris and ensure the long-term viability of LEO satellite constellations.

In the realm of global connectivity, LEO satellites will have bridged the digital divide, offering affordable and accessible internet services to billions of people worldwide unconnected today. In 2023 the estimate is that there are about 3 billion people, almost 40% of all people in the world today, that have never used internet. In the next decade, it must be our ambition that with LEO satellite networks this number is brought down to very near Zero. This will have profound implications for education, healthcare, economic development, and global collaboration.

FURTHER READING.

  1. A. Vanelli-Coralli, N. Chuberre, G. Masini, A. Guidotti, M. El Jaafari, “5G Non-Terrestrial Networks.”, Wiley (2024). A recommended reading for deep diving into NTN networks of satellites, typically the LEO kind, and High-Altitude Platform Systems (HAPS) such as stratospheric drones.
  2. I. del Portillo et al., “A technical comparison of three low earth orbit satellite constellation systems to provide global broadband,” Acta Astronautica, (2019).
  3. Nils Pachler et al., “An Updated Comparison of Four Low Earth Orbit Satellite Constellation Systems to Provide Global Broadband” (2021).
  4. Starlink, “Starlink specifications” (Starlink.com page). The following Wikipedia resource is quite good as well: Starlink.
  5. Quora, “How much does a satellite cost for SpaceX’s Starlink project and what would be the cheapest way to launch it into space?” (June 2023). This link includes a post from Elon Musk commenting on the cost involved in manufacturing the Starlink satellite and the cost of launching SpaceX’s Falcon 9 rocket.
  6. Michael Baylor, “With Block 5, SpaceX to increase launch cadence and lower prices.”, nasaspaceflight.com (May, 2018).
  7. Gwynne Shotwell, TED Talk from May 2018. She quotes here a total of USD 10 billion as a target for the 12,000 satellite network. This is just an amazing visionary talk/discussion about what may happen by 2028 (in 4-5 years ;-).
  8. Juliana Suess, “Guo Wang: China’s Answer to Starlink?”, (May 2023).
  9. Makena Young & Akhil Thadani, “Low Orbit, High Stakes, All-In on the LEO Broadband Competition.”, Center for Strategic & International Studies CSIS, (Dec. 2022).
  10. AST SpaceMobile website: https://ast-science.com/ Constellation Areas: Internet, Direct-to-Cell, Space-Based Cellular Broadband, Satellite-to-Cellphone. 243 LEO satellites planned. 2 launched.
  11. Lynk Global website: https://lynk.world/ (see also FCC Order and Authorization). It should be noted that Lynk can operate within 617 to 960 MHz (Space-to-Earth) and 663 to 915 MHz (Earth-to-Space). However, only outside the USA. Constellation Area: IoT / M2M, Satellite-to-Cellphone, Internet, Direct-to-Cell. 8 LEO satellites out of 10 planned.
  12. Omnispace website: https://omnispace.com/ Constellation Area: IoT / M2M, 5G. Ambition to have the world’s first global 5G non-terrestrial network. Initial support 3GPP-defined Narrow-Band IoT radio interface. Planned 200 LEO and <15 MEO satellites. So far, only 2 satellites have been launched.
  13. NewSpace Index: https://www.newspace.im/ I find this resource to have excellent and up-to-date information on commercial satellite constellations.
  14. R.K. Mailloux, “Phased Array Antenna Handbook, 3rd Edition”, Artech House, (September 2017).
  15. A.K. Singh, M.P. Abegaonkar, and S.K. Koul, “Metamaterials for Antenna Applications”, CRC Press (September 2021).
  16. T.L. Marzetta, E.G. Larsson, H. Yang, and H.Q. Ngo, “Fundamentals of Massive MIMO”, Cambridge University Press, (November 2016).
  17. G.Y. Slepyan, S. Vlasenko, and D. Mogilevtsev, “Quantum Antennas”, arXiv:2206.14065v2, (June 2022).
  18. R. Huntley, “Quantum Rydberg Receiver Shakes Up RF Fundamentals”, EE Times, (January 2022).
  19. Y. Du, N. Cong, X. Wei, X. Zhang, W. Lou, J. He, and R. Yang, “Realization of multiband communications using different Rydberg final states”, AIP Advances, (June 2022). Demonstrating the applicability of the Rydberg effect in digital transceivers in the Ku and Ka bands.

ACKNOWLEDGEMENT.

I greatly acknowledge my wife, Eva Varadi, for her support, patience, and understanding during the creative process of writing this article.

Stratospheric Drones & Low Earth Satellites: Revolutionizing Terrestrial Rural Broadband from the Skies?

Posted on by
5

“From an economic and customer experience standpoint, deploying stratospheric drones may be significantly more cost effective than establishing extra terrestrial infrastructures”.

This article, in a different and somewhat shorter format, has also been published by New Street Research under the title “Stratospheric drones: A game changer for rural networks?”. You will need to register with New Street Research to get access.

As a mobile cellular industry expert and a techno-economist, the first time I was presented with the concept of stratospheric drones, I feel the butterflies in my belly. That tingling feeling that I was seeing something that could be a huge disruptor of how mobile cellular networks are being designed and built. Imagine getting rid of the profitability-challenged rural cellular networks (i.e., the towers, the energy consumption, the capital infrastructure investments), and, at the same time, offering much better quality to customers in rural areas than is possible with the existing cellular network we have deployed there. A technology that could fundamentally change the industry’s mobile cellular cost structure for the better at a quantum leap in quality and, in general, provide economical broadband services to the unconnected at a fraction of the cost of our traditional ways of building terrestrial cellular coverage.

Back in 2015, I got involved with Deutsche Telekom AG Group Technology, under the leadership of Bruno Jacobfeuerborn, in working out the detailed operational plans, deployment strategies, and, of course, the business case as well as general economics of building a stratospheric cellular coverage platform from scratch with the UK-based Stratospheric Platform Ltd [2] in which Deutsche Telekom is an investor. The investment thesis was really in the way we expected the stratospheric high-altitude platform to make a large part of mobile operators’ terrestrial rural cellular networks obsolete and how it might strengthen mobile operator footprints in countries where rural and remote coverage was either very weak or non-existing (e.g., The USA, an important market for Deutsche Telekom AG).

At the time, our thoughts were to have an operational stratospheric coverage platform operationally by 2025, 10 years after kicking off the program, with more than 100 high-altitude platforms covering a major Western European country serving rural areas. As it so often is, reality is unforgiving, as it often is with genuinely disruptive ideas. Getting to a stage of deployment and operation at scale of a high-altitude platform is still some years out due to the lack of maturity of the flight platform, including regulatory approvals for operating a HAP network at scale, increasing the operating window of the flight platform, fueling, technology challenges with the advanced antenna system, being allowed to deployed terrestrial-based cellular spectrum above terra firma, etc. Many of these challenges are progressing well, although slowly.

Globally, various companies are actively working on developing stratospheric drones to enhance cellular coverage. These include aerospace and defense giants like Airbus, advancing its Zephyr drone, and BAE Systems, collaborating with Prismatic for their PHASA-35 UAV. One of the most exciting HAPS companies focusing on developing world-leading high-altitude aircraft that I have come across during my planning work on how to operationalize a Stratospheric cellular coverage platform is the German company Leichtwerk AG, which has their hydrogen-fueled StratoStreamer as well as a solar-powered platform under development with the their StratoStreamer being close to production-ready. Telecom companies like Deutsche Telekom AG and BT Group are experimenting with hydrogen-powered drones in partnership with Stratospheric Platforms Limited. Through its subsidiary HAPSMobile, SoftBank is also a significant player with its Sunglider project. Additionally, entities like China Aerospace Science and Technology Corporation and Cambridge Consultants contribute to this field by co-developing enabling technologies (e.g., advanced phased-array antenna, fuel technologies, material science, …) critical for the success and deployability of high-altitude platforms at scale, aiming to improve connectivity in rural, remote, and underserved areas.

The work on integrating High Altitude Platform (HAP) networks with terrestrial cellular systems involves significant coordination with international regulatory bodies like the International Telecommunication Union Radiocommunication Sector (ITU-R) and the World Radiocommunication Conference (WRC). This process is crucial for securing permission to reuse terrestrial cellular spectrum in the stratosphere. Key focus areas include negotiating the allocation and management of frequency bands for HAP systems, ensuring they don’t interfere with terrestrial networks. These efforts are vital for successfully deploying and operating HAP systems, enabling them to provide enhanced connectivity globally, especially in rural areas where terrestrial cellular frequencies are already in use and remote and underserved regions. At the latest WRC-2023 conference, Softbank successfully gained approval within the Asia-Pacific region to use mobile spectrum bands for stratospheric drone-based mobile broadband cellular services.

Most mobile operators have at least 50% of their cellular network infrastructure assets in rural areas. While necessary for providing the coverage that mobile customers have come to expect everywhere, these sites carry only a fraction of the total mobile traffic. Individually, rural sites have poor financial returns due to their proportional operational and capital expenses.

In general, the Opex of the cellular network takes up between 50% and 60% of the Technology Opex, and at least 50% of that can be attributed to maintaining and operating the rural part of the radio access network. Capex is more cyclical than Opex due to, for example, the modernization of radio access technology. Nevertheless, over a typical modernization cycle (5 to 7 years), the rural network demands a little bit less but a similar share of Capex overall as for Opex. Typically, the Opex share of the rural cellular network may be around 10% of the corporate Opex, and its associated total cost is between 12% and 15% of the total expenses.

The global telecom towers market size in 2023 is estimated at ca. 26+ billion euros, ca. 2.5% of total telecom turnover, with a projected growth of CAGR 3.3% from now to 2030. The top 10 Tower management companies manage close to 1 million towers worldwide for mobile CSPs. Although many mobile operators have chosen to spin off their passive site infrastructure, there are still some remaining that may yet to spin off their cellular infrastructure to one of many Tower management companies, captive or independent, such as American Tower (224,019+ towers), Cellnex Telecom (112,737+ towers), Vantage Towers (46,100+ towers), GD Towers (+41,600 towers), etc…

IMAGINE.

Focusing on the low- or no-profitable rural cellular coverage.

Imagine an alternative coverage technology to the normal cellular one all mobile operators are using that would allow them to do without the costly and low-profitable rural cellular network they have today to satisfy their customers’ expectations of high-quality ubiquitous cellular coverage.

For the alternative technology to be attractive, it would need to deliver at least the same quality and capacity as the existing terrestrial-based cellular coverage for substantially better economics.

If a mobile operator with a 40% EBITDA margin did not need its rural cellular network, it could improve its margin by a sustainable 5% and increase its cash generation in relative terms by 50% (i.e., from 0.2×Revenue to 0.3×Revenue), assuming a capex-to-revenue ratio of 20% before implementing the technology being reduced to 15% after due to avoiding modernization and capacity investments in the rural areas.

Imagine that the alternative technology would provide a better cellular quality to the consumer for a quantum leap reduction of the associated cost structure compared to today’s cellular networks.

Such an alternative coverage technology might also impact the global tower companies’ absolute level of sustainable tower revenues, with a substantial proportion of revenue related to rural site infrastructure being at risk.

Figure 1 An example of an unmanned autonomous stratospheric coverage platform. Source: Cambridge Consultants presentation (see reference [2]) based on their work with Stratospheric Platforms Ltd (SPL) and SPL’s innovative high-altitude coverage platform.

TERRESTRIAL CELLULAR RURAL COVERAGE – A MATTER OF POOR ECONOMICS.

When considering the quality we experience in a terrestrial cellular network, a comprehensive understanding of various environmental and physical factors is crucial to predicting the signal quality accurately. All these factors generally work against cellular signal propagation regarding how far the signal can reach from the transmitting cellular tower and the achievable quality (e.g., signal strength) that a customer can experience from a cellular service.

Firstly, the terrain plays a significant role. Rural landscapes often include varied topographies such as hills, valleys, and flat plains, each affecting signal reach differently. For instance, hilly or mountainous areas may cause signal shadowing and reflection, while flat terrains might offer less obstruction, enabling signals to travel further.

At higher frequencies (i.e., above 1 GHz), vegetation becomes an increasingly critical factor to consider. Trees, forests, and other dense foliage can absorb and scatter radio waves, attenuating signals. The type and density of vegetation, along with seasonal changes like foliage density in summer versus winter, can significantly impact signal strength.

The height and placement of transmitting and receiving antennas are also vital considerations. In rural areas, where there are fewer tall buildings, the height of the antenna can have a pronounced effect on the line of sight and, consequently, on the signal coverage and quality. Elevated antennas mitigate the impact of terrain and vegetation to some extent.

Furthermore, the lower density of buildings in rural areas means fewer reflections and less multipath interference than in urban environments. However, larger structures, such as farm buildings or industrial facilities, must be factored in, as they can obstruct or reflect signals.

Finally, the distance between the transmitter and receiver is fundamental to signal propagation. With typically fewer cell towers spread over larger distances, understanding how signal strength diminishes with distance is critical to ensuring reliable coverage at a high quality, such as high cellular throughput, as the mobile customer expects.

The typical way for a cellular operator to mitigate the environmental and physical factors that inevitably result in loss of signal strength and reduced cellular quality (i.e., sub-standard cellular speed) is to build more sites and thus incur increasing Capex and Opex in areas that in general will have poor economical payback associated with any cellular assets. Thus, such investments make an already poor economic situation even worse as the rural cellular network generally would have very low utilization.

Figure 2 Cellular capacity or quality measured by the unit or total throughput is approximately driven by the amount of spectrum (in MHz) times the effective spectral efficiency (in Mbps/MHz/units) times the number of cells or capacity units deployed. When considering the effective spectral efficiency, one needs to consider the possible “boost” that a higher order MiMo or Advanced Antenna System will bring over and above the Single In Single Out (SISO) antenna would result in.

As our alternative technology also would need to provide at least the same quality and capacity it is worth exploring what can be expected in terms of rural terrestrial capacity. In general, we have that the cellular capacity (and quality) can be written as (also shown in Figure 2 above):

Throughput (in Mbps) =
Spectral Bandwidth in MHz ×
Effective Spectral Efficiency in Mbps/MHz/Cell ×
Number of Cells

We need to keep in mind that an additional important factor when considering quality and capacity is that the higher the operational frequency, the lower the radius (all else being equal). Typically, we can improve the radius at higher frequencies by utilizing advanced antenna beam forming, that is, concentrate the radiated power per unit coverage area, which is why you will often hear that the 3.6 GHz downlink coverage radius is similar to that of 1800 MHz (or PCS). This 3.6 GHz vs. 1.8 GHz coverage radius comparison is made when not all else is equal. Comparing a situation where the 1800 MHz (or PCS) radiated power is spread out over the whole coverage area compared to a coverage situation where the 3.6 GHz (or C-band in general) solution makes use of beamforming, where the transmitted energy density is high, allowing to reach the customer at a range that would not be possible if the 3.6 GHz radiated power would have been spread out over the cell like the example of the 1800 MHz.

As an example, take an average Western European rural 5G site with all cellular bands between 700 and 2100 MHz activated. The site will have a total of 85 MHz DL and 75 MHz UL, with a 10 MHz difference between DL and UL due to band 38 Supplementary Downlink SDL) operational on the site. In our example, we will be optimistic and assume that the effective spectral efficiency is 2 Mbps per MHz per cell (average over all bands and antenna configurations), which would indicate a fair amount of 4×4 and 8×8 MiMo antenna systems deployed. Thus, the unit throughput we would expect to be supplied by the terrestrial rural cell would be 170 Mbps (i.e., 85 MHz × 2.0 Mbps/MHz/Cell). With a rural cell coverage radius between 2 and 3 km, we then have an average throughput per square kilometer of 9 Mbps/km2. Due to the low demand and high-frequency bandwidth per active customer, DL speeds exceeding 100+ Mbps should be relatively easy to sustain with 5G standalone, with uplink speeds being more compromised due to larger coverage areas. Obviously, the rural quality can be improved further by deploying advanced antenna systems and increasing the share of higher-order MiMo antennas in general, as well as increasing the rural site density. However, as already pointed out, this would not be an economically reasonable approach.

THE ADVANTAGE OF SEEING FROM ABOVE.

Figure 3 illustrates the difference between terrestrial cellular coverage from a cell tower and that of a stratospheric drone or high-altitude platform (“Antenna-in-the-Sky”). The benefit of seeing the world from above is that environmental and physical factors have substantially less impact on signal propagation and quality primarily being impacted by distance as it approximates free space propagation. This situation is very different for a terrestrial-based cellular tower with its radiated signal being substantially impacted by the environment as well as physical factors.

It may sound silly to talk about an alternative coverage technology that could replace the need for the cellular tower infrastructure that today is critical for providing mobile broadband coverage to, for example, rural areas. What alternative coverage technologies should we consider?

If, instead of relying on terrestrial-based tower infrastructure, we could move the cellular antenna and possibly the radio node itself to the sky, we would have a situation where most points of the ground would be in the line of sight to the “antenna-in-the-sky.” The antenna in the sky idea is a game changer in terms of coverage itself compared to conventional terrestrial cellular coverage, where environmental and physical factors dramatically reduce signal propagation and signal quality.

The key advantage of an antenna in the sky (AIS) is that the likelihood of a line-of-sight to a point on the ground is very high compared to establishing a line-of-sight for terrestrial cellular coverage that, in general, would be very low. In other words, the cellular signal propagation from an AIS closely approximates that of free space. Thus, all the various environmental signal loss factors we must consider for a standard terrestrial-based mobile network do not apply to our antenna in the sky.

Over the last ten years, we have gotten several technology candidates for our antenna-in-the-sky solution, aiming to provide terrestrial broadband services as a substitute, or enhancement, for terrestrial mobile and fixed broadband services. In the following, I will describe two distinct types of antenna-in-the-sky solutions: (a) Low Earth Orbit (LEO) satellites, operating between 500 to 2000 km above Earth, that provide terrestrial broadband services such as we know from Starlink (SpaceX), OneWeb (Eutelsat Group), and Kuiper (Amazon), and (b) So-called, High Altitude Platforms (HAPS), operating at altitudes between 15 to 30 km (i.e., in the stratosphere). Such platforms are still in the research and trial stages but are very promising technologies to substitute or enhance rural network broadband services. The HAP is supposed to be unmanned, highly autonomous, and ultimately operational in the stratosphere for an extended period (weeks to months), fueled by green hydrogen and possibly solar. The high-altitude platform is thus also an unmanned aerial vehicle (UAV), although I will use the term stratospheric drone and HAP interchangeably in the following.

Low Earth Orbit (LEO) satellites and High Altitude Platforms (HAPs) represent two distinct approaches to providing high-altitude communication and observation services. LEO satellites, operating between 500 km and 2,000 km above the Earth, orbit the planet, offering broad global coverage. The LEO satellite platform is ideal for applications like satellite broadband internet, Earth observation, and global positioning systems. However, deploying and maintaining these satellites involves complex, costly space missions and sophisticated ground control. Although, as SpaceX has demonstrated with the Starlink LEO satellite fixed broadband platform, the unitary economics of their satellites significantly improve by scale when the launch cost is also considered (i.e., number of satellites).

Figure 4 illustrates a non-terrestrial network architecture consisting of a Low Earth Orbit (LEO) satellite constellation providing fixed broadband services to terrestrial users. Each hexagon represents a satellite beam inside the larger satellite coverage area. Note that, in general, there will be some coverage overlap between individual satellites, ensuring a continuous service including interconnected satellites. The user terminal (UT) dynamically aligns itself, aiming at the best quality connection provided by the satellites within the UT field of vision.

Figure 4 Illustrating a Non-Terrestrial Network consisting of a Low Earth Orbit (LEO) satellite constellation providing fixed broadband services to terrestrial users (e.g., Starlink, Kuiper, OneWeb,…). Each hexagon represents a satellite beam inside the larger satellite coverage area. Note that, in general, there will be some coverage overlap between individual satellites, ensuring a continuous service. The operating altitude of a LEO satellite constellation is between 300 and 2,000 km. It is assumed that the satellites are interconnected, e.g., laser links. The User Terminal antenna (UT) is dynamically orienting itself after the best line-of-sight (in terms of signal quality) to a satellite within UT’s field-of-view (FoV). The FoV has not been shown in the picture above so as not to overcomplicate the illustration. It should be noted just like with the drone it is possible to integrate the complete gNB on the LEO satellite. There might even be applications (e.g., defense, natural & unnatural disaster situations, …) where a standalone 5G SA core is integrated.

On the other hand, HAPs, such as unmanned (autonomous) stratospheric drones, operate at altitudes of approximately 15 km to 30 km in the stratosphere. Unlike LEO satellites, the stratospheric drone can hover or move slowly over specific areas, often geostationary relative to the Earth’s surface. This characteristic makes them more suitable for localized coverage tasks like regional broadband, surveillance, and environmental monitoring. The deployment and maintenance of the stratospheric drones are managed from the Earth’s surface and do not require space launch capabilities. Furthermore, enhancing and upgrading the HAPs is straightforward, as they will regularly be on the ground for fueling and maintenance. Upgrades are not possible with an operational LEO satellite solution where any upgrade would have to wait on a subsequent generation and new launch.

Figure 5 illustrates the high-level network architecture of an unmanned autonomous stratospheric drone-based constellation providing terrestrial cellular broadband services to terrestrial mobile users delivered to their normal 5G terminal equipment. Each hexagon represents a beam arising from the phased-array antenna integrated into the drone’s wingspan. To deliver very high-availability services to a rural area, one could assign three HAPs to cover a given area. The drone-based non-terrestrial network is drawn consistent with the architectural radio access network (RAN) elements from Open RAN, e.g., Radio Unit (RU), Distributed Unit (DU), and Central Unit (CU). It should be noted that the whole 5G gNB (the 5G NodeB), including the CU, could be integrated into the stratospheric drone, and in fact, so could the 5G standalone (SA) packet core, enabling full private mobile 5G networks for defense and disaster scenarios or providing coverage in very remote areas with little possibility of ground-based infrastructure (e.g., the arctic region, or desert and mountainous areas).

Figure 5 illustrates a Non-Terrestrial Network consisting of a stratospheric High Altitude Platform (HAP) drone-based constellation providing terrestrial Cellular broadband services to terrestrial mobile users delivered to their normal 5G terminal equipment. Each hexagon represents a beam inside the larger coverage area of the stratospheric drone. To deliver very high-availability services to a rural area, one could assign three HAPs to cover a given area. The operating altitude of a HAP constellation is between 10 to 50 km with an optimum of around 20 km. It is assumed that there is inter-HAP connectivity, e.g., via laser links. Of course, it is also possible to contemplate having the gNB (full 5G radio node) in the stratospheric drone entirely, which would allow easier integration with LEO satellite backhauls, for example. There might even be applications (e.g., defense, natural & unnatural disaster situations, …) where a standalone 5G SA core is integrated.

The unique advantage of the HAP operating in the stratosphere is (1) The altitude is advantageous for providing wider-area cellular coverage with a near-ideal quality above and beyond what is possible with conventional terrestrial-based cellular coverage because of very high line-of-sight likelihood due to less environment and physical issues that substantially reduces the signal propagation and quality of a terrestrial coverage solution, and (2) More stable atmospheric conditions characterize the stratosphere compared to the troposphere below it. This stability allows the stratospheric drone to maintain a consistent position and altitude with less energy expenditure. The stratosphere offers more consistent and direct sunlight exposure for a solar-powered HAP with less atmospheric attenuation. Moreover, due to the thinner atmosphere at stratospheric altitudes, the stratospheric drone will experience a lower air resistance (drag), increasing the energy efficiency and, therefore, increasing the operational airtime.

Figure 6 illustrates Leichtwerk AG’s StratoStreamer HAP design that is near-production ready. Leichtwerk AG works closely together with AESA towards the type certificate that would make it possible to operationalize a drone constellation in Europe. The StratoStreamer has a wingspan of 65 meter and can carry a payload of 100+ kg. Courtesy: Leichtwerk AG.

Each of these solutions has its unique advantages and limitations. LEO satellites provide extensive coverage but come with higher operational complexities and costs. HAPs offer more focused coverage and are easier to manage, but they need the global reach of LEO satellites. The choice between these two depends on the specific requirements of the intended application, including coverage area, budget, and infrastructure capabilities.

In an era where digital connectivity is indispensable, stratospheric drones could emerge as a game-changing technology. These unmanned (autonomous) drones, operating in the stratosphere, offer unique operational and economic advantages over terrestrial networks and are even seen as competitive alternatives to low earth orbit (LEO) satellite networks like Starlink or OneWeb.

STRATOSPHERIC DRONES VS TERRESTRIAL NETWORKS.

Stratospheric drones positioned much closer to the Earth’s surface than satellites, provide distinct signal strength and latency benefits. The HAP’s vantage point in the stratosphere (around 20 km above the Earth) ensures a high probability of line-of-sight with terrestrial user devices, mitigating the adverse effects of terrain obstacles that frequently challenge ground-based networks. This capability is particularly beneficial in rural areas in general and mountainous or densely forested areas, where conventional cellular towers struggle to provide consistent coverage.

Why the stratosphere? The stratosphere is the layer of Earth’s atmosphere located above the troposphere, which is the layer where weather occurs. The stratosphere is generally characterized by stable, dry conditions with very little water vapor and minimal horizontal winds. It is also home to the ozone layer, which absorbs and filters out most of the Sun’s harmful ultraviolet radiation. It is also above the altitude of commercial air traffic, which typically flies at altitudes ranging from approximately 9 to 12 kilometers (30,000 to 40,000 feet). These conditions (in addition to those mentioned above) make operating a stratospheric platform very advantageous.

Figure 6 illustrates the coverage fundamentals of (a) a terrestrial cellular radio network with the signal strength and quality degrading increasingly as one moves away from the antenna and (b) the terrestrial coverage from a stratospheric drone (antenna in the sky) flying at an altitude of 15 to 30 km. The stratospheric drone, also called a High-Altitude Platform (HAP), provides near-ideal signal strength and quality due to direct line-of-sight (LoS) with the ground, compared to the signal and quality from a terrestrial cellular site that is influenced by its environment and physical factors and the fact that LoS is much less likely in a conventional terrestrial cellular network. It is worth keeping in mind that the coverage scenarios where a stratospheric drone and a low earth satellite may excel in particular are in rural areas and outdoor coverage in more dense urban areas. In urban areas, the clutter, or environmental features and objects, will make line-of-site more challenging, impacting the strength and quality of the radio signals.

Figure 6 The chart above illustrates the coverage fundamentals of (a) a terrestrial cellular radio network with the signal strength and quality degrading increasingly as one moves away from the antenna and (b) the terrestrial coverage from a stratospheric drone (antenna in the sky) flying at an altitude of 15 to 30 km. The stratospheric drone, also called a High Altitude Platform (HAP), provides near-ideal signal strength and quality due to direct line-of-sight (LoS) with the ground, compared to the signal & quality from a terrestrial cellular site that is influenced by its environment and physical factors and the fact that LoS is much less likely in a conventional terrestrial cellular network.

From an economic and customer experience standpoint, deploying stratospheric drones may be significantly more cost-effective than establishing extensive terrestrial infrastructure, especially in remote or rural areas. The setup and operational costs of cellular towers, including land acquisition, construction, and maintenance, are substantially higher compared to the deployment of stratospheric drones. These aerial platforms, once airborne, can cover vast geographical areas, potentially rendering numerous terrestrial towers redundant. At an operating height of 20 km, one would expect a coverage radius ranging from 20 km up to 500 km, depending on the antenna system, application, and business model (e.g., terrestrial broadband services, surveillance, environmental monitoring, …).

The stratospheric drone-based coverage platform, and by platform, I mean the complete infrastructure that will replace the terrestrial cellular network, will consist of unmanned autonomous drones with a considerable wingspan (e.g., 747-like of ca. 69 meters). For example, European (German) Leichtwerk’s StratoStreamer has a wingspan of 65 meters and a wing area of 197 square meters with a payload of 120+ kg (note: in comparison a Boing 747 has ca. 500+ m2 wing area but its payload is obviously much much higher and in the range of 50 to 60 metric tons). Leichtwerk AG work closely together with AESA in order to achieve the European Union Aviation Safety Agency (EASA) type certificate that would allow the HAPS to integrate into civil airspace (see refs. [34] for what that means).

An advanced antenna system is positioned under the wings (or the belly) of the drone. I will assume that the coverage radius provided by a single drone is 50 km, but it can dynamically be made smaller or larger depending on the coverage scenario and use case. The drone-based advanced antenna system breaks up the coverage area (ca. six thousand five hundred plus square kilometers) into 400 patches (i.e., a number that can be increased substantially), averaging approx. 16 km2 per patch and a radius of ca. 2.5 km. Due to its near-ideal cellular link budget, the effective spectral efficiency is expected to be initially around 6 Mbps per MHz per cell. Additionally, the drone does not have the same spectrum limitations as a rural terrestrial site and would be able to support frequency bands in the downlink from ~900 MHz up to 3.9 GHz (and possibly higher, although likely with different antenna designs). Due to the HAP altitude, the Earth-to-HAP uplink signal will be limited to a lower frequency spectrum to ensure good signal quality is being received at the stratospheric antenna. It is prudent to assume a limit of 2.1 GHz to possibly 2.6 GHz. All under the assumption that the stratospheric drone operator has achieved regulatory approval for operating the terrestrial cellular spectrum from their coverage platform. It should be noted that today, cellular frequency spectrum approved for terrestrial use cannot be used at an altitude unless regulatory permission has been given (more on this later).

Let’s look at an example. We would need ca. 46 drones to cover the whole of Germany with the above-assumed specifications. Furthermore, if we take the average spectrum portfolio of the 3 main German operators, this will imply that the stratospheric drone could be functioning with up to 145 MHz in downlink and at least 55 MHz uplink (i.e., limiting UL to include 2.1 GHz). Using the HAP DL spectral efficiency and coverage area we get a throughput density of 70+ Mbps/km2 and an effective rural cell throughput of 870 Mbps. In terrestrial-based cellular coverage, the contribution to quality at higher frequencies is rapidly degrading as a function of the distance to the antenna. This is not the case for HAP-based coverage due to its near-ideal signal propagation.

In comparison, the three incumbent German operators have on average ca. 30±4k sites per operator with an average terrestrial coverage area of 12 km2 and a coverage radius of ca. 2.0 km (i.e., smaller in cities, ~1.3 km, larger in rural areas, ~2.7 km). Assume that the average cost of ownership related only to the passive part of the site is 20+ thousand euros and that 50% of the 30k sites (expect a higher number) would be redundant as the rural coverage would be replaced by stratospheric drones. Such a site reduction quantum conservatively would lead to a minimum gross monetary reduction of 300 million euros annually (not considering the cost of the alternative technology coverage solution).

In our example, the question is whether we can operate a stratospheric drone-based platform covering rural Germany for less than 300 million euros yearly. Let’s examine this question. Say the stratospheric drone price is 1 million euros per piece (similar to the current Starlink satellite price, excluding the launch cost, which would add another 1.1 million euros to the satellite cost). For redundancy and availability purposes, we assume we need 100 stratospheric drones to cover rural Germany, allowing me to decommission in the radius of 15 thousand rural terrestrial sites. The decommissioning cost and economical right timing of tower contract termination need to be considered. Due to the standard long-term contracts may be 5 (optimistic) to 10+ years (realistic) year before the rural network termination could be completed. Many Telecom businesses that have spun out their passive site infrastructure have done so in mutual captivity with the Tower management company and may have committed to very “sticky” contracts that have very little flexibility in terms of site termination at scale (e.g., 2% annually allowed over total portfolio).

We have a capital expense of 100 million for the stratospheric drones.  We also have to establish the support infrastructure (e.g., ground stations, airfield suitability rework, development, …), and consider operational expenses. The ballpark figure for this cost would be around 100 million euros for Capex for establishing the supporting infrastructure and another 30 million euros in annual operational expenses. In terms of steady-state Capex, it should be at most 20 million per year. In our example, the terrestrial rural network would have cost 3 billion euros, mainly Opex, over ten years compared to 700 million euros, a little less than half as Opex, for the stratospheric drone-based platform (not considering inflation).

The economical requirements of a stratospheric unmanned and autonomous drone-based coverage platform should be superior compared to the current cellular terrestrial coverage platform. As the stratospheric coverage platform scales and increasingly more stratospheric drones are deployed, the unit price is also likely to reduce accordingly.

Spectrum usage rights yet another critical piece.

It should be emphasized that the deployment of cellular frequency spectrum in stratospheric and LEO satellite contexts is governed by a combination of technical feasibility, regulatory frameworks, coordination to prevent interference, and operational needs. The ITU, along with national regulatory bodies, plays a central role in deciding the operational possibilities and balancing the needs and concerns of various stakeholders, including satellite operators, terrestrial network providers, and other spectrum users. Today, there are many restrictions and direct regulatory prohibitions in repurposing terrestrially assigned cellular frequencies for non-terrestrial purposes.

The role of the World Radiocommunications Conference (WRC) role is pivotal in managing the global radio-frequency spectrum and satellite orbits. Its decisions directly impact the development and deployment of various radiocommunication services worldwide, ensuring their efficient operation and preventing interference across borders. The WRC’s work is fundamental to the smooth functioning of global communication networks, from television and radio broadcasting to cellular networks and satellite-based services. The WRC is typically held every three to four years, with the latest one, WRC-23, held in Dubai at the end of 2023, reference [13] provides the provisional final acts of WRC-23 (December 2023). In landmark recommendation, WRC-23 relaxed the terrestrial-only conditions for the 698 to 960 MHz and 1,71 to 2.17 GHz, and 2.5 to 2.69 GHz frequency bands to also apply for high-altitude platform stations (HAPS) base stations (“Antennas-in -Sky”). It should be noted that there are slightly different frequency band ranges and conditions, depending on which of the three ITU-R regions (as well as exceptions for particular countries within a region) the system will be deployed in. Also the HAPS systems do not enjoy protection or priority over existing use of those frequency bands terrestrially. It is important to note that the WRC-23 recommendation only apply to coverage platforms (i.e., HAPS) in the range from 20 to 50 km altitude. These WRC-23 frequency-bands relaxation does not apply to satellite operation. With the recognized importance of non-terrestrial networks and the current standardization efforts (e.g., towards 6G), it is expected that the fairly restrictive regime on terrestrial cellular spectrum may be relaxed further to also allow mobile terrestrial spectrum to be used in “Antenna-in-the-Sky” coverage platforms. Nevertheless, HAPS and terrestrial use of cellular frequency spectrum will have to be coordinated to avoid interference and resulting capacity and quality degradation.

SoftBank announced recently (i.e., 28 December 2023 [11]), after deliberations at the WRC-23, that they had successfully gained approval within the Asia-Pacific region (i.e., ITU-R region 3) to use mobile spectrum bands, namely 700-900MHz, 1.7GHz, and 2.5GHz, for stratospheric drone-based mobile broadband cellular services (see also refs. [13]). As a result of this decision, operators in different countries and regions will be able to choose a spectrum with greater flexibility when they introduce HAPS-based mobile broadband communication services, thereby enabling seamless usage with existing smartphones and other devices.

Another example of re-using terrestrial licensed cellular spectrum above ground is SpaceX direct-to-cell capable 2nd generation Starlink satellites.

On January 2nd, 2024, SpaceX launched their new generation of Starlink satellites with direct-to-cell capabilities to close a connection to a regular mobile cellular phone (e.g., smartphone). The new direct-to-cell Starlink satellites use T-Mobile US terrestrial licensed cellular frequency band (i.e., 2×5 MHz Band 25, PCS G-block) and will work, according to T-Mobile US, with most of their existing mobile phones. The initial direct-to-cell commercial plans will only support low-bandwidth text messaging and no voice or more bandwidth-heavy applications (e.g., streaming). Expectations are that the direct-to-cell system would deliver up to 18.3 Mbps (3.66 Mbps/MHz/cell) downlink and up to 7.2 Mbps (1.44 Mbps/MHz/cell) uplink over a channel bandwidth of 5 MHz (maximum).

Given that terrestrial 4G LTE systems struggle with such performance, it will be super interesting to see what the actual performance of the direct-to-cell satellite constellation will be.

COMPARISON WITH LEO SATELLITE BROADBAND NETWORKS.

When juxtaposed with LEO satellite networks such as Starlink (SpaceX), OneWeb (Eutelsat Group), or Kuiper (Amazon), stratospheric drones offer several advantages. Firstly, the proximity to the Earth’s surface (i.e., 300 – 2,000 km) results in lower latency, a critical factor for real-time applications. While LEO satellites, like those used by Starlink, have reduced latency (ca. 3 ms round-trip-time) compared to traditional geostationary satellites (ca. 240 ms round-trip-time), stratospheric drones can provide even quicker response times (one-tenth of an ms in round-trip-time), making the stratospheric drone substantially more beneficial for applications such as emergency services, telemedicine, and high-speed internet services.

A stratospheric platform operating at 20 km altitude and targeting surveillance, all else being equal, would be 25 times better at distinguishing objects apart than an LEO satellite operating at 500 km altitude. The global aerial imaging market is expected to exceed 7 billion euros by 2030, with a CAGR of 14.2% from 2021. The flexibility of the stratospheric drone platform allows for combining cellular broadband services and a wide range of advanced aerial imaging services. Again, it is advantageous that the stratospheric drone regularly returns to Earth for fueling, maintenance, and technology upgrades and enhancements. This is not possible with an LEO satellite platform.

Moreover, the deployment and maintenance of stratospheric drones are, in theory, less complex and costly than launching and maintaining a constellation of satellites. While Starlink and similar projects require significant upfront investment for satellite manufacturing and rocket launches, stratospheric drones can be deployed at a fraction of the cost, making them a more economically viable option for many applications.

The Starlink LEO satellite constellation currently is the most comprehensive satellite (fixed) broadband coverage service. As of November 2023, Starlink had more than 5,000 satellites in low orbit (i.e., ca. 550 km altitude), and an additional 7,000+ are planned to be deployed, with a total target of 12+ thousand satellites. The current generation of Starlink satellites has three downlink phased-array antennas and one uplink phase-array antenna. This specification translates into 48 beams downlink (satellite to ground) and 16 beams uplink (ground to satellite). Each Starlink beam covers approx. 2,800 km2 with a coverage range of ca. 30 km, over which a 250 MHz downlink channel (in the Ku band) has been assigned. According to Portillo et al. [14], the spectral efficiency is estimated to be 2.7 Mbps per MHz, providing a total throughput of a maximum of 675 Mbps in the coverage area or a throughput density of ca. 0.24 Mbps per km2.

According to the latest Q2-2023 Ookla speed test it is found that “among the 27 European countries that were surveyed, Starlink had median download speeds greater than 100 Mbps in 14 countries, greater than 90 Mbps in 20 countries, and greater than 80 in 24 countries, with only three countries failing to reach 70 Mbps” (see reference [18]). Of course, the actual customer experience will depend on the number of concurrent users demanding resources from the LEO satellite as well as weather conditions, proximity of other users, etc. Starlink themselves seem to have set an upper limit of 220 Mbps download speed for their so-called priority service plan or otherwise 100 Mbps (see [19] below). Quite impressive performance if there are no other broadband alternatives available.

According to Elon Musk, SpaceX aims to reduce each Starlink satellite’s cost to less than one million euros. However, according to Elon Musk, the unit price will depend on the design, capabilities, and production volume. The launch cost using the SpaceX Falcon 9 launch vehicle starts at around 57 million euros, and thus, the 50 satellites would add a launch cost of ca. 1.1 million euros per satellite. SpaceX operates, as of September 2023, 150 ground stations (“Starlink Gateways”) globally that continue to connect the satellite network with the internet and ground operations. At Starlink’s operational altitude, the estimated satellite lifetime is between 5 and 7 years due to orbital decay, fuel and propulsion system exhaustion, and component durability. Thus, a LEO satellite business must plan for satellite replacement cycles. This situation differs greatly from the stratospheric drone-based operation, where the vehicles can be continuously maintained and upgraded. Thus, they are significantly more durable, with an expected useful lifetime exceeding ten years and possibly even 20 years of operational use.

Let’s consider our example of Germany and what it would take to provide LEO satellite coverage service targeting rural areas. It is important to understand that a LEO satellite travels at very high speeds (e.g., upwards of 30 thousand km per hour) and thus completes an orbit around Earth in between 90 to 120 minutes (depending on the satellite’s altitude). It is even more important to remember that Earth rotates on its axis (i.e., 24 hours for a full rotation), and the targeted coverage area will have moved compared to a given satellite orbit (this can easily be several hundreds to thousands of kilometers). Thus, to ensure continuous satellite broadband coverage of the same area on Earth, we need a certain number of satellites in a particular orbit and several orbits to ensure continuous coverage at a target area on Earth. We would need at least 210 satellites to provide continuous coverage of Germany. Most of the time, most satellites would not cover Germany, and the operational satellite utilization will be very low unless other areas outside Germany are also being serviced.

Economically, using the Starlink numbers above as a guide, we incur a capital expense of upwards of 450 million euros to realize a satellite constellation that could cover Germany. Let’s also assume that the LEO satellite broadband operator (e.g., Starlink) must build and launch 20 satellites annually to maintain its constellation and thus incur an additional Capex of ca. 40+ million euros annually. This amount does not account for the Capex required to build the ground network and the operations center. Let’s say all the rest requires an additional 10 million euros Capex to realize and for miscellaneous going forward. The technology-related operational expenses should be low, at most 30 million euros annually (this is a guesstimate!) and likely less. So, covering Germany with an LEO broadband satellite platform over ten years would cost ca. 1.3 billion euros. Although substantially more costly than our stratospheric drone platform, it is still less costly than running a rural terrestrial mobile broadband network.

Despite being favorable compared in economic to the terrestrial cellular network, it is highly unlikely to make any operational and economic sense for a single operator to finance such a network, and it would probably only make sense if shared between telecom operators in a country and even more so over multiple countries or states (e.g., European Union, United States, PRC, …).

Despite the implied silliness of a single mobile operator deploying a satellite constellation for a single Western European country (irrespective of it being fairly large), the above example serves two purposes; (1) To illustrates how economically in-efficient rural mobile networks are that a fairly expansive satellite constellation could be more favorable. Keep in mind that most countries have 3 or 4 of them, and (2) It also shows that the for operators to share the economics of a LEO satellite constellation over larger areal footprint may make such a strategy very attractive economically,

Due to the path loss at 550 km (LEO) being substantially higher than at 20 km (stratosphere), all else being equal, the signal quality of the stratospheric broadband drone would be significantly better than that of the LEO satellite. However, designing the LEO satellite with more powerful transmitters and sensitive receivers can compensate for the factor of almost 30 in altitude difference to a certain extent. Clearly, the latency performance of the LEO satellite constellation would be inferior to that of the stratospheric drone-based platform due to the significantly higher operating altitude.

It is, however, the capacity rather than shared cost could be the stumbling block for LEOs: For a rural cellular network or stratospheric drone platform, we see the MNOs effectively having “control” over the capex costs of the network, whether it be the RAN element for a terrestrial network, or the cost of whole drone network (even if it in the future, this might be able to become a shared cost).

However, for the LEO constellation, we think the economics of a single MNO building a LEO constellation even for their own market is almost entirely out of the question (ie multiple €bn capex outlay). Hence, in this situation, the MNOs will rely on a global LEO provider (ie Starlink, or AST Space Mobile) and will “lend” their spectrum to their in their respective geography in order to provide service. Like the HAPs, this will also require further regulatory approvals in order to free up terrestrial spectrum for satellites in rural areas.

We do not yet have the visibility of the payments the LEOs will require, so there is the potential that this could be a lower cost alternative again to rural networks, but as we show below, we think the real limitation for LEOs might not be the shared capacity rental cost, but that there simply won’t be enough capacity available to replicate what a terrestrial network can offer today.

However, the stratospheric drone-based platform provides a near-ideal cellular performance to the consumer, close to the theoretical peak performance of a terrestrial cellular network. It should be emphasized that the theoretical peak cellular performance is typically only experienced, if at all, by consumers if they are very near the terrestrial cellular antenna and in a near free-space propagation environment. This situation is a very rare occurrence for the vast majority of mobile consumers.

Figure 7 summarizes the above comparison between a rural terrestrial cellular network with the non-terrestrial cellular networks such as LEO satellites and Stratospheric drones.

Figure 7 Illustrating a comparison between terrestrial cellular coverage with stratospheric drone-based (“Antenna-in-the-sky”) cellular coverage and Low Earth Orbit (LEO) satellite coverage options.

While the majority of the 5,500+ Starlink constellation is 13 GHz (Ku-band), at the beginning of 2024, Space X launched a few 2nd generation Starlink satellites that support direct connections from the satellite to a normal cellular device (e.g., smartphone), using 5 MHz of T-Mobile USA’s PCS band (1900 MHz). The targeted consumer service, as expressed by T-Mobile USA, is providing texting capabilities over areas with no or poor existing cellular coverage across the USA. This is fairly similar to services at similar cellular coverage areas presently offered by, for example, AST SpaceMobile, OmniSpace, and Lynk Global LEO satellite services with reported maximum speed approaching 20 Mbps. The so-called Direct-2-Device, where the device is a normal smartphone without satellite connectivity functionality, is expected to develop rapidly over the next 10 years and continue to increase the supported user speeds (i.e., utilized terrestrial cellular spectrum) and system capacity in terms of smaller coverage areas and higher number of satellite beams.

Table 1 below provides an overview of the top 10 LEO satellite constellations targeting (fixed) internet services (e.g., Ku band), IoT and M2M services, and Direct-to-Device (or direct-to-cell) services. The data has been compiled from the NewSpace Index website, which should be with data as of 31st of December 2023. The Top-10 satellite constellation rank has been based on the number of launched satellites until the end of 2023. Two additional Direct-2-Cell (D2C or Direct-to-Device, D2D) LEO satellite constellations are planned for 2024-2025. One is SpaceX Starlink 2nd generation, which launched at the beginning of 2024, using T-Mobile USA’s PCS Band to connect (D2D) to normal terrestrial cellular handsets. The other D2D (D2C) service is Inmarsat’s Orchestra satellite constellation based on L-band (for mobile terrestrial services) and Ka for fixed broadband services. One new constellation (Mangata Networks) targeting 5G services. With two 5G constellations already launched, i.e., Galaxy Space (Yinhe) launched 8 LEO satellites, 1,000 planned using Q- and V-bands (i.e., not a D2D cellular 5G service), and OmniSpace launched two satellites and have planned 200 in total. Moreover, currently, there is one planned constellation targeting 6G by the South Korean Hanwha Group (a bit premature, but interesting nevertheless) with 2,000 6G LEO Satellites planned. Most currently launched and planned satellite constellations offering (or plan to provide) Direct-2-Cell services, including IoT and M2M, are designed for low-frequency bandwidth services that are unlikely to compete with terrestrial cellular networks’ quality of service where reasonable good coverage (or better) exists.

In Table 1 below, we then show 5 different services with the key input variables as cell radius, spectral efficiency and downlink spectrum. From this we can derive what the “average” capacity could be per square kilometer of rural coverage.

We focus on this metric as the best measure of capacity available once multiple users are on the service the spectrum available is shared. This is different from “peak” speeds which are only relevant in the case of very few users per cell.

  • We start with terrestrial cellular today for bands up to 2.1GHz and show that assuming a 2.5km cell radius, the average capacity is equivalent to 11Mbps per sq.km.
  • For a LEO service using Ku-band, i.e., with 250MHz to an FWA dish, the capacity could be ca. 2Mbps per sq.km.
  • For a LEO-based D2D device, what is unknown is what the ultimate spectrum allowance could be for satellite services with cellular spectrum bands, and spectral efficiency. Giving the benefit of the doubt on both, but assuming the beam radius is always going to be larger, we can get to an “optimistic” future target of 2Mbps per sq. km, i.e., 1/5th of a rural terrestrial network.
  • Finally, we show for a stratospheric drone, that given similar cell radius to a rural cell today, but with higher downlink available and greater spectral efficiency, we can reach ca. 55Mbps per sq. km, i.e. 5x what a current rural network can offer.

INTEGRATING WITH 5G AND BEYOND.

The advent of 5G, and eventually 6G, technology brings another dimension to the utility of stratospheric drones delivering mobile broadband services. The high-altitude platform’s ability to seamlessly integrate with existing 5G networks makes them an attractive option for expanding coverage and enhancing network capacity at superior economics, particularly in rural areas where the economics for terrestrial-based cellular coverage tend to be poor. Unlike terrestrial networks that require extensive groundwork for 5G rollout, the non-terrestrial network operator (NTNO) can rapidly deploy stratospheric drones to provide immediate 5G coverage over large areas. The high-altitude platform is also incredibly flexible compared to both LEO satellite constellations and conventional rural cellular network flexibility. The platform can easily be upgraded during its ground maintenance window and can be enhanced as the technology evolves. For example, upgrading to and operationalizing 6G would be far more economical with a stratospheric platform than having to visit thousands or more rural sites to modernize or upgrade the installed active infrastructure.

SUMMARY.

Stratospheric drones represent a significant advancement in the realm of wireless communication. Their strategic positioning in the stratosphere offers superior coverage and connectivity compared to terrestrial networks and low-earth satellite solutions. At the same time, their economic efficiency makes them an attractive alternative to ground-based infrastructures and LEO satellite systems. As technology continues to evolve, these high-altitude platforms (HAPs) are poised to play a crucial role in shaping the future of global broadband connectivity and ultra-high availability connectivity solutions, complementing the burgeoning 5G networks and paving the way for next-generation three-dimensional communication solutions. Moving away from today’s flat-earth terrestrial-locked communication platforms.

The strategic as well as the disruptive potential of the unmanned autonomous stratospheric terrestrial coverage platform is enormous, as shown in this article. It has the potential to make most of the rural (at least) cellular infrastructure redundant, resulting in substantial operational and economic benefits to existing mobile operators. At the same time, the HAPs could, in rural areas, provide much better service overall in terms of availability, improved coverage, and near-ideal speeds compared to what is the case in today’s cellular networks. It might also, at scale, become a serious competitive and economical threat to LEO satellite constellations, such as, for example, Starlink and Kuipers, that would struggle to compete on service quality and capacity compared to a stratospheric coverage platform.

Although the strategic, economic, as well as disruptive potential of the unmanned autonomous stratospheric terrestrial coverage platform is enormous, as shown in this article, the flight platform and advanced antenna technology are still in a relatively early development phase. Substantial regulatory work remains in terms of permitting the terrestrial cellular spectrum to be re-used above terra firma at the “Antenna-in-the-Sky. The latest developments out of WRC-23 for Asia Pacific appear very promising, showing that we are moving in the right direction of re-using terrestrial cellular spectrum in high-altitude coverage platforms. Last but not least, operating an unmanned (autonomous) stratospheric platform involves obtaining certifications as well as permissions and complying with various flight regulations at both national and international levels.

Terrestrial Mobile Broadband Network – takeaway:

  • It is the de facto practice for mobile cellular networks to cover nearly 100% geographically. The mobile consumer expects a high-quality, high-availability service everywhere.
  • A terrestrial mobile network has a relatively low area coverage per unit antenna with relatively high capacity and quality.
  • Mobile operators incur high and sustainable infrastructure costs, especially in rural areas with low or no return on that cost.
  • Physical obstructions and terrain limit performance (i.e., non-free space characteristics).
  • Well-established technology with high reliability.
  • The potential for high bandwidth and low latency in urban areas with high demand may become a limiting factor for LEO satellite constellations and stratospheric drone-based platforms. Thus, it is less likely to provide operational and economic benefits covering high-demand, dense urban, and urban areas.

LEO Satellite Network – takeaway:

  • The technology is operational and improving. There is currently some competition (e.g., Starlink, Kuiper, OneWeb, etc.) in this space, primarily targeting fixed broadband and satellite backhaul services. Increasingly, new LEO satellite-based business models are launched providing lower-bandwidth cellular-spectrum based direct-to-device (D2D) text, 4G and 5G services to regular consumer and IoT devices (i.e., Starlink, Lynk Global, AST SpaceMobile, OmniSpace, …).
  • Broader coverage, suitable for global reach. It may only make sense when the business model is viewed from a worldwide reach perspective (e.g., Starlink, OneWeb,…), resulting in much-increased satellite network utilization.
  • An LEO satellite broadband network can cover a vast area per satellite due to its high altitude. However, such systems are in nature capacity-limited, although beam-forming antenna technologies (e.g., phased array antennas) allow better capacity utilization.
  • The LEO satellite solutions are best suited for low-population areas with limited demand, such as rural and largely unpopulated areas (e.g., sea areas, deserts, coastlines, Greenland, polar areas, etc.).
  • Much higher latency compared to terrestrial and drone-based networks. 
  • Less flexible once in orbit. Upgrades and modernization only via replacement.
  • The LEO satellite has a limited useful operational lifetime due to its lower orbital altitude (e.g., 5 to 7 years).
  • Lower infrastructure cost for rural coverage compared to terrestrial networks, but substantially higher than drones when targeting regional areas (e.g., Germany or individual countries in general).
  • Complementary to the existing mobile business model of communications service providers (CSPs) with a substantial business risk to CSPs in low-population areas where little to no capacity limitations may occur.
  • Requires regulatory permission (authorization) to operate terrestrial frequencies on the satellite platform over any given country. This process is overseen by national regulatory bodies in coordination with the International Telecommunication Union (ITU) as well as national regulators (e.g., FCC in the USA). Satellite operators must apply for frequency bands for uplink and downlink communications and coordinate with the ITU to avoid interference with other satellites and terrestrial systems. In recent years, however, there has been a trend towards more flexible spectrum regulations, allowing for innovative uses of the spectrum like integrating terrestrial and satellite services. This flexibility is crucial in accommodating new technologies and service models.
  • Operating a LEO satellite constellation requires a comprehensive set of permissions and certifications that encompass international and national space regulations, frequency allocation, launch authorization, adherence to space debris mitigation guidelines, and various liability and insurance requirements.
  • Both LEO and MEO satellites is likely going to be complementary or supplementary to stratospheric drone-based broadband cellular networks offering high-performing transport solutions and possible even acts as standalone or integrated (with terrestrial networks) 5G core networks or “clouds-in-the-sky”.

Stratospheric Drone-Based Network – takeaway:

  • It is an emerging technology with ongoing research, trials, and proof of concept.
  • A stratospheric drone-based broadband network will have lower deployment costs than terrestrial and LEO satellite broadband networks.
  • In rural areas, the stratospheric drone-based broadband network offers better economics and near-ideal quality than terrestrial mobile networks. In terms of cell size and capacity, it can easily match that of a rural mobile network.
  • The solution offers flexibility and versatility and can be geographically repositioned as needed. The versatility provides a much broader business model than “just” an alternative rural coverage solution (e.g., aerial imaging, surveillance, defense scenarios, disaster area support, etc.).
  • Reduced latency compared to LEO satellites.
  • Also ideal for targeted or temporary coverage needs.
  • Complementary to the existing mobile business model of communications service providers (CSPs) with additional B2B and public services business potential from its application versatility.
  • Potential substantial negative impact on the telecom tower business as the stratospheric drone-based broadband network would make (at least) rural terrestrial towers redundant.
  • May disrupt a substantial part of the LEO satellite business model due to better service quality and capacity leaving the LEO satellite constellations revenue pool to remote areas and specialized use cases.
  • Requires regulatory permission to operate terrestrial frequencies (i.e., frequency authorization) on the stratospheric drone platform (similar to LEO satellites). Big steps have are already been made at the latest WRC-23, where the frequency bands 698 to 960 MHz, 1710 to 2170 MHz, and 2500 to 2690 MHz has been relaxed to allow for use in HAPS operating at 20 to 50 km altitude (i.e., the stratosphere).
  • Operating a stratospheric platform in European airspace involves obtaining certifications as well as permissions and (of course) complying with various regulations at both national and international levels. This includes the European Union Aviation Safety Agency (EASA) type certification and the national civil aviation authorities in Europe.

FURTHER READING.

  1. New Street Research “Stratospheric drones: A game changer for rural networks?” (January 2024).
  2. https://hapsalliance.org/
  3. https://www.stratosphericplatforms.com/, see also “Beaming 5G from the stratosphere” (June, 2023) and “Cambridge Consultants building the world’s largest  commercial airborne antenna” (2021).
  4. Iain Morris, “Deutsche Telekom bets on giant flying antenna”, Light Reading (October 2020).
  5. “Deutsche Telekom and Stratospheric Platforms Limited (SPL) show Cellular communications service from the Stratosphere” (November 2020).
  6. “High Altitude Platform Systems: Towers in the Skies” (June 2021).
  7. “Stratospheric Platforms successfully trials 5G network coverage from HAPS vehicle” (March 2022).
  8. Leichtwerk AG, “High Altitude Platform Stations (HAPS) – A Future Key Element of Broadband Infrastructure” (2023). I recommend to closely follow Leichtwerk AG which is a world champion in making advanced gliding planes. The hydrogen powered StratoStreamer HAP is near-production ready, and they are currently working on a solar-powered platform. Germany is renowned for producing some of the best gliding planes in the world (after WWII Germany was banned from developing and producing aircrafts, military as well as civil. These restrictions was only relaxed in the 60s). Germany has a long and distinguished history in glider development, dating back to the early 20th century. German manufacturers like Schleicher, Schempp-Hirth, and DG Flugzeugbau are among the world’s leading producers of high-quality gliders. These companies are known for their innovative designs, advanced materials, and precision engineering, contributing to Germany’s reputation in this field.
  9. Jerzy Lewandowski, “Airbus Aims to Revolutionize Global Internet Access with Stratospheric Drones” (December 2023).
  10. Utilities One, “An Elevated Approach High Altitude Platforms in Communication Strategies”, (October 2023).
  11. Rajesh Uppal, “Stratospheric drones to provide 5g wireless communications global internet border security and military surveillance”  (May 2023).
  12. Softbank, “SoftBank Corp.-led Proposal to Expand Spectrum Use for HAPS Base Stations Agreed at World Radiocommunication Conference 2023 (WRC-23)”, press release (December 2023).
  13. ITU Publication, World Radiocommunications Conference 2023 (WRC-23), Provisional Final Acts, (December 2023). Note 1: The International Telecommunication Union (ITU) divides the world into three regions for the management of radio frequency spectrum and satellite orbits: Region 1: includes Europe, Africa, the Middle East west of the Persian Gulf including Iraq, the former Soviet Union, and Mongolia, Region 2: covers the Americas, Greenland, and some of the eastern Pacific Islands, and Region 3: encompasses Asia (excl. the former Soviet Union), Australia, the southwest Pacific, and the Indian Ocean’s islands.
  14. Geoff Huston, “Starlink Protocol Performance” (November 2023). Note 2: The recommendations, such as those designated with “ADD” (additional), are typically firm in the sense that they have been agreed upon by the conference participants. However, they are subject to ratification processes in individual countries. The national regulatory authorities in each member state need to implement these recommendations in accordance with their own legal and regulatory frameworks.
  15. Curtis Arnold, “An overview of how Starlink’s Phased Array Antenna “Dishy McFlatface” works.”, LinkedIn (August 2023).
  16. Quora, “How much does a satellite cost for SpaceX’s Starlink project and what would be the cheapest way to launch it into space?” (June 2023).
  17. The Clarus Network Group, “Starlink v OneWeb – A Comprehensive Comparison” (October 2023).
  18. Brian Wang, “SpaceX Launches Starlink Direct to Phone Satellites”, (January 2024).
  19. Sergei Pekhterev, “The Bandwidth Of The StarLink Constellation…and the assessment of its potential subscriber base in the USA.”, SatMagazine, (November 2021).
  20. I. del Portillo et al., “A technical comparison of three low earth orbit satellite constellation systems to provide global broadband,” Acta Astronautica, (2019).
  21. Nils Pachler et al., “An Updated Comparison of Four Low Earth Orbit Satellite Constellation Systems to Provide Global Broadband” (2021).
  22. Shkelzen Cakaj, “The Parameters Comparison of the “Starlink” LEO Satellites Constellation for Different Orbital Shells” (May 2021).
  23. Mike Puchol, “Modeling Starlink capacity” (October 2022).
  24. Mike Dano, “T-Mobile and SpaceX want to connect regular phones to satellites”, Light Reading (August 2022).
  25. Starlink, “SpaceX sends first text message via its newly launched direct to cell satellites” (January 2024).
  26. GSMA.com, “New Speedtest Data Shows Starlink Performance is Mixed — But That’s a Good Thing” (2023).
  27. Starlink, “Starlink specifications” (Starlink.com page).
  28. AST SpaceMobile website: https://ast-science.com/ Constellation Areas: Internet, Direct-to-Cell, Space-Based Cellular Broadband, Satellite-to-Cellphone. 243 LEO satellites planned. 2 launched.
  29. Lynk Global website: https://lynk.world/ (see also FCC Order and Authorization). It should be noted that Lynk can operate within 617 to 960 MHz (Space-to-Earth) and 663 to 915 MHz (Earth-to-Space). However, only outside the USA. Constellation Area: IoT / M2M, Satellite-to-Cellphone, Internet, Direct-to-Cell. 8 LEO satellites out of 10 planned.
  30. Omnispace website: https://omnispace.com/ Constellation Area: IoT / M2M, 5G. World’s first global 5G non terrestrial network. Initial support 3GPP-defined Narrow-Band IoT radio interface. Planned 200 LEO and <15 MEO satellites. So far only 2 satellites launched.
  31. NewSpace Index: https://www.newspace.im/ I find this resource having excellent and up-to date information of commercial satellite constellations.
  32. Wikipedia, “Satellite constellation”.
  33. LEOLABS Space visualization – SpaceX Starlink mapping. (deselect “Debris”, “Beams”, and “Instruments”, and select “Follow Earth”). An alternative visualization service for Starlink & OneWeb satellites is the website Satellitemap.space (you might go to settings and turn on signal Intensity which will give you the satellite coverage hexagons).
  34. European Union Aviation Safety Agency (EASA). Note that an EASA-type Type Certificate is a critical document in the world of aviation. This certificate is a seal of approval, indicating that a particular type of aircraft, engine, or aviation component meets all the established safety and environmental standards per EASA’s stringent regulations. When an aircraft, engine, or component is awarded an EASA Type Certificate, it signifies a thorough and rigorous evaluation process that it has undergone. This process assesses everything from design and manufacturing to performance and safety aspects. The issuance of the certificate confirms that the product is safe for use in civil aviation and complies with the necessary airworthiness requirements. These requirements are essential to ensure aircraft operating in civil airspace safety and reliability. Beyond the borders of the European Union, an EASA Type Certificate is also highly regarded globally. Many countries recognize or accept these certificates, which facilitate international trade in aviation products and contribute to the global standardization of aviation safety.

ACKNOWLEDGEMENT.

I greatly acknowledge my wife, Eva Varadi, for her support, patience, and understanding during the creative process of writing this article.

I also owe a lot of gratitude to James Ratzer, Partner at New Street Research, for editorial suggestions, great discussions and challenges making the paper far better than it otherwise would have been. I would also like to thank Russel Waller, Pan European Telecoms and ESG Equity Analyst at New Street Research, for being supportive and insistent to get something written for NSR.

I also greatly appreciate my past collaboration and the many discussions on the topic of Stratospheric Drones in particular and advanced antenna designs and properties in general that I have had with Dr. Jaroslav Holis, Senior R&D Manager (Group Technology, Deutsche Telekom AG) over the last couple of years. When it comes to my early involvement in Stratospheric Drones activities with Group Technology Deutsche Telekom AG, I have to recognize my friend, mentor, and former boss, Dr. Bruno Jacobfeuerborn, former CTO Deutsche Telekom AG and Telekom Deutschland, for his passion and strong support for this activity since 2015. My friend and former colleague Rachid El Hattachi deserves the credit for “discovering” and believing in the opportunities that a cellular broadband-based stratospheric drone brings to the telecom industry.

Many thanks to CEO Dr. Reiner Kickert of Leichtwerk AG for providing some high resolution pictures of his beautiful StratoStreamer.

Thanks to my friend Amit Keren for suggesting a great quote that starts this article.

Any errors or unclarities are solely due to myself and not the collaborators and colleagues that have done their best to support this piece.

Telco energy consumption – a path to a greener future?

Posted on by
Reply

To my friend Rudolf van der Berg this story is not about how volumetric demand (bytes or bits) results in increased energy consumption (W·h). That notion is silly, as we both “violently” agree on ;-). I recommend that readers also check out Rudolf’s wonderful presentation, “Energy Consumption of the Internet (May 2023),” which he delivered at the RIPE86 student event this year in 2023.

Recently, I had the privilege to watch a presentation by a seasoned executive talk about what his telco company is doing for the environment regarding sustainability and CO2 reduction in general. I think the company is doing something innovative beyond compensating shortfalls with buying certificates and (mis)use of green energy resources.

They replace (reasonably) aggressively their copper infrastructure (country stat for 2022: ~90% of HH/~16% subscriptions) with green sustainable fiber (country stat for 2022: ~78%/~60%). This is an obvious strategy that results in a quantum leap in customer experience potential and helps reduce overall energy consumption resulting from operating the ancient copper network.

Missing a bit imo, was the consideration of and the opportunity to phase out the HFC network (country stat for 2022: ~70%/~60%) and reduce the current HFC+Fibre overbuild of 1.45 and, of course, reduce the energy consumption and operational costs (and complexity) of operating two fixed broadband technologies (3 if we include the copper). However, maybe understandably enough, substantial investments have been made in upgrading to Docsis 3.1. An investment that possibly still is somewhat removed from having been written off.

The “wtf-moment” (in an otherwise very pleasantly and agreeable session) came when the speaker alluded that as part of their sustainability and CO2 reduction strategy, the telco was busy migrating from 4G LTE to 5G with the reasoning that 5G is 90% more energy efficient compared to 4G.

Firstly, it is correct that 5G is (in apples-for-apples comparisons!) ca. 90% more efficient in delivering a single bit compared to 4G. The metric we use is Joules-per-bit or Watts-seconds-per-bit. It is also not uncommon at all to experience Telco executives hinting at the relative greenness of 5G (it is, in my opinion, decidedly not a green broadband communications technology … ).

Secondly, so what! Should we really care about relative energy consumption? After all, we pay for absolute energy consumption, not for whatever relativized measure of consumed energy.

I think I know the answer from the CFO and the in-the-know investors.

If the absolute energy consumption of 5G is higher than that of 4G, I will (most likely) have higher operational costs attributed to that increased power consumption with 5G. If I am not in an apples-for-apples situation, which rarely is the case, and I am anyway really not in, the 5G technology requires substantially more power to provide for new requirements and specifications. I will be worse off regarding the associated cost in absolute terms of money. Unless I also have a higher revenue associated with 5G, I am economically worse off than I was with the older technology.

Having higher information-related energy efficiency in cellular communications systems is a feature of the essential requirement of increasingly better spectral efficiency all else being equal. It does not guarantee that, in absolute monetary terms, a Telco will be better off … far from it!

THE ENERGY OF DELIVERING A BIT.

Energy, which I choose to represent in Joules, is equal to the Power (in Watt or W) that I need to consume per time-unit for a given output unit (e.g., a bit) times the unit of time (e.g., a second) it took to provide the unit.

Take a 4G LTE base station that consumes ca. 5.0kW to deliver a maximum throughput of 160 Mbps per sector (@ 80 MHz per sector). The information energy efficiency of the specific 4G LTE base station (e.g., W·s per bit) would be ca. 10 µJ/bit. The 4G LTE base station requires 10 micro (one millionth) Joules to deliver 1 bit (in 1 second).

In the 5G world, we would have a 5G SA base station, using the same frequency bands as 4G and with an additional 10 MHz @ 700MHz and 100 MHz @ 3.5 GHz included. The 3.5 GHz band is supported by an advanced antenna system (AAS) rather than a classical passive antenna system used for the other frequency bands. This configuration consumes 10 kW with ~40% attributed to the 3.5 GHz AAS, supporting ~1 Gbps per sector (@ 190 MHz per sector). This example’s 5G information energy efficiency would be ca. 0.3 µJ/bit.

In this non-apples-for-apples comparison, 5G is about 30 times more efficient in delivering a bit than 4G LTE (in the example above). Regarding what an operator actually pays for, 5G is twice as costly in energy consumption compared to 4G.

It should be noted that the power consumption is not driven by the volumetric demand but by the time that demand exists and the load per unit of time. Also, base stations will have a power consumption even when idle with the degree depending on the intelligence of the energy management system applied.

So, more formalistic, we have

E per bit = P (in W) · time (in sec) per bit, or in the basic units

J / bit = W·s / bit = W / (bit/s) = W / bps = W / [ MHz · Mbps/MHz/unit · unit-quantity ]

E per bit = P (in W) / [ Bandwidth (in MHz) · Spectral Efficiency (in Mbps/MHz/unit) · unit-quantity ]

It is important to remember that this is about the system spec information efficiency and that there is no direct relationship between the Power that you need and the outputted information your system will ultimately support bit-wise.

\frac{E_{4G}}{bit} \; = \; \frac {\; P_{4G} \;} {\; B_{4G} \; \cdot \; \eta_{4G,eff} \; \cdot N \;\;\;} and \;\;\; \frac{E_{5G}}{bit} \; = \; \frac {\; P_{5G} \;} {\; B_{5G} \; \cdot \; \eta_{5G,eff} \; \cdot N \;}

Thus, the relative efficiency between 4G and 5G is

\frac{E_{4G}/bit}{E_{5G}/bit} \; = \; \frac{\; P_{4G} \;}{\; P_{5G}} \; \cdot \; \frac{\; B_{5G} \;}{\; B_{4G}} \; \cdot \; \frac{\; \eta_{5G,eff} \;}{\; \eta_{4G,eff}}

Currently (i.e., 2023), the various components of the above are approximately within the following ranges.

\frac{P_{4G}}{P_{5G}} \; \lesssim \; 1

\frac{B_{5G}}{B_{4G}} \; > \;2

\frac{\; \eta_{5G,eff} \;}{\; \eta_{4G,eff}} \; \approx \; 10

The power consumption of a 5G RAT is higher than that of a 4G RAT. As we add higher frequency spectrum (e.g., C-band, 6GHz, 23GHz,…) to the 5G RAT, increasingly more spectral bandwidth (B) will be available compared to what was deployed for 4G. This will increase the bit-wise energy efficiency of 5G compared to 4G, although the power consumption is also expected to increase as higher frequencies are supported.

If the bandwidth and system power consumption is the same for both radio access technologies (RATs), then we have the relative information energy efficiency is

\frac{E_{4G}/bit}{E_{5G}/bit} \; \approx \; \frac{\; \eta_{5G,eff} \;}{\; \eta_{4G,eff}} \; \gtrsim \; 10

Depending on the relative difference in spectral efficiency. 5G is specified and designed to have at least ten times (10x) the spectral efficiency of 4G. If you do the math (assuming apples-to-apples applies), it is no surprise that 5G is specified to be 90% more efficient in delivering a bit (in a given unit of time) compared to 4G LTE.

And just to emphasize the obvious,

E_{RAT} \; = \; P_{RAT} \; \cdot \; t \; \approx \; E_{idle} \; + \; P_{BB, RAT} \; \cdot \; t \; +\sum_{freq}P_{freq,\; antenna\; type}\; \cdot \; t_{freq} \;

RAT refers to the radio access technology, BB is the baseband, freq the cellular frequencies, and idle to the situation where the system is not being utilized.

Volume in Bytes (or bits) does not directly relate to energy consumption. As frequency bands are added to a sector (of a base station), the overall power consumption will increase. Moreover, the more computing is required in the antenna, such as for advanced antenna systems, including massive MiMo antennas, the more power will be consumed in the base station. The more the frequency bands are being utilized in terms of time, the higher will the power consumption be.

Indirectly, as the cellular system is being used, customers consume bits and bytes (=8·bit) that will depend on the effective spectral efficiency (in bps/Hz), the amount of effective bandwidth (in Hz) experienced by the customers, e.g., many customers will be in a coverage situation where they may not benefit for example from higher frequency bands), and the effective time they make use of the cellular network resources. The observant reader will see that I like the term “effective.” The reason is that customers rarely enjoy the maximum possible spectral efficiency. Likely, not all the frequency spectrum covering customers is necessarily being applied to individual customers, depending on their coverage situation.

In the report “A Comparison of the Energy Consumption of Broadband Data Transfer Technologies (November 2021),” the authors show the energy and volumetric consumption of mobile networks in Finland over the period from 2010 to 2020. To be clear, I do not support the author’s assertion of causation between volumetric demand and energy consumption. As I have shown above, volumetric usage does not directly cause a given power consumption level. Over the 10-year period shown in the report, they observe a 70% increase in absolute power consumption (from 404 to 686 GWh, CAGR ~5.5%) and a factor of ~70 in traffic volume (~60 TB to ~4,000 TB, CAGR ~52%). Caution should be made in resisting the temptation to attribute the increase in energy over the period to be directly related to the data volume increase, however weak it is (i.e., note that the authors did not resist that temptation). Rudolf van der Berg has raised several issues with the approach of the above paper (as well as with many other related works) and indicated that the data and approach of the authors may not be reliable. Unfortunately, in this respect, it appears that systematic, reliable, and consistent data in the Telco industry is hard to come by (even if that data should be available to the individual telcos).

Technology change from 2G/3G to 4G, site densification, and more frequency bands can more than easily explain the increase in energy consumption (and all are far better explanations than data volume). It should be noted that there will also be reasons that decrease power consumption over time, such as more efficient electronics (e.g., via modernization), intelligent power management applications, and, last but not least, switching off of older radio access technologies.

The factors that drive a cell site’s absolute energy consumption is

  • Radio access technology with new technologies generally consumes more energy than older ones (even if the newer technologies have become increasingly more spectrally efficient).
  • The antenna type and configuration, including computing requirements for advanced signal processing and beamforming algorithms (that will improve the spectral efficiency at the expense of increased absolute energy consumption).
  • Equipment efficiency. In general, new generations of electronics and systems designs tend to be more energy-efficient for the same level of performance.
  • Intelligent energy management systems that allow for effective power management strategies will reduce energy consumption compared to what it would have been without such systems.
  • The network optimization goal policy. Is the cellular network planned and optimized for meeting the demands and needs of the customers (i.e., the economic design framework) or for providing the peak performance to as many customers as possible (i.e., the Umlaut/Ookla performance-driven framework)? The Umlaut/Ookla-optimized network, maxing out on base station configuration, will observe substantially higher energy consumption and associated costs.
The absolute cellular energy consumption has continued to rise as new radio access technologies (RAT) have been introduced irrespective of the leapfrog in those RATS spectral (bits per Hz) and information-related (Joules per bit) efficiencies.

WHY 5G IS NOT A GREEN TECHNOLOGY?

Let’s first re-acquaint ourselves with the 2015 vision of the 5G NGMN whitepaper;

“5G should support a 1,000 times traffic increase in the next ten years timeframe, with energy consumption by the whole network of only half that typically consumed by today’s networks. This leads to the requirement of an energy efficiency increase of x2000 in the next ten years timeframe.” (Section 4.2.2 Energy Efficiency, 5G White Paper by NGMN Alliance, February 2015).

The bold emphasis is my own and not in the paper itself. There is no doubt that the authors of the 5G vision paper had the ambition of making 5G a sustainable and greener cellular alternative than historically had been the case.

So, from the above statement, we have two performance figures that illustrate the ambition of 5G relative to 4G. Firstly, we have a requirement that the 5G energy efficiency should be 2000x higher than 4G (as it was back in the beginning of 2015).

\frac{E_{4G}/bit}{E_{5G}/bit} \; = \; \frac{\; P_{4G} \;}{\; P_{5G}} \; \cdot \; \frac{\; B_{5G} \;}{\; B_{4G}} \; \cdot \; \frac{\; \eta_{5G,eff} \;}{\; \eta_{4G,eff}} \; \geq \; 2,000

or

\frac{\; P_{4G} \;}{\; P_{5G}} \; \cdot \; \frac{\; B_{5G} \;}{\; B_{4G}} \; \geq \; 200

if

\frac{\; \eta_{5G,eff} \;}{\; \eta_{4G,eff}} \; \approx \; 10

Getting more spectrum bandwidth is relatively trivial as you go up in frequency and into, for example, the millimeter wave range (and beyond). However, getting 20+ GHz (e.g., 200+x100 MHz @ 4G) of additional practically usable spectrum bandwidth would be rather (=understatement) ambitious.

And that the absolute energy consumption of the whole 5G network should be half of what it was with 4G

\frac{E_{5G}}{E_{4G}} \; = \; \frac{\; P_{5G} \; \cdot \; t\;}{\; P_{4G} \; \cdot \; t}\; \approx \; \frac{\; P_{5G} \;}{\; P_{4G} \; } \; \leq \; \frac{1}{2}

If you think about this for a moment. Halfing the absolute energy consumption is an enormous challenge, even if it would have been with the same RAT. It requires innovation leapfrogs across the RAT electronic architecture, design, and material science underlying all of it. In other words, fundamental changes are required in the RF frontend (e.g., Power amplifiers, transceivers), baseband processing, DSP, DAC, ADC, cooling, control and management systems, algorithms, compute, etc…

But reality eats vision for breakfast … There really is no sign that the super-ambitious goal set by the NGMN Alliance in early 2015 is even remotely achievable even if we would give it another ten years (i.e., 2035). We are more than two orders of magnitude away from the visionary target of NGMN, and we are almost at the 10-year anniversary of the vision paper. We more or less get the benefit of the relative difference in spectral efficiency (x10), but no innovation beyond that has contributed very much to quantum leap cellular energy efficiency bit-wise.

I know many operators who will say that from a sustainability perspective, at least before the energy prices went through the roof, it really does not matter that 5G, in absolute terms, leads to substantial increases in energy consumption. They use green energy to supply the energy demand from 5G and pay off $CO_2$ deficits with certificates.

First of all, unless the increased cost can be recovered with the customers (e.g., price plan increase), it is a doubtful economic venue to pursue (and has a bit of a Titanic feel to it … going down together while the orchestra is playing).

Second, we should ask ourselves whether it is really okay for any industry to greedily consume sustainable and still relatively scarce green resources without being incentivized (or encouraged) to pursue alternatives and optimize across mobile and fixed broadband technologies. Particularly when fixed broadband technologies, such as fiber, are available, that would lead to a very sizable and substantial reduction in energy consumption … as customers increasingly adapt to fiber broadband.

Fiber is the greenest and most sustainable access technology we can deploy compared to cellular broadband technologies.

SO WHAT?

5G is a reality. Telcos are and will continue to invest substantially into 5G as they migrate their customers from 4G LTE to what ultimately will be 5G Standalone. The increase in customer experience and new capabilities or enablers are significant. By now, most Telcos will (i.e., 2023) have a very good idea of the operational expense associated with 5G (if not … you better do the math). Some will have been exploring investing in their own green power plants (e.g., solar, wind, hydrogen, etc.) to mitigate part of the energy surge arising from transitioning to 5G.

I suspect that as Telcos start reflecting on Open RAN as they pivot towards 6G (-> 2030+), above and beyond what 6G, as a RAT, may bring of additional operational expense pain, there will be new energy consumption and sustainability surprises to the cellular part of Telcos P&L. In general, breaking up an electronic system into individual (non-integrated) parts, as opposed to being integrated into a single unit, is likely to result in an increased power consumption. Some of the operational in-efficiencies that occur in breaking up a tightly integrated design can be mitigated by power management strategies. Though in order to get such power management strategies to work at the optimum may force a higher degree of supplier uniformity than the original intent of breaking up the tightly integrated system.

However, only Telcos that consider both their mobile and fixed broadband assets together, rather than two silos apart, will gain in value for customers and shareholders. Fixed-mobile (network) conversion should be taken seriously and may lead to very different considerations and strategies than 10+ years ago.

With increasing coverage of fiber and with Telcos stimulating aggressive uptake, it will allow those to redesign the mobile networks for what they were initially supposed to do … provide convenience and service where there is no fixed network present, such as when being mobile and in areas where the economics of a fixed broadband network makes it least likely to be available (e.g., rural areas) although LEO satellites (i.e., here today), maybe stratospheric drones (i.e., 2030+), may offer solid economic alternatives for those places. Interestingly, further simplifying the cellular networks supporting those areas today.

TAKE AWAY.

Volume in Bytes (or bits) does not directly relate to the energy consumption of the underlying communications networks that enable the usage.

The duration, the time scale, of the customer’s usage (i.e., the use of the network resources) does cause power consumption.

The bit-wise energy efficiency of 5G is superior to that of 4G LTE. It is designed that way via its spectral efficiency. Despite this, a 5G site configuration is likely to consume more energy than a 4G LTE site in the field and, thus, not a like-for-like in terms of number of bands and type of antennas deployed.

The absolute power consumption of a 5G configuration is a function of the number of bands deployed, the type of antennas deployed, intelligent energy management features, and the effective time 5G resources that customers have demanded.

Due to its optical foundation, Fiber is far more energy efficient in both bit-wise relative terms and absolute terms than any other legacy fixed (e.g., xDSL, HFC) or cellular broadband technology (e.g., 4G, 5G).

Looking forward and with the increasing challenges of remaining sustainable and contributing to CO2 reduction, it is paramount to consider an energy-optimized fixed and mobile converged network architecture as opposed to today’s approach of optimizing the fixed network separately from the cellular network. As a society, we should expect that the industry works hard to achieve an overall reduction in energy consumption, relaxing the demand on existing green energy infrastructures.

With 5G as of today, we are orders of magnitude from the original NGMN vision of energy consumption of only half of what was consumed by cellular networks ten years ago (i.e., 2014), requiring an overall energy efficiency increase of x2000.

Be aware that many Telcos and Infrastructure providers will use bit-wise energy efficiency when they report on energy consumption. They will generally report impressive gains over time in the energy that networks consume to deliver bits to their customers. This is the least one should expect.

Last but not least, the telco world is not static and is RAT-wise not very clean, as mobile networks will have several RATs deployed simultaneously (e.g., 2G, 4G, and 5G). As such, we rarely (if ever) have apples-to-apples comparisons on cellular energy consumption.

ACKNOWLEDGEMENT.

I greatly acknowledge my wife, Eva Varadi, for her support, patience, and understanding during the creative process of writing this article. I also greatly appreciate the discussion on this topic that I have had with Rudolf van der Berg over the last couple of years. I thank him for pointing out and reminding me (when I forget) of the shortfalls and poor quality of most of the academic work and lobbying activities done in this area.

PS

If you are aiming at a leapfrog in absolute energy reduction of your cellular network, above and beyond what you get with your infrastructure suppliers (e.g., Nokia, Ericsson, Huawei…), I really recommend you take a look at Opanga‘s machine learning-based Joule ML solution. The Joules ML has been proven to reduce RAN energy costs by 20% – 40% on top of what the RAT supplier’s (e.g., Ericsson, Nokia, Huawei, etc.) own energy management solutions may bring.

Disclosure: I am associated with Opanga and on their Industry Advisory Board.

Spectrum in the USA – An overview of Today and a new Tomorrow.

Posted on by
1

This week (Week 17, 2023), I submitted my comments and advice titled “Development of a National Spectrum Strategy (NSS)” to the United States National Telecommunications & Information Administration (NTIA) related to their work on a new National Spectrum Strategy.

Of course, one might ask why, as a European, bother with the spectrum policy of the United States. So hereby, a bit of reasoning for bothering with this super interesting and challenging topic of spectrum policy on the other side of the pond.

A EUROPEAN IN AMERICA.

As a European coming to America (i.e., USA) for the first time to discuss the electromagnetic spectrum of the kind mobile operators love to have exclusive access to, you quickly realize that Europe’s spectrum policy/policies, whether you like them or not, are easier to work with and understand. Regarding spectrum policy, whatever you know from Europe is not likely to be the same in the USA (though physics is still fairly similar).

I was very fortunate to arrive back in the early years of the third millennium to discuss cellular capacity and, as it quickly evolves (“escalates”), too, having a discussion of available cellular frequencies, the associated spectral bandwidth, and whether they really need that 100 million US dollar for radio access expansions.

Why fortunate?

I was one of the first (from my company) to ask all those “stupid” questions whenever I erroneously did not just assume things surely must be the same as in Europe and ended up with the correct answer that in the USA, things are a “little” different and a lot more complicated in terms of the availability of frequencies and what feeds the demand … the spectrum bandwidth. My arrival was followed by “hordes” of other well-meaning Europeans with the same questions and presumptions, using European logic to solve US challenges. And that doesn’t really work (surprised you not should be). I believe my T-Mobile US colleagues and friends over the years surely must have felt like Groundhog Day all over again at every new European visit.

COMPARING APPLES AND ORANGES.

Looking at US spectrum reporting, it is important to note that it is customary to provide the total amount of spectrum. Thus, for FDD spectrum bands, including both the downlink spectrum portion and uplink spectrum part of the cellular frequency band in question. For example, when a mobile network operator (MNO) reports that it has, e.g., 40 MHz of AWS1 spectrum in San Diego (California), it means that it has 2×20 MHz (or 20+20 MHz). Thus, 20 MHz of downlink (DL) services and 20 MHz of uplink (UL) services. For FDD, both the DL and the UL parts are counted. In Europe, historically, we mainly would talk about half the spectrum for FDD spectrum bands. This is one of the first hurdles to get over in meetings and discussions. If not sorted out early can lead to some pretty big misunderstandings (to say the least). To be honest, and in my opinion, providing the full spectrum holding, irrespective of whether a band is used as FDD or TDD, is less ambiguous than the European tradition.

The second “hurdle” is to understand that a USA-based MNO is likely to have a substantial variation in its spectrum holdings across the US geography. An MNO may have a 40 MHz (i.e., 2×20 MHz) PCS spectrum in Los Angeles (California) and only 30 MHz (2×15 MHz) of the same spectrum in New York or only 20 MHz (2×10 MHz) in Miami (Florida). For example, FCC (i.e., the regulator managing non-federal spectrum) uses 734 so-called Cellular Market Areas or CMAs, and there is no guarantee that a mobile operator’s spectrum position will remain the same over these 734 CMAs. Imagine Dutch (or other European) mobile operators having a varying 700 MHz (used for 5G) spectrum position across the 342 municipalities of The Netherlands (or another European country). It takes a lot of imagination … right? And maybe why, we Europeans, shake our heads at the US spectrum fragmentation, or market variation, as opposed to our nice, neat, and tidy market-wise spectrum uniformity. But is the European model so much better (apart from being neat & tidy)? …

… One may argue that the US model allows for spectrum acquisition to be more closely aligned with demand, e.g., less spectrum is needed in low-population density areas and more is required in high-density population areas (where demand will be much more intense). As evidenced by many US auctions, the economics matched the demand fairly well. While the European model is closely aligned with our good traditions of being solid on average … with our feet in the oven and our head in the freezer … and on average all is pretty much okay in Europe.

Figure 1 and 2 below illustrates a mobile operator difference between its spectrum bandwidth spread across the 734 US-defined CMAs in the AWS1 band and how that would look in Europe.

Figure 1 illustrates the average MNO distribution of (left chart) USA AWS1 band (band 4) distribution over the 734 Cellular Market Areas (CMA) defined by the FCC. (right chart) Typical European 3 MNO 2100-band (band-1) distribution across the country’s geographical area. As a rule of thumb for European countries, the spectrum is fairly uniformly distributed across the national MNOs. E.g., if you have 3 mobile operators, the 120 MHz available to band-1 will be divided equally among the 3, and If there are 4 MNOs, then it will be divided by 4. Nevertheless, in Europe, an MNO spectrum position is fixed across the geography.

Figure 2 below is visually an even stronger illustration of mobile operator bandwidth variation across the 734 cellular market areas. The dashed white horizontal line is if the PCS band (a total of 120 MHz or 2×60 MHz) would be shared equally between 4 main nationwide mobile operators ending up at 30 MHz per operator across all CMAs. This would resemble what today is more or less a European situation, i.e., irrespective of regional population numbers, the mobile operator’s spectrum bandwidth at a given carrier frequency would be the same. The European model, of course, also implies that an operator can provide the same quality in peak bandwidth before load may become an issue. The high variation in the US operator’s spectrum bandwidth may result in a relatively big variation in provided quality (i.e., peak speed in Mbps) across the different CMAs.

There is an alternative approach to spectrum acquisition that may also be more spectrally efficient, which the US model is much more suitable for. Aim at a target Hz per Customer (i.e., spectral overhead) and keep this constant within the various market. Of course, there is a maximum realistic amount of bandwidth to acquire, governed by availability (e.g., for PCS, that is, 120 MHz) and competitive bidders’ strength. There will also be a minimum bandwidth level determined by the auction rules (e.g., 5 MHz) and a minimum acceptable quality level (e.g., 10 MHz). However, Figure 2 below reflects more opportunistic spectrum acquisition in CMAs with less than a million population as opposed to a more intelligent design (possibly reflecting the importance of, or lack of, different CMAs to the individual operators).

Figure 2 illustrates the bandwidth variation (orange dots) across the 734 cellular market areas for 4 nationwide mobile network operators in the United States. The horizontal dashed white line is if the four main nationwide operators would equally share the 120 MHz of PCS spectrum (fairly similar to a European situation). MNOs would have the same spectral bandwidth across every CMA. The Minimum – Growing – Maximum dashed line illustrates a different spectrum acquisition strategy, where the operator has fixed the amount of spectrum per customer required and keeps this as a planning rule between a minimum level (e.g., a unit of minimum auctioned bandwidth) and a realistic maximum level (e.g., determined by auction competition, auction ruling, and availability).

Thirdly, so-called exclusive use frequency licenses (as opposed to shared frequencies), as issued by FCC, can be regarded accounting-wise as an indefinitely-lived intangible asset. Thus, once a US-based cellular mobile operator has acquired a given exclusive-use license, that license can be considered disposable to the operator in perpetuity. It should be noted that FCC licenses typically would be issued for a fixed (limited) period, but renewals are routine.

This is a (really) big difference from European cellular frequency licenses that typically expire after 10 – 20 years, with the expired frequency bands being re-auctioned. A European mobile operator cannot guarantee its operation beyond the expiration date of the spectrum acquired, posing substantial existential threats to business and shareholder value. In the USA, cellular mobile operators have a substantially lower risk regarding business continuity as their spectrum, in general, can be regarded as theirs indefinitely.

FCC also operates with a shared-spectrum license model, as envisioned by the Citizens Broadband Radio Service (CBRS) in the 3.55 to 3.7 GHz frequency range (i.e., the C-band). A shared-spectrum license model allows for several types of users (e.g., Federal and non-Federal) and use-cases (e.g., satellite communications, radar applications, national cellular services, local community broadband services, etc..) to co-exist within the same spectrum band. Usually, such shared licenses come with firm protection of federal (incumbent) users that allows commercial use to co-exist with federal use, though with the federal use case taking priority over the non-federal. A really good overview of the CBRS concept can be found in “A Survey on Citizens Broadband Radio Service (CBRS)” by P. Agarwal et al.. Wireless Innovation Forum published on 2022 a piece on “Lessons Learned from CBRS” which provides a fairly nuanced, although somewhat negative, view on spectrum sharing as observed in the field and within the premises of the CBRS priority architecture and management system.

Recent data around FCC’s 3.5 GHz (CBRS) Auction 105 would indicate that shared-licensed spectrum is valued at a lower USD-per-MHz-pop (i.e., 0.14 USD-per-MHz-pop) than exclusive-use license auctions in 3.7 GHz (Auction 107; 0.88 USD-per-MHz-pop) and 3.45 GHz (Auction 110; 0.68 USD-per-MHz-pop). The duration of the shared-spectrum license in the case of the Auction 105 spectrum is 10 years after which it is renewed. Verizon and Dish Networks were the two main telecom incumbents that acquired substantial spectrum in Auction 105. AT&T did not acquire and T-Mobile US only picked close to nothing (i.e., 8 licenses).

THE STATE OF CELLULAR PERFORMANCE – IN THE UNITED STATES AND THE REST OF THE WORLD.

Irrespective of how one feels about the many mobile cellular benchmarks around in the industry (e.g., Ookla Speedtest, Umaut benchmarking, OpenSignal, etc…), these benchmarks do give an indication of the state of networks and how those networks utilize the spectral resources that mobile companies have often spend hundreds of millions, if not billions, of US dollars acquiring and not to underestimate in cost and time, spectrum clearing or perfecting a “second-hand” spectrum may incur for those operators.

So how do US-based mobile operators perform in a global context? We can get an impression, although very 1-dimensional, from Figure 1 below.

Figure 3 illustrates the comparative results of Ookla Speedtest data in median downlink speed (Mbps) for various countries. The selection of countries provides a reasonable representation of maximum and minimum values. To give an impression of the global ranking as of February 2023; South Korea (3), Norway (4), China (7), Canada (17), USA (19), and Japan (48). As a reminder, the statistic is based on the median of all measurements per country. Thus, half of the measurements were above the median speed value, and the other half were below. Note: median values from 2020 to 2017 are estimated as Ookla did only provide average numbers.

Ookla’s Speedtest rank (see Figure 3 above) positions the United States cellular mobile networks (as an average) among the Top-20. Depending on the ambition level, that may be pretty okay or a disappointment. However, over the last 24 months, thanks to the fast 5G deployment pace at 600 MHz, 2.5 GHz, and C-band, the US has leapfrogged (on average) its network quality which for many years did not improve much due to little spectrum availability and huge capital investment levels. Something that the American consumer can greatly enjoy irrespective of the relative mobile network ranking of the US compared to the rest of the world. South Korea and Norway are ranked 3 and 4, respectively, regarding cellular downlink (DL) speed in Mbps. The above figure also shows a significant uplift in the speed at the time of introducing 5G in the cellular operators’ networks worldwide.

How to understand the supplied cellular network quality and capacity that the consumer demand and hopefully also enjoy? Let start with the basics:

Figure 4 illustrates one of the most important (imo) to understand about creating capacity & quality in cellular networks. You need frequency bandwidth (in MHz), the right technology boosting your spectral efficiency (i.e., the ability to deliver bits per unit Hz), and sites (sectors, cells, ..) to deploy the spectrum and your technology. That’s pretty much it.

We might be able to understand some of the dynamics of Figure 3 using Figure 4, which illustrates the fundamental cellular quality (and capacity) relationship with frequency bandwidth, spectral efficiency, and the number of cells (or sectors or sites) deployed in a given country.

Thus, a mobile operator can improve its cellular quality (and capacity) by deploying more spectrum acquired on its existing network, for example, by auctions, leasing, sharing, or other arrangements within the possibilities of whatever applicable regulatory regime. This option will exhaust as the operator’s frequency spectrum pool is deployed across the cellular network. It leaves an operator to wait for an upcoming new frequency auction or, if possible, attempt to purchase additional spectrum in the market (if regulation allows) that may ultimately include a merger with another spectrum-rich entity (e.g., AT&T attempt to take over T-Mobile US). All such spectrum initiatives may take a substantial amount of time to crystalize, while customers may experience a worsening in their quality. In Europe, the licensed spectrum becomes available in cycles of 10 – 20 years. In the USA, exclusive-use licensed spectrum typically would be a once-only opportunity to acquire (unless you acquire another spectrum-holding entity later, e.g., Metro PCS, Sprint, AT&T’s attempt to acquire T-Mobile, …).

Another part of the quality and capacity toolkit is for the mobile operator to choose appropriately spectral efficient technologies that are supported by a commercially available terminal ecosystem. Firstly, migrate frequency and bandwidth away from currently deployed legacy radio-access technology (e.g., 2G, 3G, …) to newer and spectrally more efficient ones (e.g., 4G, 5G, …). This migration, also called spectral re-farming, requires a balancing act between current legacy demand versus the future expectations of demand in the newer technology. In a modern cellular setting, the choice of antenna technology (e.g., massive MiMo, advanced antenna systems, …) and type (e.g., multi-band) is incredibly important for boosting quality and capacity within the operators’ cellular networks. Given that such choices may result in redesigning existing site infrastructure, it provides an opportunity to optimize the existing infrastructure for the best coverage of the consolidated spectrum pool. It is likely that the existing infra was designed with a single or only a few frequencies in mind (e.g., PCS, PCS+AWS, …) as well as legacy antennas, and the cellular performance is likely improved by considering the complete pool of frequencies in the operator’s spectrum holding. The mobile operator’s game should always be to achieve the best possible spectral efficiency considering demand and economics (i.e., deploying 64×64 massive MiMo all over a network may be the most spectrally efficient solution, theoretically, but both demand and economics would rarely support such an apparently “silly” non-engineering strategy). In general, this will be the most frequently used tool in the operators’ quality/capacity toolkit. I expect to see an “arms race” between operators deploying the best and most capable antennas (where it matters), as it will often be the only way to differentiate in quality and capacity (if everything else is almost equal).

Finally, the mobile operator can deploy more site locations (macro and small cells), if permitting allows, or more sectors by sectorization (e.g., 3 → 4, 4 → 5 sectors) or cell split if the infrastructure and landlord allows. If there remains unused spectral bandwidth in the operator’s spectrum pool, the operator may likely choose to add another cell (i.e., frequency band) to the existing site. Particular adding new site locations (macro or small cell) is the most complex path to be taken and, of course, also often the least economic path.

Thus, to get a feeling for the Ookla Speedtest, which is a country average, results of Figure 3, we need, as a starting point, to have the amount of spectral bandwidth for the average cellular mobile operator. This is summarised in below’s Table 1.

Table 1 provides, per country, the average amount of Low-band (≤ 1 GHz), Mid-band (1 GHz to 2.1 GHz), 2.3 & 2.5 GHz bands, Sub-total bandwidth before including the C-band, the C-band (3.45 to 4.2 GHz) and the Total bandwidth. The table also includes the Ookla Global Speedtest DL Mbps and Global Rank as of February 2023. I have also included the in-country mobile operator variation within the different categories, which may indicate what kind of performance range to expect within a given country.

It does not take too long to observe that there is only an apparently rather weak correlation between spectrum bandwidth (sub-total and total) and the observed DL speed (even after rescaling to downlink spectrum only). Also, what is important is, of course, how much of the spectrum is deployed. Typically low and medium bands will be deployed extensively, while other high-frequency bands may only have been selectively deployed, and the C-band is only in the process of being deployed (where it is available). What also plays a role is to what degree 5G has been rollout across the network, how much bandwidth has been dedicated to 5G (and 4G), and what type of advanced antenna system or massive MiMo capabilities has been chosen. And then, to provide a great service, a network must have a certain site density (or coverage) compared to the customer’s demand. Thus, it is to be expected that the number of mobile site locations, and the associated number of frequency cells and sectors, will play a role in the average speed performance of a given country.

Figure 5 illustrates how the DL speed in Mbps correlates with the (a) total amount of spectrum excluding the C-band (still not widely deployed), (b) Customers per Site that provides a measure of the customer load at the site location level. The more customers load a site or compete for radio resources (i.e., MHz), the lower the experience. Finally, (c) The higher the Site times, the bandwidth is compared to the number of customers. More quality can be provided (as observed with the positive correlation). The data is from Table 1.

Figure 5 shows that load (e.g., customers per site) and available capacity (e.g., sites x bandwidth) relative to customers are strongly correlated with the experienced quality (e.g., speed in Mbps). The comparison between the United States and China is interesting as both countries with a fairly similar surface area (i.e., 9.8 vs. 9.6 million sq. km), the USA has a little less than a quarter of the population, and the average mobile US operator would have about one-third of the customers compared to the average Chinese operator (note: China mobile dominates the average). The Chinese operator, ignoring C-band, would have ca. 25 MHz or ~+20% (~50 MHz or ca. +10% if C-band is included) more than the US operator. Regarding sites, China Mobile has been reported to have millions of cell site locations (incl. lots of small cells). The US operator’s site count is in the order of hundreds of thousands (though less than 200k currently, including small cells). Thus, Chinese mobile operators have between 5x to 10x the number of site locations compared to the American ones. While the difference in spectrum bandwidth has some significance (i.e., China +10% to 20% higher), the huge relative difference in site numbers is one of the determining factors in why China (i.e., 117 Mbps) gets away with a better speed test score that is better than the American one (i.e., 85 Mbps). While theoretically (and simplistically), one would expect that the average Chinese mobile operator should be able to provide more than twice the speed as compared to the American mobile operator instead of “only” about 40% more, it stands to show that the radio environment is a “bit” more complex than the simplistic view.

Of course, the US-based operator could attempt to deploy even more sites where it matters. However, I very much doubt that this would be a feasible strategy given permitting and citizen resistance to increasing site density in areas where it actually would be needed to boost the performance and customer experience.

Thus, the operator in the United States must acquire more spectrum bandwidth and deploy that where it matters to their customers. They also need to continue to innovate on leapfrogging the spectral efficiency of the radio access technologies and deploy increasingly more sophisticated antenna systems across their coverage footprint.

In terms of sectorization (at existing locations), cell split (adding existing spectrum to an existing site), and/or adding more sophisticated antenna systems is a matter of Capex prioritization and possibly getting permission from the landlord. Acquiring new spectrum … well, that depends on such new spectrum somehow becomes available.

Where to “look” for more spectrum?

WHERE COULD MORE SPECTRUM COME FROM?

Within the so-called “beachfront spectrum” covering the frequency range from 225 MHz to 4.2 GHz (according to NTIA), only about 30% (ca. 1GHz of bandwidth within the frequency range from 600 MHz to 4.2 GHz) is exclusively non-Federal, and mainly with the mobile operators as exclusive use licenses deployed for cellular mobile services across the United States. Federal authorities exclusively use a bit less than 20% (~800 MHz) for communications, radars, and R&D purposes. This leaves ca. 50% (~2 GHz) of the beachfront spectrum shared between Federal authorities and commercial entities (i.e., non-Federal).

For cellular mobile operators, exclusive use licenses would be preferable (note: at least at the current state of the relevant technology landscape) as it provides the greatest degree of operational control and possibility to optimize spectral efficiency, avoiding unacceptable levels of interference either from systems or towards systems that may be sharing a given frequency range.

The options for re-purposing the Federal-only spectrum (~800 MHz) could, for example, be either (a) moving radar systems’ operational frequency range out of the beachfront spectrum range to the degree innovation and technology supports such a migration, (b) modernizing radar systems with a focus of making these substantially more spectrally efficient and interference-resistant, (c) migrated federal-only communications services to commercially available systems (e.g., 5G federal-only slicing) similar to the trend of migrating federal legacy data centers to the public cloud. Within the shared frequency portion with the ~2 GHz of bandwidth, it may be more challenging as considerable commercial interests (other than mobile operators) have positioned that business at and around such frequencies, e.g., within the CBRS frequency range. This said, there might also be opportunities within the Federal use cases to shift applications towards commercially available communication systems or to shift them out of the beachfront range. Of course, in my opinion, it always makes sense to impose (and possibly finance) stricter spectral efficiency conditions, triggering innovation on federal systems and commercial systems alike within the shared portion of the beachfront spectrum range. With such spectrum strategies, it appears compelling that there are high likelihood opportunities for creating more spectrum for exclusive license use that would safeguard future consumer and commercial demand and continuous improvement of customer experience that comes with the future demand and user expectations of the technology that serves them.

I believe that the beachfront should be extended beyond 4.2 GHz. For example aligning with band-79, whose frequency range extends from 4.4 GHz to 5 GHz, allows for a bandwidth of 600 MHz (e.g., China Mobile has 100 MHz in the range from 4.8 GHz to 4.9 GHz). Exploring additional re-purposing opportunities for exclusive use licenses in what may be called the extended beachfront frequency range from 4.2 GHz up to 7.2 GHz should be conducted with priority. Such a study should also consider the possibility of moving the spectrum under exclusive and shared federal use to other frequency bands and optimizing the current federal frequency and spectrum allocation.

The NTIA, that is, the National Telecommunications and Information Administration, is currently (i.e., 2023) for the United States developing a National Spectrum Strategy (NSS) and the associated implementation plan. Comments and suggestions to the NSS were possible until the 18th of April, 2023. The National Spectrum Strategy should address how to create a long-term spectrum pipeline. It is clear that developing a coherent national spectrum strategy is critical to innovation, economic competition, national security, and maybe re-capture global technology leadership.

So who is the NTIA? What do they do that FCC doesn’t already do? (you may possibly ask).

WHO MANAGES WHAT SPECTRUM?

Two main agencies in the US manage the frequency spectrum, the FCC and the NTIA.The Federal Communications Commission, the FCC for short, is an independent agency that exclusively regulates all non-Federal spectrum use across the United States. FCC allocates spectrum licenses for commercial use, typically through spectrum auctions. A new or re-purposed commercialized spectrum has been reclaimed from other uses, both from federal uses and existing commercial uses. Spectrum can be re-purposed either because newer, more spectrally efficient technologies become available (e.g., the transition from analog to digital broadcasting) or it becomes viable to shift operation to other spectrum bands with less commercial value (and, of course, without jeopardizing existing operational excellence). It is also possible that spectrum, previously having been for exclusive federal use (e.g., military applications, fixed satellite uses, etc..), can be shared, such as the case with Citizens Broadband Radio Service (CBRS), which allows non-federal parties access to 150 MHz in the 3.5 GHz band (i.e., band 48). However, it has recently been concluded that (centralized) dynamic spectrum sharing only works in certain use cases and is associated with considerable implementation complexities. Multiple parties with possible vastly different requirements co-existence within a given band is very much work-in-progress and may not be consistent with the commercialized spectrum operation required for high-quality broadband cellular operation.

In parallel with the FCC, we have the National Telecommunications and Information Administration, NTIA for short. NTIA is solely responsible for authorizing Federal spectrum use. It also acts as the President of the United State’s principal adviser on telecommunications policies, coordinating the views of the Executive Branch. NTIA manages about 2,398 MHz (69%) within the so-called “beachfront spectrum” range of 225 MHz to 3.7 GHz (note: I would let that Beachfront go to 7 GHz, to be honest). Of the total of 3,475 MHz, 591 MHz (17%) is exclusively for Federal use, and 1,807 MHz (52%) is shared (or coordinated) between Federal and non-Federal. Thus, leaving 1,077 MHz (31%) for exclusive commercial use under the management of the FCC.

NTIA, in collaboration with the FCC, has been instrumental in the past in freeing up substantial C-band spectrum, 480 MHz in total, of which 100 MHz is conditioned on prioritized sharing (i.e., Auction 105), for commercial and shared use that subsequently has been auctioned off over the last 3 years raising USD 109 billion. In US Dollar (USD) per MHz per population count (pop) we have on average ca. USD 0.68 per MHz-pop from the C-band auctions in the US, compared to USD 0.13 per MHz-pop in Europe C-band auctions, and USD 0.23 per MHz-pop in APAC auctions. It should be remember that the United States exclusive-use spectrum licenses can be regarded as an indefinite-lived intangible asset while European spectrum rights expire between 10 and 20 years. This may explain a big part of the pricing difference between US-based spectrum pricing and that of Europe and Asia.

NTIA and FCC jointly manage all the radio spectrum, licensed (e.g., cellular mobile frequencies, TV signals, …) and unlicensed (e.g., WiFi, MW Owens, …) of the United States, NTIA for Federal use, and FCC for non-Federal use (put simply). FCC is responsible for auctioning spectrum licenses and is also authorized to redistribute licenses.

RESPONSE TO NTIA’s National Spectrum Strategy Request for Comments

Here are some of key points to consider for developing a National Spectrum Strategy (NSS).

  • The NTIA National Spectrum Strategy (NSS) should focus on creating a long-term spectrum pipeline. Developing a coherent national spectrum strategy is critical to innovation, economic competition, national security, and global technology leadership.
  • NTIA should aim at significant amounts of spectrum to study and clear to build a pipeline. Repurposing at least 1,500 Mega Hertz of spectrum perfected for commercial operations is good initial target allowing it to continue to meet consumer, business, and societal demand. It requires more than 1,500 Mega Hertz to be identified for study.
  • NTIA should be aware that the mobile network quality strongly correlates with the mobile operators’ spectrum available for their broadband mobile service in a global setting.
  • NTIA must remember that not all spectrum is equal. As it thinks about a pipeline, it must ensure its plans are consistent with the spectrum needs of various use cases of the wireless sectors. The NSS is a unique opportunity for NTIA to establish a more reliable process and consistent policy for making the federal spectrum available for commercial use. NTIA should reassert its role, and that of the FCC, as the primary federal and commercial regulator of spectrum policy.

A balanced spectrum policy is the right approach. Given the current spectrum dynamics, the NSS should prioritize identifying exclusive-use licensed spectrum instead of, for example, attempting co-existence between commercial and federal use.

Spectrum-band sharing between commercial communications networks and federal communications, or radar systems, may impact the performance of all the involved systems. Such practice compromises the level of innovation in modern commercialized communications networks (e.g., 5G or 6G) to co-exist with the older legacy systems. It also discourages the modernization of legacy federal equipment.

Only high-power licensed spectrum can provide the performance necessary to support nationwide wireless with the scale, reliability, security, resiliency, and capabilities consumers, businesses, and public sector customers expect.

Exclusive use of licensed spectrum provides unique benefits compared to unlicensed and shared spectrum. Unlicensed spectrum, while important, is only suitable for some types of applications, and licensed spectrum under shared access frameworks by CBRS is unsuited for serving as the foundation for nationwide mobile wireless networks.

Allocating new spectrum bands for the exclusive use of licensed spectrum positively impacts the entire wireless ecosystem, including downstream investments by equipment companies and others who support developing and deploying wireless networks. Insufficient licensed spectrum means increasingly deteriorating customer experience and lost economic growth, jobs, and innovation.

Other countries are ahead of the USA in developing plans for licensed spectrum allocations, targeting the full potential of the spectrum range from 300 MHz up to 7 GHz (i.e., the beachfront spectrum range), and those countries will lead the international conversation on licensed spectrum allocation. The NSS offers an opportunity to reassert U.S. leadership in these debates.

NTIA should also consider the substantial benefits and economic value of leading the innovation in modernizing the legacy spectrally in-efficient non-commercial communications and radar systems occupying vast spectrum resources.

Exclusive-use licensed spectrum has inherent characteristics that benefit all users in the wireless ecosystem.

Consumer demand for mobile data is at an all-time high and only continues to surge as demand grows for lightning-fast and responsive wireless products and services enabled by licensed spectrum.

With an appropriately designed and well-sized spectrum pipeline, demand will remain sustainable as supplied spectrum capacity compared to the demand will remain or exceed today’s levels.

Networks built on licensed spectrum are the backbone of next-generation innovative applications like precision agriculture, telehealth, advanced manufacturing, smart cities, and our climate response.

Licensed spectrum is enhancing broadband competition and bridging the digital divide by enabling 5G services like 5G Fixed Wireless Access (FWA) in areas traditionally dominated by cable and in rural areas where fiber is not cost-effective to deploy.

NTIA should identify the midband spectrum (e.g., ~2.5GHz to ~7GHz) and, in particular, frequencies above the C-band for licensed spectrum. That would be the sweet spot for leapfrogging broadband speed and capacity necessary to power 5G and future generations of broadband communications networks.

The National Spectrum Strategy is an opportunity to improve the U.S. Government’s spectrum management process.

The NSS allows NTIA to develop a more consistent and better process for allocating spectrum and providing dispute resolution.

The U.S. should handle mobile networks without a new top-down government-driven industrial policy to manage mobile networks. A central planning model would harm the nation, severely limiting innovation and private sector dynamism.

Instead, we need a better collaboration between government agencies with NTIA and the FCC as the U.S. Government agencies with clear authority over the nation’s spectrum. The NSS also should explore mechanisms to get federal agencies (and their associated industry sectors) to surface their concerns about spectrum allocation decisions early in the process and accept NTIA’s role as a mediator in any dispute.

ACKNOWLEDGEMENT.

I greatly acknowledge my wife, Eva Varadi, for her support, patience, and understanding during the creative process of writing this article. Of course, throughout the years of being involved in T-Mobile US spectrum strategy, I have enjoyed many discussions and debates with US-based spectrum professionals, bankers, T-Mobile US colleagues, and very smart regulatory policy experts in Deutsche Telekom AG. I have the utmost respect for their work and the challenges they have faced and face. For this particular work, I cannot thank Roslyn Layton, PhD enough for nudging me into writing the comments to NTIA. By that nudge, this little article is a companion to my submission about the US Spectrum as it stands today and what I would like to see with the upcoming National Spectrum Strategy. I very much recommend reading Roslyn’s far more comprehensive and worked-through comments to the NTIA NSS request for advice. A final thank you to John Strand (who keeps away from Linkedin;-) of Strand Consult for challenging my way of thinking and for always stimulating new ways of approaching problems in our telecom sector. I very much appreciate our discussions.

ADDITIONAL MATERIAL.

  1. Kim Kyllesbech Larsen, “NTIA-2023-003. Development of a National Spectrum Strategy (NSS)”, National Spectrum Strategy Request for Comment Responses April 2023. See all submissions here.
  2. Roslyn Layton, “NTIA–2023–0003. Development of a National Spectrum Strategy (NSS)”, National Spectrum Strategy Request for Comment Responses April 2023..
  3. Ronald Harry Coase, “The Federal Communications Commission”, The Journal of Law & Economics, Vol. 2 (October 1959), pp. 1- 40. In my opinion, a must-read for anyone who wants to understand the US spectrum regulation and how it came about.
  4. Kenneth R. Carter, “Policy Lessons from Personal Communications Services: Licensed vs. Unlicensed Spectrum Access,” 2006, Columbus School of Law. An interesting perspective on licensed and unlicensed spectrum access.
  5. Federal Communication Commission (FCC) assigned areas based on the relevant radio licenses. See also FCC Cellular Market Areas (CMAs).
  6. FCC broadband PCS band plan, UL:1850-1910 MHz & DL:1930-1990 MHz, 120 MHz in total or 2×60 MHz.
  7. Understanding Federal Spectrum Use is a good piece from NTIA about the various federal use of spectrum in the United States.
  8. Ookla’s Speedtest Global Index for February 2023. In order to get the historical information use the internet archive, also called “The Wayback Machine.”
  9. I make extensive use of the Spectrum Monitoring site, which I can recommend as one of the most comprehensive sources of frequency allocation data worldwide that I have come across (and is affordable to use).
  10. FCC Releases Rules for Innovative Spectrum Sharing in 3.5 GHz Band.
  11. 47 CFR Part 96—Citizens Broadband Radio Service. Explain the hierarchical spectrum-sharing regime of and priorities given within the CBRS.