Ferkans — Interactive Telecom Tutor

The Constellation as a Cell-Free Network

A LEO constellation above a ground UE is, in effect, a sparse cell-free network: multiple satellites simultaneously visible, each within link-budget range. Rather than picking one "serving satellite" and handing off at horizon-exit, the UE can receive from several satellites in parallel. This is multi-satellite macro-diversity: the space-domain analog of cell-free OTFS (Chapter 17). The Buzzi-Caire-Colavolpe CommIT framework extends cell-free to this orbital setting. With OTFS handling the extreme Doppler per-satellite and macro-diversity smoothing across satellites, the resulting link is robust against blockage, handover, and high mobility at once.

,

Definition:
Multi-Satellite Macro-Diversity

At time $t$ , the UE has $S(t)$ simultaneously-visible satellites (above minimum elevation $\theta_{\min}$ ). The per-satellite DD channel is $\mathbf{H}^{(s, k)}(t) \;=\; a_s(t)\, \delta(\tau - \tau_s(t))\, \delta(\nu - \nu_s(t)) \,+\, \text{multipath}, \qquad s = 1, \ldots, S.$ Each satellite contributes one dominant path (LOS) plus occasional multipath. The channel to UE from all satellites is the aggregate of $S$ such per-satellite channels.

Receive processing: the UE combines signals from all visible satellites via cell-free OTFS principles (Chapter 17): conjugate beamforming + coherent combining. Diversity gain: $S$ -fold.

Coordination: satellites coordinate timing/phase with ground stations via ephemeris predictions + cross-sat links.

Typical $S$ : 3-5 simultaneously visible over most of the sky for Starlink-class constellations; up to 10 over equatorial regions.

Theorem: LEO Multi-Satellite Diversity Gain

For a UE with $S$ visible LEO satellites and per-satellite SINR $\mathrm{SINR}_s$ , the aggregate SINR after multi-satellite macro-diversity combining is $\mathrm{SINR}_{\text{agg}} \;=\; \sum_{s=1}^{S} \mathrm{SINR}_s \;\approx\; S \cdot \overline{\mathrm{SINR}}.$ The reliability benefit (against blockage / weather fade): $\mathbb{P}(\mathrm{SINR}_{\text{agg}} < \gamma) \;\leq\; \prod_{s} \mathbb{P}(\mathrm{SINR}_s < \gamma/S),$ which vanishes rapidly as $S$ grows.

Numerical: at $S = 4$ , per-satellite $\mathrm{SINR} = 15$ dB with 10% outage probability:

Single-satellite outage: 10%.
Multi-satellite aggregate outage: $\sim 0.01\%$ (100 $\times$ better).

The link availability improves from "good" (90%) to "near-perfect" (99.99%) — the key operational advantage of LEO constellations over GEO.

Each satellite link has its own reliability profile: blockage by buildings, fading from weather, satellite outage. If one satellite is unavailable, others typically are. Multi-satellite diversity exploits this statistical independence — if $S$ satellites each have 90% availability, the probability that all $S$ are simultaneously unavailable is $\sim (10^{-1})^S$ . For $S = 4$ : $10^{-4}$ , a 1000x improvement.

Proof

Sum SINR

Coherent combining at the receiver: signal powers add constructively ( $\sum_s |a_s|^2 P_s$ ), noise adds linearly ( $S \sigma_w^2$ ). SINR ratio: linearly $S \cdot \overline{P_s/\sigma_w^2}$ .

Outage

Link outage if aggregate below threshold. Independent-satellite assumption: $\mathbb{P}(\text{all below}) = \prod_s \mathbb{P}(\text{s below})$ .

Rapid decay

For $S$ -fold redundancy at individual outage $p$ : aggregate outage $p^S$ . $p = 0.1$ , $S = 4$ : $10^{-4}$ . $\blacksquare$

Key Takeaway

Multi-satellite macro-diversity delivers 99.99%-reliability LEO links. Combining $S = 4$ simultaneously-visible satellites with per-sat 90% availability: aggregate availability 99.99%. Compare GEO single-satellite: 99.5% typical. LEO with macro- diversity decisively beats GEO — the structural advantage of constellations.

Definition:
Soft Handover in LEO Constellations

In classical satellite networks, handover is hard: UE drops old satellite, acquires new one. Brief service interruption (seconds) during acquisition.

Soft handover (cell-free LEO): UE maintains links with multiple satellites continuously. As one satellite sets, its contribution naturally diminishes; as a new one rises, its contribution grows. The transition is smooth — no service interruption.

Implementation: cell-free OTFS conjugate beamforming. Ground station coordinates which satellites serve which UEs. As satellites enter/exit visibility, clusters update automatically.

Latency benefit: no acquisition time between passes. Critical for mobile UEs (ships, aircraft) that may fail to acquire a new satellite during a hard handover (signal strength drops during manoever).

Theorem: Continuous Rate Under Handover

For a UE under continuous LEO service with handover between satellites, the sustained rate is $R_{\text{sust}} \;=\; (1 - \eta_{\text{ho}}) \cdot R_{\text{peak}},$ where $\eta_{\text{ho}}$ is the handover overhead fraction.

Hard handover: $\eta_{\text{ho}} = T_{\text{acquisition}}/T_{\text{pass}} \sim 1\%$ per handover. At 5-minute pass duration: 1 handover every 5 min. $\eta_{\text{agg}} \sim 1\%$ .

Soft handover (cell-free): $\eta_{\text{ho}} = 0$ . Continuous rate = peak rate. No rate loss from handover.

Consequence: Soft handover provides $\sim 1\%$ rate improvement on top of the other cell-free gains. For URLLC applications (no tolerance for interruption): soft handover is mandatory.

This is a small quantitative win over the macro-diversity gain, but qualitatively important for latency-critical services. A ship at sea, a UAV, an emergency vehicle — all need continuous service without the ~0.1-second acquisition delay of hard handover. Soft handover via cell-free OTFS LEO is the only way to meet URLLC targets in non-terrestrial networks.

Proof

Hard handover overhead

Service interruption during acquisition: $\sim 0.1$ - $1$ seconds. Number of handovers per minute: $1/T_{\text{pass}}$ $\sim$ 12/hour. Per-hour overhead: 12 × 1 sec = 12 seconds / hour = 0.3%. Measurable rate loss.

Soft handover

Multiple satellites always contribute. No dedicated acquisition time. Overhead: 0.

Net gain

Soft handover adds $\sim 0.3\%$ rate improvement. Combined with macro-diversity's reliability gain: very robust link. $\blacksquare$

Example: Starlink-Scale Multi-Satellite Coverage

Starlink constellation: $\sim 6000$ satellites at $\sim 550$ km altitude. Compute per-UE: (a) Number of simultaneously-visible satellites $S$ . (b) Multi-sat macro-diversity gain. (c) Outage probability vs GEO baseline.

Solution

Visible satellites

Per-site area: $\sim 4\pi R^2 / 6000 \sim 85{,}000$ km² per satellite. Visible slant area at 10° min elevation: $\sim 5000$ km² footprint. Ratio: ~6-10 satellites simultaneously visible on average ( $S \approx 5$ ).

Diversity gain

$S = 5$ . Per-sat SINR at 15 dB with 5% outage. Aggregate SINR: $5 \cdot 15 = 75$ dB (22 dB in actual SNR terms, i.e., cell-free combining gives ~7 dB aggregate). Aggregate outage: $(0.05)^5 = 3 \times 10^{-7}$ — essentially never outage.

GEO comparison

GEO at 35{,}786 km: single satellite, 99.5% availability (rain fades at 28 GHz). At 10 Mbps: periodic 5-minute outages. LEO cell-free OTFS: ~99.9999% availability.

Rate

GEO peak: 50 Mbps per UE, average 45 Mbps (10% outage). LEO cell-free: 200 Mbps per UE, average 199.9 Mbps. 4x the rate, 200x the reliability.

LEO Multi-Satellite Diversity Gain

Plot aggregate SINR and outage probability as a function of $S$ (simultaneously-visible satellites). Compare single-sat, 3-sat, 5-sat, 10-sat combining.

Parameters

S

visible4

Per-sat SNR (dB)15

Outage threshold (dB)5

Multi-Satellite Combining for LEO-OTFS

Input: S simultaneously-visible satellites {s_1, ..., s_S}

Per-satellite DD channel estimates {ĥ^(s, k)}

Per-satellite received signal {y^(s)}

Output: Combined estimate ẑ

1. ESTIMATE PER-SAT CHANNEL:

For each satellite s:

ĥ^(s, k) = channel estimate from pilot or sensing

2. PHASE ALIGNMENT:

Ensure all satellites contribute coherent phase. Use time-sync

from GPS + predicted ephemeris phase.

3. CONJUGATE BEAMFORMING:

For each DD cell (ℓ, k):

z^(s)[ℓ, k] = (ĥ^(s, k)[ℓ, k])^* / ||ĥ^(s, k)|| · y^(s)[ℓ, k]

4. COHERENT SUM:

ẑ[ℓ, k] = (1/S) Σ_s z^(s)[ℓ, k]

5. DETECT:

Apply DD-domain MP detection on ẑ.

6. HANDOVER MANAGEMENT:

Monitor per-sat quality. Add entering satellites to cluster;

remove setting satellites. Transition continuously.

Complexity: O(S · MN · P) for combining. S = 5, MN = 10⁵, P = 3:

1.5 × 10⁶ ops per frame. Trivial on modern SoC.

🚨Critical Engineering Note

Synchronization Across the Constellation

For multi-satellite coherent combining, satellites must be time- and phase-synchronized:

Time sync across satellites:

Inter-satellite laser links (ISLs) exchange timing data.
Ground-to-satellite PTP: from gateway stations.
Accuracy: $\pm 10$ ns typical. Sufficient for 10 ns $\Delta\tau$ at 100 MHz bandwidth.

Phase sync:

Each satellite has GPS-disciplined oscillator. Reference to GNSS.
Inter-satellite phase error: $\pm 0.1°$ achievable.
Phase error from ephemeris uncertainty (position): $\sim 5$ - $10°$ at 28 GHz. Must be calibrated.

Calibration protocols:

Ground station transmits reference signal; multiple satellites receive it.
Each satellite reports received phase.
Calibration solves for inter-satellite phase errors.
Update every minute during active operation.

Deployed in Starlink Gen 2 (2023+). Calibration overhead: ~0.1% of satellite capacity.

Practical Constraints

•
Time sync: ±10 ns via ISL + GPS
•
Phase sync: ±0.1° with GPS-disciplined oscillator
•
Ephemeris phase correction: ±5° calibrated
•
Calibration overhead: ~0.1% of capacity

Common Mistake: Handover Race Conditions

Mistake:

Assuming soft handover is universally better than hard handover. For UEs transitioning between orbital planes (e.g., polar vs equatorial), the satellite set changes dramatically — all cluster members leave, new ones arrive. A pure soft handover approach loses continuity.

Correction:

Use hybrid handover:

Intra-plane (satellites of same orbital plane): soft handover. Gradual.
Inter-plane (satellites from different orbital plane): hard handover. Brief interruption, but unavoidable geometry.

For critical applications, schedule inter-plane handovers during non-active periods (UEs asleep, etc.). For continuous 24/7 service: ~1-2 brief interruptions per hour. Tolerable for most applications. Starlink design example: 3 orbital planes, inter-plane handover every 20 minutes.

Multi-Satellite Macro-Diversity