R1-2601811 discussion

Discussion on downlink transmission scheme(s) for downlink shared channels for 6GR

From Spreadtrum
Status: not treated
WI: FS_6G_Radio
Agenda: 10.5.2.2
Release: Rel-20
Source: 3gpp.org ↗
Spreadtrum's prior position on 10.5.2.2 at RAN1#124 · AI-synthesized, paraphrased
verify sources →
Proposes unified 6G downlink transmission schemes to minimize UE and NW implementation complexity, explicitly opposing multiple schemes for the same scenario and scheme enhancements yielding less than 10% performance gain. Retains NR baselines for maximum layers (8), maximum codewords (2), DMRS port count, and PN sequence for CP-OFDM, while requiring a single DMRS type for both DL and UL. For AI/ML-based DMRS overhead reduction, presents simulation results showing AI-based channel estimation achieving similar BLER with reduced DMRS overhead (2-symbol DMRS matching 3-symbol non-AI performance in time domain, 1/3 RE reduction in frequency domain) and proposes prioritizing sparse orthogonal DMRS (sub-case A) over superimposed pilots (SIP), which they consider insufficiently evaluated for multi-user interference and standardization impact. For NR-6G multi-RAT spectrum sharing, proposes semi-static FDM/TDM and rate matching mechanisms while opposing SDM between NR and 6G due to negative performance impacts.

Summary

Spreadtrum presents a comprehensive analysis of downlink transmission schemes for 6G, containing 21 proposals and 5 observations covering PDSCH design, DMRS, PTRS, and MRSS aspects. The document argues for simplifying 6G designs relative to 5G NR by unifying schemes, minimizing configurations, and avoiding marginal enhancements.

Position

Spreadtrum proposes maintaining 5G NR layer limits (max 8 layers, 2 codewords) and a baseline layer mapping rule where the first half layers map to CW0 and remaining to CW1, with a single MCS per codeword. They require a single PDSCH mapping type to reduce receiver complexity and oppose multiple transmission schemes for the same scenario, arguing that only Rel-15 schemes achieved good commercialization while later enhanced schemes were not implemented. For DMRS, they propose defining a single DMRS type for DL and supporting up to 32 ports via increased CDM groups or longer FD-OCC, and present technical case against SIP (superimposed pilot) by deprioritizing its study due to insufficient demonstrated gains and increased AI model complexity requiring 1.48M parameters and 80M FLOPs. For MRSS, they oppose signals/channels sharing between NR and 6G as well as SDM between the two RATs, and propose semi-static FDM/TDM and rate matching of 6GR PDSCH around NR signals/channels.

Key proposals

Your notes

Private to your account