Direct-to-Cell: A First Look into Starlink’s Direct Satellite-to-Device Radio Access Network through Crowdsourced Measurements

Low Earth Orbit (LEO) satellite mega-constellations have emerged as a viable access solution for broadband connectivity in underserved areas. In 2024, Starlink, in partnership with T-Mobile, began beta testing an SMS-only Supplemental Coverage from Space (SCS) service. This marks the first largescale deployment of Direct Satellite-to-Device (DS2D) communications, allowing unmodified smartphones to connect directly to spaceborne base stations. This paper presents the first measurement study of deployed DS2D technologies. Using crowdsourced mobile network data from the U.S. between October 2024 and July 2025, we provide evidence-based insights into the capabilities, limitations, and future evolution of DS2D technologies for extending mobile connectivity. We find a strong correlation between the number of satellites deployed, the number of unique cell identifiers measured, and the volume of measurements, concentrated in accessible areas with poor terrestrial network coverage, such as national parks and sparsely populated counties. Stable physical-layer measurements were observed throughout the period, with a 24-dB lower median RSRP and a 3-dB higher RSRQ compared to terrestrial networks, reflecting the SMS-only usage of the DS2D network during this period. Based on the SINR measurements collected, we estimate the expected performance of the announced DS2D mobile data service to be around 3 Mbps per beam in outdoor conditions. We also discuss strategies to expand this capacity up to 18 Mbps in the future, depending on key regulatory and business decisions, including allowable out-ofband emissions, permitted number of satellites, and availability of spectrum and orbital resources.

低地球轨道（LEO）卫星巨型星座已成为服务欠缺地区宽带连接的可行接入解决方案。2024年，Starlink与T-Mobile合作，开始对一项仅支持短信（SMS）的“来自太空的补充覆盖”（SCS）服务进行Beta测试。这标志着卫星直连设备（DS2D）通信的首次大规模部署，允许未经修改的智能手机直接连接到星载基站。

本文对已部署的DS2D技术进行了首次测量研究。我们使用2024年10月至2025年7月期间来自美国的众包移动网络数据，为DS2D技术在扩展移动连接性方面的能力、局限性和未来演进提供了基于实证的见解。

我们发现，部署的卫星数量、测量到的唯一小区标识符数量以及测量数据量之间存在很强的相关性，这些测量数据集中在地面网络覆盖较差但（用户）可达的区域，例如国家公园和人口稀少的县。在整个（测量）期间，观察到了稳定的物理层测量值，与地面网络相比，其RSRP（参考信号接收功率）中值低24 dB，RSRQ（参考信号接收质量）中值高3 dB，这反映了DS2D网络在此期间仅限短信的使用情况。

基于收集到的SINR（信号与干扰加噪声比）测量值，我们 估计已宣布的DS2D移动数据服务在室外条件下的预期性能约为每波束3 Mbps。 我们还讨论了未来将该容量扩展至18 Mbps的策略，这取决于关键的监管和商业决策，包括允许的带外发射、许可的卫星数量以及频谱和轨道资源的可用性。

Introduction

Over the past decade, mobile networks have become an essential means of accessing the Internet for billions of users worldwide. Despite a remarkable growth in mobile network coverage, significant portions of the global population remain underserved by terrestrial cellular infrastructure, particularly in remote, rural, and sparsely populated areas.

In recent years, particularly from 2020, Low Earth Orbit (LEO) mega-constellations have become a viable and increasingly widespread Non-Terrestrial Network (NTN) Internet access solution for fixed broadband services. Networks like SpaceX’s Starlink, Eutelsat’s OneWeb or Amazon’s Kuiper are successfully addressing longstanding limitations of traditional geostationary satellite communications. By operating at significantly lower altitudes (500–1200 km), these constellations offer latency levels comparable to those of terrestrial networks (typically tens of milliseconds vs. the hundreds associated with geostationary satellites), supporting real-time applications such as video conferencing and online gaming [1].

Following technological developments in fixed broadband connectivity, satellite-based cellular communications, known as Direct Satellite-to-Device (DS2D), have recently emerged as a promising solution to provide seamless connectivity directly to standard unmodified smartphones. Starlink, the satellite communications division of SpaceX, has emerged as the global leader in DS2D communications by rapidly deploying a non-3GPP-NTN-compliant solution that leverages terrestrial International Mobile Telecomunications (IMT) spectrum. In partnership with T-Mobile and using its spectrum, Starlink has been the first company to launch large-scale beta testing of a Supplemental Coverage from Space (SCS) service. The system is initially limited to Short Message Service (SMS), with support for voice and data services planned by late 2025.

在过去十年中，移动网络已成为全球数十亿用户接入互联网的重要途径。尽管移动网络覆盖范围实现了显著增长，但全球仍有很大一部分人口未被地面蜂窝基础设施充分覆盖，特别是在偏远、农村和人口稀少的地区。

近年来，特别是自2020年以来，低地球轨道（LEO）巨型星座已成为一种可行且日益普及的、用于固定宽带服务的非地面网络（NTN）互联网接入解决方案。诸如SpaceX的Starlink、Eutelsat的OneWeb或Amazon的Kuiper等网络，正在成功解决传统地球静止轨道卫星通信的长期局限性。通过在显著更低的高度（500-1200公里）运行，这些星座提供了与地面网络相当的延迟水平（通常为几十毫秒，而地球静止轨道卫星则为数百毫秒），从而支持了视频会议和在线游戏等实时应用[1]。

继固定宽带连接技术发展之后，被称为“卫星直连设备”（DS2D）的星基蜂窝通信近来已成为一种有前景的解决方案，可直接为标准、未经修改的智能手机提供无缝连接。

SpaceX的卫星通信部门Starlink，通过快速部署一个利用地面国际移动通信（IMT）频谱的非3GPP-NTN兼容解决方案，已成为DS2D通信领域的全球领导者。Starlink与T-Mobile合作并使用其频谱，已成为第一家对“来自太空的补充覆盖”（SCS）服务进行大规模Beta测试的公司。该系统初期仅限于短信服务（SMS），并计划于2025年底前支持语音和数据服务。

In this article, we examine Starlink’s SCS services and characterize its DS2D network—commonly referred to by Starlink as Direct-to-Cell. Using a large-scale dataset provided by the company Weplan Analytics comprising millions of crowdsourced measurements collected from October 2024 to July 2025, our analysis offers valuable insight into the implications of DS2D network design for mobile network planning and spectrum policy. The main contributions of this work are:

• The first measurement-based study of a commercially deployed DS2D network, including both spatial and temporal analyses.

• An empirical evaluation of physical-layer Radio Access Network (RAN) parameters, namely Reference Signal Received Power (RSRP), Reference Signal Received Quality (RSRQ) and Signal-to-Interference-plus-Noise Ratio (SINR).

• An assessment of the projected performance of future data services and strategies for capacity expansion.

在本文中，我们研究了Starlink的SCS服务，并对其DS2D网络——即Starlink通常所称的“直连蜂窝”（Direct-to-Cell）——进行了特性刻画。我们使用了由Weplan Analytics公司提供的大规模数据集，其中包含从2024年10月到2025年7月收集的数百万条众包测量数据。我们的分析为DS2D网络设计对移动网络规划和频谱政策的影响提供了宝贵的见解。本文的主要贡献包括：

• 对商用DS2D网络的首次基于测量的研究，包括空间和时间维度的分析
• 对物理层无线接入网（RAN）参数的实证评估，即参考信号接收功率（RSRP）、参考信号接收质量（RSRQ）和信号与干扰加噪声比（SINR）
• 对未来数据服务的预期性能评估以及容量扩展策略的探讨

The remainder of this paper is structured as follows. Section II provides background on the evolution of LEO satellite networks and summarizes related research. Section III describes the dataset and our analytical approach. Section IV presents and discusses the results. Finally, Section V offers conclusions and suggests future research.

本文的其余部分结构如下：第二节介绍了LEO卫星网络的演进背景并总结了相关研究。第三节描述了数据集和我们的分析方法。第四节展示并讨论了结果。最后，第五节提供了结论并提出了未来的研究方向。

Background

A. Technological Evolution and Emerging Ecosystem

LEO satellite communications have progressed considerably in recent decades, driven by satellite miniaturization, improved launch capabilities, and the growing demand for ubiquitous connectivity. Early LEO constellations in the late 1990s were primarily designed to deliver voice and low-speed data services to remote and underserved areas. A new phase in LEO communications emerged in the late 2010s with the advent of megaconstellations [2]. Initiatives such as SpaceX’s Starlink and OneWeb began deploying extensive LEO satellite networks to provide broadband Internet access globally.

DS2D has been further enabled by significant progress in electronics, embedded systems, digital signal processing, and beamforming technologies, among others [3]. Specifically, DS2D connectivity relies on the satellites being equipped with advanced phased array antennas, highly sensitive radio receivers, and high-power transmitters. Moreover, compensation for Doppler shift and delay is essential to overcome the physical constraints of mobile-to-satellite links, and softwaredefined radio (SDR) is necessary to adapt the system to varying frequencies and technologies.

3GPP specifications have recently incorporated NTN in Release 17. Protocols have been enhanced to address the distinctive characteristics of satellite-based communications, such as the larger propagation distance between user terminals and satellites, which impacts procedures including hybrid automatic repeat request (HARQ) and random access. In addition, system information blocks have been extended to convey satellite-specific parameters, such as orbital position. Subsequent releases introduce further enhancements, focusing on improved coverage (potentially extending to indoor environments), support for neighbor cell measurements, and reduced battery consumption.

Starlink has pioneered DS2D deployments with a strategy focused on a dense LEO constellation at the lowest feasible orbital altitudes (currently primarily ∼ 550 km and selectively ∼ 350 km subject to coordination with the NASA). In contrast, its main competitor, AST SpaceMobile relies on a less dense constellation at higher altitudes (∼ 700 km), supported by very large antenna arrays. Although both operators have so far relied primarily on IMT spectrum obtained through agreements with mobile network operators in different countries, they are now pursuing their own terrestrial and satellite spectrum. Other actors in the emerging DS2D ecosystem include Lynk Global, Skylo, and Viasat-Inmarsat. Key use cases comprise ubiquitous connectivity for IoT, supplementary coverage for terrestrial mobile networks, SOS alerts, emergency communications, and maritime and aeronautical connectivity.

近几十年来，在卫星小型化、发射能力提高以及对泛在连接需求日益增长的推动下，LEO卫星通信取得了长足的进步。20世纪90年代末的早期LEO星座主要设计为向偏远和服务欠缺地区提供语音和低速数据服务。随着巨型星座的出现，LEO通信在21世纪10年代末进入了一个新阶段[2]。诸如SpaceX的Starlink和OneWeb等项目开始部署广泛的LEO卫星网络，以在全球范围内提供宽带互联网接入。

DS2D的发展还得益于电子学、嵌入式系统、数字信号处理和波束成形技术等方面的重大进展[3]。具体而言，DS2D连接依赖于卫星装备先进的相控阵天线、高灵敏度射频接收机和高功率发射机。此外，多普勒频移和延迟的补偿对于克服移动设备到卫星链路的物理限制至关重要，而软件定义无线电（SDR）则是使系统适应不同频率和技术所必需的。

3GPP规范在Release 17中最近纳入了NTN。协议已经得到增强，以应对星基通信的独有特性，例如用户终端和卫星之间更远的传播距离，这会影响包括混合自动重传请求（HARQ）和随机接入在内的程序。此外，系统信息块（SIB）也得到了扩展，以传递卫星特定的参数，例如轨道位置。后续的版本将引入进一步的增强，重点关注改善覆盖范围（可能扩展到室内环境）、支持邻近小区测量以及降低电池消耗。

Starlink率先采用了DS2D部署策略，专注于在最低可行轨道高度（目前主要为～550公里，并根据与NASA的协调选择性地采用～350公里）部署密集的LEO星座。相比之下，其主要竞争对手AST SpaceMobile则依赖于更高轨道（～700公里）上密度较低的星座，并辅以超大型天线阵列。尽管迄今为止，这两家运营商主要依赖通过与不同国家的移动网络运营商达成协议获得的IMT频谱，但它们现在正寻求获得自己的地面和卫星频谱。这个新兴DS2D生态系统中的其他参与者包括Lynk Global、Skylo和Viasat-Inmarsat。关键用例包括物联网（IoT）的泛在连接、地面移动网络的补充覆盖、SOS警报、应急通信以及海事和航空连接。

B. Starlink’s Direct Satellite-to-Device (DS2D)

Approach Starlink has reshaped the satellite telecommunication sector by deploying over 7 000 satellites since its first launches in 2019. The bent-pipe architecture initially adopted [4] enabled rapid deployment using simpler satellite designs at the cost of deploying a dense network of ground stations to ensure continuous coverage. From 2022 onwards, Starlink progressively introduced laser inter-satellite links (ISL), fully adopted in its second-generation (“Gen2”) fleet, enabling onboard signal processing and space-based routing. This transition from a ground-dependent relay to a space-based mesh brings expanded coverage, improved resilience, and enhanced endto-end performance [1].

Rather than awaiting the completion of 3GPP standardisation of NTN, Starlink pursued an early deployment strategy by integrating conventional LTE eNodeB payloads into its satellites, complemented with proprietary adaptations for NTN operation. This allows smartphones to connect natively using terrestrial mobile standards, while traffic is handled either through the company’s ground station network or, when necessary, routed across the constellation via ISLs [5].

SpaceX announced the completion of its first DS2D constellation in December 2024, consisting of approximately 400 satellites—barely 6% of their deployed fleet. In addition to the SCS service provided in the U.S. through the partnership with T-Mobile, the constellation also offers SCS connectivity in New Zealand and Australia via a partnership with operators and is currently under testing with operators in several other countries.

In November 2024, the FCC granted partial approval for Starlink’s DS2D operations to initiate the SMS-only beta testing, but imposed several conditions to prevent harmful interference with terrestrial and satellite incumbents, including the requirement to stop transmission immediately upon any substantiated claim of interference, limitations on power flux density levels for aggregate out-of-band emissions (OOBE), and mandatory coordination procedures with incumbents.1

Subsequently, in March 2025, the FCC approved 2 a conditional waiver allowing Starlink to increase its OOBE by 9.4 dB, despite formal objections from terrestrial operators, including AT&T and Verizon, and EchoStar. The waiver stipulated that Starlink must continue to adhere to strict OOBE limits and engage in active interference resolution with affected parties.

In the U.S., Starlink’s DS2D beta operations have relied exclusively on the PCS G Block, a 2 × 5 MHz channel at 1910–1915 MHz (uplink) and 1990–1995 MHz (downlink), which is part of the IMT spectrum licensed to T-Mobile across the contiguous United States. 3 In September 2025, having concluded the beta phase, Starlink acquired Mobile Satellite Service (MSS) spectrum rights from Echostar in the AWS4 band (2000–2020 MHz uplink, 2180–2200 MHz downlink, 2×20 MHz) and IMT PCS H Block (1915–1920 MHz uplink, 1995–2000 MHz downlink, 2 × 5 MHz). These additional allocations significantly expand the spectrum available for its Direct-to-Cell operations, particularly since the AWS-4 band overlaps with the 5G NR NTN band n252.

自2019年首次发射以来，Starlink已部署了超过7000颗卫星，重塑了卫星通信行业。其最初采用的“弯管”（bent-pipe）架构[4]使得能够使用更简单的卫星设计快速部署，但代价是需要部署密集的地面站网络以确保连续覆盖。从2022年起，Starlink逐步引入了激光星间链路（ISL），并在其第二代（“Gen2”）卫星中全面采用，实现了星上信号处理和天基路由。这种从依赖地面的中继向天基网状网络（space-based mesh）的转变，带来了扩大的覆盖范围、更高的弹性和增强的端到端性能[1]。

Starlink没有等待3GPP完成NTN的标准化，而是通过将传统的LTE eNodeB有效载荷集成到其卫星中，并辅以针对NTN操作的专有适配，推行了早期部署策略。这使得智能手机能够使用地面移动标准进行本地连接，而流量则通过该公司的地面站网络处理，或在必要时通过ISL在星座中路由[5]。

SpaceX于2024年12月宣布完成了其首个DS2D星座的部署，该星座由大约400颗卫星组成——仅占其已部署卫星总数的6%。 除了通过与T-Mobile合作在美国提供SCS服务外，该星座还通过与运营商的合作在新西兰和澳大利亚提供SCS连接，并且目前正在与其它几个国家的运营商进行测试。

2024年11月，美国联邦通信委员会（FCC）部分批准了Starlink的DS2D运营，以启动仅限SMS的Beta测试，但施加了若干条件以防止对现有地面和卫星系统产生有害干扰，包括要求在收到任何经证实的干扰申诉后立即停止传输、限制带外发射（OOBE）总量的功率通量密度水平，以及与现有运营方强制执行协调程序。

随后，在2025年3月，尽管遭到了包括AT&T、Verizon和EchoStar在内的地面运营商的正式反对，FCC还是批准了一项有条件的豁免，允许Starlink将其OOBE提高9.4 dB。该豁免规定，Starlink必须继续遵守严格的OOBE限制，并与受影响方积极解决干扰问题。

在美国，Starlink的DS2D Beta运营完全依赖于PCS G频段，这是一个2 × 5 MHz的信道，上行链路为1910–1915 MHz，下行链路为1990–1995 MHz，该频段是T-Mobile在美国本土（contiguous United States）获得许可的IMT频谱的一部分。2025年9月，在结束Beta阶段后，Starlink从Echostar收购了AWS-4频段（上行2000–2020 MHz，下行2180–2200 MHz，2×20 MHz）的移动卫星服务（MSS）频谱权，以及IMT PCS H频段（上行1915–1920 MHz，下行1995–2000 MHz，2 × 5 MHz）。这些额外的频谱分配显著扩展了其“直连蜂窝”业务可用的频谱资源，特别是因为AWS-4频段与5G NR NTN的n252频段重叠。

C. Related Work

Mirroring the rapid pace of innovation of LEO satellite communication technologies, a growing body of research has emerged from multiple perspectives to explore the design, deployment, and operation of these systems. Early studies focused primarily on the technologies that underpin LEO broadband services, addressing alternative network architectures (e.g., bent-pipe versus ISL designs), hybrid beamforming strategies, interference coordination mechanisms, and resource management challenges [2]. Although much of this work was initially centered on fixed satellite broadband services, recent attention has shifted toward DS2D communications [6], which is reflected in dedicated research and publication initiatives [7]. Alongside theoretical and architectural advances, new simulation tools, such as Hypatia and StarPerf, have been developed to support realistic modeling, design, and performance evaluation of LEO satellite networks under operational conditions.

Following initial deployments and the establishment of operational networks, a growing number of measurement studies have emerged, primarily focusing on Starlink’s fixed satellite broadband services [1], [8]–[10]. These studies have primarily evaluated connectivity delivered through dedicated fixed user terminals (i.e., satellite “dishes”), with particular attention to metrics such as throughput and latency, a major historical limitation of traditional geostationary satellite systems.

Despite growing research on LEO satellite networks, to the best of our knowledge, there are no empirical evaluations of commercial DS2D deployments. This paper presents the first measurement-based analysis of a real-world DS2D network, using large-scale crowdsourced mobile data collected during the beta testing phase of Starlink’s service in the U.S.

与LEO卫星通信技术的快速创新步伐相呼应，学术界已从多个角度涌现出越来越多探索这些系统设计、部署和运营的研究。早期研究主要集中在支撑LEO宽带服务的技术上，涉及替代网络架构（例如，“弯管”与ISL设计）、混合波束成形策略、干扰协调机制和资源管理挑战[2]。尽管这其中许多工作最初都集中在固定卫星宽带服务上，但近期的注意力已转向DS2D通信[6]，这体现在专门的研究和出版计划中[7]。伴随着理论和架构的进步，新的仿真工具（如Hypatia和StarPerf）也已被开发出来，以支持在运行条件下对LEO卫星网络进行真实的建模、设计和性能评估。

随着初步部署和运营网络的建立，越来越多的测量研究开始出现，主要集中在Starlink的固定卫星宽带服务上[1]、[8]–[10]。这些研究主要评估了通过专用固定用户终端（即卫星“天线”）提供的连接性，特别关注吞吐量和延迟等指标，这是传统地球静止轨道卫星系统的主要历史局限。

尽管对LEO卫星网络的研究日益增多，但据我们所知，目前尚无对商用DS2D部署的实证评估。本文利用在Starlink美国服务Beta测试阶段收集的大规模众包移动数据，对真实的DS2D网络进行了首次基于测量的分析。

Methodology

In this study, we adopt a data-driven, user-centric methodology based on large-scale crowdsourced mobile measurements. The dataset is provided by the crowdsourcing company Weplan Analytics, whose data collection framework is specifically designed to capture and evaluate RAN performance. Crowdsourced measurements offer a unique vantage point for evaluating cellular network performance, capturing realworld user experience even in sparsely populated areas [11]. Subsection III-A describes the data sources and the data preprocessing; Subsection III-B presents how we estimate the potential share of SCS services; and Subsection III-C details how we calculate the expected performance of DS2D communications based on SINR measurements.

在本研究中，我们采用了一种基于大规模众包移动测量的数据驱动、以用户为中心的方法。该数据集由众包公司Weplan Analytics提供，其数据收集框架专门用于捕获和评估无线接入网（RAN）的性能。众包测量为评估蜂窝网络性能提供了一个独特的视角，即使在人口稀少的地区也能捕捉到真实的用户体验[11]。第三-A小节描述了数据源和数据预处理；第三-B小节介绍了我们如何估计SCS服务的潜在占比；第三-C小节详细说明了我们如何基于SINR测量值计算DS2D通信的预期性能。

A. Data sources

Our primary data source consists of LTE crowdsourced measurements passively collected from Android user devices between October 2024 and July 2025. All data were anonymized at collection time by the crowdsourcing provider to ensure full compliance with applicable data protection regulations. We used the following network variables: (i) RAN parameters RSRP, RSRQ, SINR, EARFCN, ECI (E-UTRAN Cell Identifier) as defined in 3GPP specifications [12], and (ii) PLMN (Public Land Mobile Network) identifier, comprising Mobile Country Code (MCC) and Mobile Network Code (MNC). We filter with MCC = 310 and MNC = 830 or 210 for Starlink measurements, and MNC = 260 for T-Mobile’s terrestrial network measurements, which we compare with Starlink’s.

In addition, we incorporate three complementary data sources that help assess the consistency and reliability of the crowdsourced dataset. First, we compiled a dataset tracking the evolution of the cumulative number of Starlink’s DS2Dcapable satellites in orbit by aggregating information from official SpaceX mission reports, which we publish openly as supplementary material. Second, we used the publicly available T-Mobile coverage map to evaluate how well the spatial distribution of DS2D measurements aligns with areas designated as lacking terrestrial coverage. Finally, we used official open-data vector files for U.S. counties and national parks to quantify the SCS share, as detailed next, and official county-level population data from the U.S. Census Bureau to study the correlation of SCS share and population density.

我们的主要数据源包含在2024年10月至2025年7月期间，从Android用户设备被动收集的LTE众包测量数据。所有数据在收集时均由众包提供商进行了匿名化处理，以确保完全符合适用的数据保护法规。我们使用了以下网络变量：（i）3GPP规范[12]中定义的RAN参数RSRP、RSRQ、SINR、EARFCN、ECI（E-UTRAN小区标识符），以及（ii）PLMN（公共陆地移动网络）标识符，包括移动国家码（MCC）和移动网络码（MNC）。我们使用MCC = 310和MNC = 830或210来筛选Starlink的测量数据，并使用MNC = 260来筛选T-Mobile的地面网络测量数据，以便与Starlink进行比较。

此外，我们还纳入了三个补充数据源，以帮助评估众包数据集的一致性和可靠性。首先，我们通过汇总SpaceX官方任务报告中的信息，编制了一个跟踪Starlink在轨具备DS2D能力卫星的累积数量演变的数据集，并将其作为补充材料公开发布。其次，我们使用了公开的T-Mobile覆盖图，以评估DS2D测量数据的空间分布与被指定为缺乏地面覆盖的区域的吻合程度。最后，我们使用了美国各县和国家公园的官方开放数据矢量文件来量化SCS占比（详见下文），并使用了来自美国人口普查局的官方县级人口数据来研究SCS占比与人口密度之间的相关性。

B. Measuring SCS share

To estimate the likelihood of relying on Starlink SCS services in a particular geographic area, we define a spatial indicator called SCS share. This metric quantifies the relative prevalence of Starlink DS2D connectivity in our dataset by computing the ratio of Starlink DS2D measurements to the total number of measurements (i.e., Starlink DS2D and T-Mobile terrestrial network measurements across all technologies and frequency bands) observed. It thus indicates the degree to which Starlink serves as the primary mobile coverage provider in a given area.

为了 估计在特定地理区域依赖Starlink SCS服务的可能性，我们定义了一个称为“SCS占比”（SCS share）的空间指标。

该指标通过计算Starlink DS2D测量值与观测到的测量总值（即Starlink DS2D和T-Mobile所有技术和频段的地面网络测量值）的比率，来量化Starlink DS2D连接在我们的数据集中的相对普遍程度。

因此， 它表明了Starlink在特定区域作为主要移动覆盖提供商的程度。

C. Estimating DS2D performance

To assess the performance of future DS2D data services, we use the modified Shannon formula for LTE introduced in [13] to estimate the downlink spectral efficiency of the system, η (bps/Hz per beam). This formula, for the case of no spatial multiplexing, is η = min { s a log 2 (1 + SINR/b), m } , where s is a bandwidth efficiency factor, which accounts for all system overhead including guard bands, pilot symbols and signaling channels; a, b are fitting coefficients, which model implementation-related losses; and m is a limit imposed by the highest modulation and coding scheme (MCS) that the system can use. The terms s and m can be computed theoretically, whereas a, b are adjusted from simulation. We consider a static (no fading) channel model with AWGN and no spatial multiplexing. In these conditions, [13] gives the values s = 0.57, a = 0.9, b = 1.25. The parameter m depends on the specific technology used by the LTE network. The highest supported MCS can be 64-, 256- or 1024-QAM, with a coding rate close to 1. As a reference, we use 256-QAM, for which the best channel-quality indicator (CQI) that the mobile terminal can report corresponds to 7.41 information bits per symbol [14, Table 7.2.3-2]. Taking into account the efficiency factor s, which already incorporates the effect of the cyclic prefix (as well as other types of overhead), this yields m = 4.22 bps/Hz. Thus, the expression for spectral efficiency results in: η = min { 0.51 log 2 (1 + SINR/1.25), 4.22 } . Following [13], the spectral efficiency is computed for each SINR value, and then averaged over the SINR distribution.

为了评估未来DS2D数据服务的性能，我们使用[13]中引入的LTE修正香农公式来估计系统的下行链路频谱效率 \(\eta\)（单位：bps/Hz 每波束）

在没有空间复用的情况下，该公式为：\(\eta = \min \{ s a \log_2(1 + \text{SINR}/b), m \}\)

其中: \(s\) 是一个带宽效率因子，它考虑了所有系统开销，包括保护带、导频符号和信令信道；\(a\) 和 \(b\) 是拟合系数，用于模拟与实现相关的损耗；\(m\) 是系统可使用的最高调制与编码方案（MCS）所施加的限制。\(s\) 和 \(m\) 可以从理论上计算得出，而 \(a\) 和 \(b\) 则通过仿真进行调整

我们考虑一个静态（无衰落）、无空间复用的加性高斯白噪声（AWGN）信道模型。在这些条件下，[13]给出的值为\(s = 0.57\)，\(a = 0.9\)，\(b = 1.25\)。参数\(m\)取决于LTE网络使用的具体技术。支持的最高MCS可以是64-、256-或1024-QAM，编码率接近1。作为参考，我们使用256-QAM，此时移动终端可报告的最佳信道质量指示符（CQI）对应于每符号7.41个信息比特[14, 表7.2.3-2]。考虑到已经包含了循环前缀（以及其他类型开销）影响的效率因子\(s\)，可得出\(m = 4.22 \text{ bps/Hz}\)。因此，频谱效率的表达式为：\(\eta = \min \{ 0.51 \log_2(1 + \text{SINR}/1.25), 4.22 \}\)。遵循[13]的方法，我们对每个SINR值计算频谱效率，然后对SINR的分布取平均值。

Results

AI 总结一下吧:

A. Network Expansion vs. Constellation Deployment

观测到的独特小区标识符（ECI）的数量变化与第二代（Gen2）卫星的部署时间表呈现密切相关性。这表明众包测量可以有效地作为大规模表征 DS2D 网络的工具，并且其洞察力集中在缺乏稳健地面网络覆盖的稀疏人口地区。

每月独特 ECI 数量的演变趋势，该趋势与在轨运行的卫星总数（Gen2 卫星发射时间线）保持紧密相关:

B. Starlink’s DS2D Spatial Analysis

DS2D 服务的覆盖范围和利用率存在显著的空间差异，主要集中在地面网络覆盖不足或无法服务的地区

1. 在星座完成之前，服务激活在地理上受到限制
2. 集中在 T-Mobile 指定为“仅卫星覆盖”的区域，并且集中在靠近人口稠密区但地面服务不足的、可达到的地区
3. SCS 份额与县级人口密度之间存在中等强度的负对数相关性

C. Analysis of DS2D Key Performance Indicators

与地面网络相比，DS2D 网络的物理层 RAN 指标反映出高传播损耗，但干扰较小，这符合其在低负载下提供覆盖受限服务的特性

D. Performance assessment of future DS2D services

基于现有 SINR 测量，当前的 DS2D 服务容量足以满足基本的短信需求，但远低于地面网络；未来通过提高发射功率、降低轨道高度和扩展带宽，容量有显著提升的潜力

Conclusion

In this article, we presented the first empirical analysis of a commercially deployed DS2D network, focusing on Starlink’s SCS Service in the U.S. in partnership with T-Mobile. Using large-scale crowdsourced measurements between October 2024 and July 2025, we characterized the spatio-temporal evolution and RAN performance of the DS2D network, which only supported SMS during the observed period.

Our results confirm a strong correlation between the number of DS2D-capable satellites launched and the number of unique cell identifiers observed, as well as a spatial distribution of measurements consistent with the operator’s coverage gaps. We observe a significant number of measurements concentrated around areas that are relatively accessible yet underserved by terrestrial networks, such as national parks. Furthermore, we find significant negative correlation between the prevalence of Starlink measurements and population density, reflecting its role as a complementary solution in sparsely populated regions.

The measurements indicate slight increases in RSRP, RSRQ, and SINR throughout the beta testing period, which may be attributable to network changes, including the higher OOBE levels authorized by the FCC and the deployment of some satellites at VLEO altitudes. Based on the SINR measurements, we estimate current average per-beam throughput to be approximately 3.1 Mbps, which is sufficient for basic services but significantly below terrestrial averages. Looking ahead, per-beam throughput could increase up to 4.2 Mbps by using VLEO satellites (around 350 km), and up to 4.5 Mbps for LEO, assuming that OOBEs relaxation has not been applied yet. In the short term, recent spectrum acquisitions will enable doubling per-beam capacity (to 6.2 Mbps), with the potential to reach 18.6 Mbps in the medium term once NTN bands are integrated into commercial smartphone modems. However, the extent to which these enhancements translate into higher peruser throughput will ultimately depend on the expansion of the satellite constellation and the evolution of service demand.

Our work illustrates the value of crowdsourced data to monitor emerging network technologies. Beyond characterizing deployment and performance, such data can help detect underserved zones, assess coverage gaps, and inform infrastructure planning, offering a promising tool for both technical and policy research on mobile communication systems. Future work will involve monitoring the evolution of DS2D networks and services, tracking the evolution of Starlink’s network as new services are introduced, and exploring other emerging DS2D deployments.

在本文中，我们对一个已商用部署的DS2D网络进行了首次实证分析，重点关注Starlink在美国与T-Mobile合作推出的SCS服务。利用2024年10月至2025年7月期间的大规模众包测量数据，我们刻画了该DS2D网络的时空演进和RAN性能特性，该网络在观测期内仅支持短信服务。

我们的结果证实，具备DS2D能力的卫星发射数量与观测到的唯一小区标识符数量之间存在强相关性，并且测量数据的空间分布与运营商的覆盖空白区域相一致。我们观察到大量测量数据集中在那些相对可达但地面网络服务不足的区域，例如国家公园。此外，我们发现Starlink测量的普遍性与人口密度之间存在显著的负相关关系，这反映了其作为人口稀少地区补充解决方案的角色。

测量数据表明，在整个Beta测试期间，RSRP、RSRQ和SINR均有轻微增长，这可能归因于网络的变化，包括FCC批准的更高OOBE（带外发射）水平以及部分卫星在VLEO（极低地球轨道）高度的部署。基于SINR测量值，我们估计当前平均每波束吞吐量约为3.1 Mbps，这足以支持基本服务，但显著低于地面网络的平均水平。展望未来，通过使用VLEO卫星（约350公里），每波束吞吐量可增至4.2 Mbps；假设OOBE放宽尚未实施，LEO（卫星）的（吞吐量）最高可达4.5 Mbps。短期内，近期的频谱收购将使每波束容量翻倍（达到6.2 Mbps），而一旦NTN频段被集成到商用智能手机调制解调器中，中期则有潜力达到18.6 Mbps。然而，这些增强在多大程度上能转化为更高的每用户吞吐量，最终将取决于卫星星座的扩展和服务需求的演变。

我们的工作展示了众包数据在监测新兴网络技术方面的价值。除了描述部署情况和性能特性外，此类数据还有助于检测服务不足的区域、评估覆盖差距，并为基础设施规划提供参考，从而为移动通信系统的技术和政策研究提供了一个有前景的工具。未来的工作将包括监测DS2D网络和服务的演进，跟踪Starlink网络在引入新服务时的演变，以及探索其他新兴的DS2D部署。