Chapter Summary

Chapter Summary

Key Points

• 1.

  LEO is the interesting NTN regime. Delays of $2$–$20$ ms are compatible with interactive services, but orbital velocities of $\approx 7.5$ km/s create peak Doppler shifts of $\approx 700$ kHz at Ka band and force continuous handover every $5$–$15$ minutes per satellite pass. MEO and GEO avoid the Doppler and handover problems but cannot meet the latency and throughput bar of modern broadband. The rest of the chapter therefore concentrates on LEO.

• 2.

  The LEO channel is LOS-dominated with deterministic Doppler. Rician K factors of $10$–$20$ dB make the channel essentially a single specular path with a known complex gain. The large raw Doppler shift is deterministic — it can be computed from the satellite ephemeris and pre-compensated on both uplink and downlink. The residual Doppler after pre-compensation is $O(100)$ Hz, within the tolerance of 5G NR FR2 OFDM numerologies.

• 3.

  Ka band wins on bandwidth, not on Doppler. The $2$–$3$ GHz contiguous spectrum available at Ka band overrides the $14\times$ Doppler penalty compared to S band. The design burden is therefore to handle Ka-band Doppler via pre-compensation and to handle Ka-band rain fade via link margin and macro-diversity.

• 4.

  Macro-diversity across $M$ visible satellites gives linear SNR gain. Theorem TMacro-Diversity SNR Gain with Coherent Combining establishes that coherent joint transmission with matched precoding yields a post-combining SNR proportional to $M$, while selection- diversity only yields a logarithmic $\\log M$ improvement. The $\\approx 6$ dB gain at $M = 4$ is large and is the core technical claim of the Buzzi-Caire-Colavolpe contribution.

• 5.

  Cell-free user-centric architecture is the right system abstraction. At any instant a typical LEO terminal has $M \approx 5$–$20$ simultaneously visible satellites. Serving it from a cluster of $M^\star \approx 3$–$6$ of them yields the linear SNR gain, an order-of-magnitude reliability improvement under rain fade, and a soft-handover mechanism that eliminates the service interruptions of single-satellite schemes. The user-centric formulation of Chapter 12 ports directly to orbit.

• 6.

  OFDM works only with pre-compensation and narrow beams. Raw Doppler and wide-beam differential delay both exceed the budgets of 5G NR FR2 OFDM. Real deployments (Starlink, OneWeb) handle this with aggressive ephemeris-driven pre-compensation and per-spot-beam carrier assignment, keeping the residual ICI and ISI in check. This is the practical deployment path through Release 17–18 NTN.

• 7.

  OTFS localizes the LEO channel to a single delay-Doppler tap. For a pure LOS channel with Doppler shift $\\Delta f_D$ and delay $\\tau_0$, OTFS represents the effect as a single shift on the delay-Doppler lattice. Equalization is then trivial and the Doppler tolerance is set by lattice resolution rather than by subcarrier spacing. The ICI floor that degrades OFDM at high residual Doppler disappears. The cost is that OTFS is not yet in any 3GPP standard.

• 8.

  Feeder-link load is the practical bottleneck. Coherent macro-diversity requires distributing each user's data to all $M$ serving satellites, which multiplies the feeder-link load by $M$. Optical inter-satellite links, now being deployed at Starlink-scale, let the master satellite distribute data within the constellation rather than through multiple ground gateways. The ISL-routed architecture is the default assumption for 6G NTN research proposals.

• 9.

  Handover becomes cluster reselection. Instead of the Starlink-style hard handover every few minutes, the user-centric cell-free architecture reshuffles its cluster members smoothly: a departing satellite's contribution fades down while a new satellite's contribution fades up, with an overlap window of one round-trip time. Service is uninterrupted, and the master-satellite role rotates infrequently.