Integration of Satellites in 5G through LEO Constellations

I. INTRODUCTION

• Satellite integration is presented as a response to 5G demands for greater throughput and global, seamless connectivity, with satellites complementing terrestrial networks across dense and rural areas.The paper notes that past independent evolution of terrestrial and satellite networks made later integration difficult.Source: §I · p. 1
• Prior work considered GEO-related challenges and LEO satellite effects on LTE; this paper shifts the assessment to 5G PHY/MAC procedures.The paper also situates hybrid satellite–terrestrial resource management in NFV/SDN work.Source: §I · p. 1
• The study assesses how large delays and Doppler shifts in a Ku-band LEO mega-constellation affect 5G waveform design, Random Access, and HARQ procedures.It extends the analysis in to the emerging 5G air interface.Source: §I · p. 1

II. SYSTEM ARCHITECTURE

• Type-1 Relay Nodes provide terrestrial access and backhaul through transparent satellites to a donor gNB connected to the 5G Core Network.Relay Nodes act as gNBs toward UEs and as UEs toward the donor gNB.Source: §II · pp. 1–2
• The assumed backhaul is FDD and outband, while the UE–RN and RN–DgNB interfaces may use different numerologies.The satellite gateway has ideal feeder links, and one donor gNB may connect to multiple Relay Nodes.Source: §II · p. 2

III. SATELLITE CHANNEL

• The channel section focuses on how large round-trip time and Doppler shifts affect 5G MAC and PHY procedures.Figure 1 identifies the g-Sat-RN architecture used as the system context.Source: §III · p. 2, Figure 1 · p. 2

A. Delay

• The RTT estimate accounts for the RN–satellite and satellite–donor-gNB distances, approximating each as the satellite altitude at the closest elevation geometry.The minimum RN–satellite distance occurs at a 90-degree elevation angle and equals the 1200 km altitude.Source: A. Delay · p. 2
• The satellite RTT is described as quite large compared with the maximum RTT foreseen in terrestrial networks, making integration with 5G a critical challenge.The section identifies large RTT and Doppler shifts as the satellite impairments under study.Source: A. Delay · p. 2, §III · p. 2

B. Doppler

• The paper treats UE–RN Doppler from relative motion separately from satellite-channel Doppler, which arises from satellite movement.For the UE–RN case, the maximum Doppler expression uses UE–RN relative speed, carrier frequency, and the angle between velocity vectors.Source: B. Doppler · p. 2
• At 4 GHz, 500 km/h UE mobility corresponds to a Doppler spread up to 1.9 kHz, which NR is described as able to estimate and decode.The 500 km/h target applies below 6 GHz.Source: B. Doppler · p. 2
• 11 GHz < fd < 14 GHz is the reported maximum Doppler-shift range for the considered Ku-band satellite scenario, significantly above terrestrial-link Doppler.The shift is attributed to satellite movement while the Relay Node and donor gNB are fixed on Earth.Source: B. Doppler · p. 2

IV. CHALLANGES OF 5G AIR INTERFACE (NR) IN SATELLITE CONTEXT

• The section assesses how large satellite-link RTT and Doppler shifts affect NR MAC and PHY procedures, against ongoing 5G standardization work.It frames the goal as identifying challenges for making NR “satellite friendly.”Source: §IV · pp. 2–3

A. Waveform

• NR's scalable subcarrier spacing and filtering/windowing support different numerologies; larger spacing improves Doppler robustness, with 15 kHz × 2^n defining the spacing values.For example, the passage gives 30 kHz spacing as sufficient for 1.9 kHz Doppler at a 4-GHz carrier frequency.Source: A. Waveform · p. 3
• Even at 480-kHz subcarrier spacing, NR tolerates only a 30.4-kHz frequency offset, leaving the satellite-link Doppler challenge unresolved on the RN–DgNB interface.The UE–RN interface does not need modification for Doppler in this scenario because RN compensates for UE-movement Doppler.Source: A. Waveform · p. 3

B. Random Access

• NR random access supports both contention-based and contention-free procedures, with contention-based access following LTE's four steps and assigning the final CRNTI in steps 3 and 4.The RAR window and contention-timer sizes had not yet been agreed in NR.Source: B. Random Access · p. 3

1) UE random access procedure:

• UE–RN contention-based random access is expected to work without modification: the NR contention timer must exceed 16 ms, and LTE allows up to 64 ms.Contention-free access uses only steps 1 and 2 and does not involve the satellite link.Source: 1) UE random access procedure · p. 3

2) RN attach procedure:

• RN startup has two phases: UE-like access followed by DgNB relay configuration; satellite RTT requires a contention timer above 16 ms and a larger RAR window than LTE's 3–15 ms.The passage notes that lower RAR-window values were expected for NR.Source: 2) RN attach procedure · p. 3
• At 45° elevation, the 320-km beam geometry yields a 208-km RN path-length difference—almost twice NR cell size and greater than LTE's 100-km largest-preamble coverage.The passage gives d1 = 1580 km and d2 = 1372 km for the longest and shortest satellite paths.Source: 2) RN attach procedure · p. 4
• RNs can be added one by one at startup, but handover or heavy rain may desynchronize them, forcing simultaneous reattachment and channel contention.The passage distinguishes this reattachment case from initial deployment.Source: 2) RN attach procedure · p. 4

C. Satellite Handover

• Satellite motion requires relay nodes to hand over, and repeated random access can consume time and reduce throughput.The proposed alternative is identified as a possible solution, but its details are not included here.Source: C. Satellite Handover · p. 4

D. Hybrid Automatic Repetition Quest (HARQ)

• A 16 ms satellite-link RTT makes propagation delay a major part of HARQ timing, unlike in the terrestrial link described.The discussion considers 8 ms total processing time and a 1 ms TTI.Source: D. Hybrid Automatic Repetition Quest… · p. 4
• NR is expected to support multiple HARQ configurations per UE, whereas FDD LTE uses up to eight parallel processes in each direction.The paper presents this configuration flexibility as useful when considering HARQ over a satellite channel.Source: D. Hybrid Automatic Repetition Quest… · p. 4

• NR should support 24 parallel HARQ processes to offer the desired peak data rate over a LEO satellite link.Source: N SAT · p. 5
• More parallel HARQ processes increase the UE soft-buffer requirement, which the paper says can be costly.The stated relationship is Nbuffer ∝ NHARQ,max * TTI.Source: N SAT · p. 5
• More processes also require wider DCI fields, increasing downlink control overhead.The passage contrasts this with LTE's three-bit DCI field for eight processes.Source: N SAT · p. 5

A. Waveform

• GNSS-equipped relay nodes estimate satellite position to compensate Doppler without changing the 5G waveform.Position-estimation error leaves a residual Doppler shift.Source: A. Waveform · p. 5
• The residual-Doppler analysis relates an estimated satellite distance and elevation angle to the true pair, accounting for position-estimation error.The analysis solves the geometric relations to obtain residual Doppler as a function of that error.Source: A. Waveform · p. 5
• Higher subcarrier spacing tolerates larger position-estimation errors, and the paper prefers high SCS on the RN-DgNB interface.At fixed estimation error, the highest Doppler occurs at 90 degrees elevation.Source: Figure 5 · p. 5, A. Waveform · p. 5

B. RA Procedure

• The RAR window size is a design consideration for RN-DgNB random access.Source: B. RA Procedure · p. 5
• A fixed RAR window above 16 ms covers the largest satellite delay but imposes unnecessary delay on terrestrial links.Source: 1) RAR window size · p. 5
• A satellite-aware option signals link presence to relay nodes through the downlink control channel so they can adjust the RAR window.The proposal requires one additional bit to indicate whether a satellite channel is present.Source: B. RA Procedure · p. 5

2) Taming advance (TA):

• Larger preamble formats can accommodate greater timing advance, though their design needs detailed analysis.The preamble format also indicates the maximum supported cell size.Source: 2) Taming advance (TA) · p. 5
• GNSS-equipped RNs can estimate satellite distance and timing advance before sending a preamble, avoiding a changed preamble format.An estimation error up to 100 km can be supported while keeping the same preamble format; the RAR then provides the exact timing advance.Source: 2) Taming advance (TA) · p. 5

C. HARQ Retransmissions

• Four solutions are identified to address the potentially large number of HARQ retransmissions.Source: C. HARQ Retransmissions · p. 5

1) Increased number of HARQ processes and buffer size:

• At least 24 parallel HARQ processes are needed for continuous transmission and peak data rate, but UE memory capability for the increased buffer requires further analysis.Source: 1) Increased number of HARQ processes… · p. 6

2) Increased number of HARQ processes with reduced buffer size:

• Two-bit feedback can indicate how close a received TB is to the original, reducing retransmissions and directly affecting buffer memory size.The feedback lets the transmitter add redundant bits according to the reported information.Source: 2) Increased number of HARQ processes… · p. 6

3) Reduced number of HARQ processes with reduced buffer size:

• Limiting the number of HARQ processes can reduce buffer requirements when the first two solutions are not possible due to UE capability, but transmission throughput will fall.Source: 3) Reduced number of HARQ processes… · p. 6

4) No HARQ protocol:

• When large satellite-channel delay prevents HARQ, replicating each TB before transmission can reduce delay impact, but significantly reduces throughput.Source: 4) No HARQ protocol · p. 6

VI. CONCLUSIONS

• The conclusion states that at least 24 parallel HARQ processes are needed for peak data rate, creating UE-buffer and DCI-field challenges that proposed solutions seek to control.Source: §VI · p. 6