R1-2600149 discussion

Transmission schemes for 6GR UL channels

From Huawei
Status: not treated
WI: FS_6G_Radio
Agenda: 10.5.2.3
Release: Rel-20
Source: 3gpp.org ↗

Summary

This Huawei/HiSilicon 3GPP Tdoc proposes physical layer enhancements for 6G Radio (6GR) uplink transmission, covering PUSCH, DMRS, and PUCCH/L1-UCI design. The document makes 34 formal proposals and 26 observations, focusing on UL MIMO codebook improvements for irregular UE antennas, high-order MU-MIMO support (up to ~100 layers), low-overhead scalable DMRS design with AI/ML receiver options, and simplified control channel mechanisms.

Position

Huawei proposes studying enhanced UL codebook designs that address performance losses from irregular UE antenna array layouts, presenting SLS results showing up to 50.3% CA and 51.1% CE spectral efficiency loss for 8Tx/4Tx NR DFT codebooks versus ideal SVD precoding. They propose high-precision UL coherent codebooks achieving 0.3–1.5 dB gains over NR R15 codebooks under improved UE coherency, and present AI/ML-based UL precoder generation with 30% SGCS gain and 10% SE gain over legacy codebooks. The company requires support for high-order MU-MIMO up to ~100 layers at 8T512R for around 7GHz deployments, achieving ~3.5x cell-average SE gains over 5G NR 4T64R, and proposes studying multi-layer DFT-s-OFDM MIMO transmission with codebook and DMRS design to be addressed in both waveform and MIMO agendas. For DMRS, they propose scalable low-overhead DMRS port numbers up to 96 with unified design for UL/DL CP-OFDM, favoring sparse orthogonal DMRS for AI/ML-based overhead reduction across both capacity and coverage scenarios while limiting SIP and DMRS-free solutions to coverage-enhancement scenarios only, and provide link-level evaluations showing 0.3–1.9 dB SNR gains with AI/ML CE/Rx receivers. They argue against moving UCI to L2 signaling, presenting a technical case that L1 UCI provides superior efficiency (avoiding 10–20x overhead increase), lower latency (avoiding retransmission delays), better coding performance with RM/Polar codes versus LDPC, and higher reliability with 1% BLER targets versus 10% for data.

Key proposals

Your notes

Private to your account