MAC in NTN — Non-Terrestrial Networks

Satellite & HAPS MAC: TA pre-compensation, HARQ adaptations, extended RACH & scheduling offsets

← MAC Overview TA → HARQ → RACH → DRX → DC → Sidelink → Energy-Saving MAC AI Scheduling ISAC Scheduling IAB MAC Cell-Free MAC

1. Propagation Delay — The NTN Challenge

NTN platforms span from HAPS (20 km) to GEO (35,786 km), creating RTTs from <1 ms to 600+ ms — compared to 1-8 ms for terrestrial. This fundamentally impacts HARQ, TA, scheduling, and RACH.

🛰️ NTN Propagation Delay — Packet Flight Time

Select a platform and watch packets travel. See how RTT varies from sub-ms (HAPS) to 600ms (GEO).

PlatformAltitudeOne-Way DelayRTTCell Radius
TerrestrialGround0.5-4 ms1-8 ms0.1-10 km
HAPS20-50 km0.07-0.17 ms0.3-1 ms50-200 km
LEO500-2000 km1.7-6.7 ms20-40 ms50-300 km
MEO8,000-20,000 km27-67 ms80-150 ms100-500 km
GEO35,786 km119 ms500-600 ms200-1000 km

2. Timing Advance — Pre-Compensation

NTN TA = Common TA (K_offset) + UE-Specific TA. UE pre-computes TA using GNSS position and satellite ephemeris (from SIB19) before first UL transmission.

📐 TA Pre-Compensation — Satellite Position & UE Distance

See how TA is computed from satellite altitude and UE position. Adjust altitude and UE offset.

550 100
Key point: Without pre-compensation, the first PRACH from a GEO UE would arrive ~600 ms late at the gNB. TA pre-compensation uses GNSS + ephemeris to compute the correct TA before any UL transmission.

3. HARQ Adaptations

NTN HARQ RTT ranges from 30 ms (LEO) to 600+ ms (GEO) — far exceeding terrestrial 8 ms. Options: extend processes (up to 32), disable HARQ (use RLC ARQ), or use blind retransmissions.

⚡ HARQ Timeline — Terrestrial vs LEO vs GEO

Compare HARQ round-trip times. See how GEO needs HARQ disabled or blind retransmissions.

Extended HARQ

Up to 32 processes. Keeps pipeline full despite 30-50 ms RTT (LEO). Still limited for GEO.

HARQ Disabled

For GEO: RTT too long for meaningful retx. RLC ARQ handles reliability (slower, but works).

Blind Retransmissions

Send N copies without waiting for ACK. Reduces latency vs stop-and-wait. Wastes resources on success.

4. Scheduling & Random Access

Scheduling offsets (K_offset) are added to K0/K1/K2 to compensate propagation delay. RACH RAR window extended up to 480 slots. Total RACH time: 50-100 ms (LEO) to 1-3 seconds (GEO).

ParameterTerrestrialLEOGEO
K_offset (slots)020-40280-560
RAR window (slots)1040-80240-480
RACH total time10-20 ms50-100 ms1-3 sec
HARQ processes1632Disabled
TA pre-compNoYes (GNSS)Yes (GNSS)
LEO handover frequency: ~1 per 2-10 minutes (vs 1 per 30-60 sec in urban terrestrial). Less frequent but more complex. Satellite ephemeris enables predictive handover.

5. Practical Field Considerations

ScenarioPlatformKey MAC Config
Rural broadband (fixed CPE)GEOHARQ disabled, large K_offset, eDRX
Maritime / AviationLEO constellationExtended HARQ (32), frequent TA update
IoT tracking (battery)LEO (IoT-NTN)eDRX (minutes), blind retx
Emergency / disasterHAPSNear-terrestrial params, standard HARQ
Hybrid NTN+TNLEO + terrestrialNTN as SCG, TN as MCG anchor

LTE vs 5G NR NTN

AspectLTE5G NR NTN
Satellite supportNone (proprietary)Standardized (Rel-17+)
TA pre-compNoYes (GNSS + ephemeris)
HARQ adaptationFixed 8 procUp to 32, disable option
Regenerative payloadN/ARel-18 (gNB on sat)
Common pitfall: Using terrestrial HARQ process count (16) for LEO → pipeline stalls. Using short RAR window → Msg2 missed → repeated RACH → congestion.

6. 6G Evolution — 3D Non-Terrestrial Integration

Mega-Constellation

10,000+ LEO satellites with THz ISL. Multi-hop space routing. MAC coordinates satellite mesh.

AI Orbit-Aware MAC

AI predicts satellite positions → pre-schedules resources → predictive HO with no measurement gaps.

Unified 3D Network

Single MAC entity: terrestrial + HAPS + LEO + GEO. Dynamic path selection per packet.

🚀 NTN Evolution — 4G to 6G

Compare NTN capabilities across generations.

Dimension4G (LTE)5G NR NTN6G (Target)
SatelliteNoneLEO + GEOMega-constellation
PayloadN/ATransparent + RegenFull gNB + ISL
HARQNoneExtended / DisabledAI-predictive
MobilityN/AEphemeris-basedAI-predicted HO
CoverageGround only2D + space3D ubiquitous