R1-2601833 discussion

On PRACH and RACH procedure

From Nokia
Status: not treated
WI: FS_6G_Radio
Agenda: 10.5.1.2
Release: Rel-20
Source: 3gpp.org ↗
Nokia's prior position on 10.5.1.2 at RAN1#124 · AI-synthesized, paraphrased
verify sources →
Proposes Zadoff-Chu sequences for 6G PRACH preamble due to CAZAC property and frequency offset robustness, and refrains from sequences lacking good cross-correlation under timing errors. Proposes 4-step RACH as the sole baseline procedure, with Msg3-based identification as baseline and Msg1-based identification restricted to limited use cases. Requires short PRACH formats based on 30 kHz SCS (baseline from NR) and long PRACH formats based on 5 kHz SCS for the 7 GHz frequency range. Proposes supporting PRACH and PUxCH overlapping to enhance RACH capacity through increased time-domain allocations. Proposes studying AI/ML-based Spatial-Domain Beam Prediction (BM-Case1) for coverage extension and PRACH attempt reduction, and Temporal-Domain Beam Prediction (BM-Case2) for improving Msg2/3/4 transmission/reception in 4-step RACH.

Summary

Nokia's contribution to 3GPP RAN1 #124bis on 6G PRACH and RACH procedure provides a comprehensive design framework with 27 proposals and 15 observations. The document covers PRACH sequence selection, preamble formats (especially for 7 GHz), configuration mechanisms (including parameterized formats and reduced table-based configurations), RACH procedure simplifications, and the application of AI/ML-based beam prediction to initial access.

Position

Nokia proposes retaining Zadoff-Chu as the PRACH sequence due to its CAZAC properties and robustness against frequency offset, while opposing sequences that lack these characteristics. For PRACH configuration, they recommend studying parametrized PRACH format configuration against static formats definitions and support reduced table-based PRACH configurations with separately configurable parameters to enable RO clustering and time adaptation. They require a unified design for terrestrial and non-terrestrial networks where feasible, and propose that SBFD RA support should not increase complexity of basic TDD RA operation. For RACH procedure, they propose 4-step RACH as the sole baseline, requiring Msg3-based identification as the default and restricting Msg1-based identification to coverage extension and OD-SIB1 use cases. For AI/ML, they propose reusing Rel-18/19 BM-Case1 (Spatial-Domain Beam Prediction) and BM-Case2 (Temporal-Domain Beam Prediction), prioritizing spatial beam prediction over temporal, and present simulation results showing 64-65% relative reduction in MSG1 failures using spatial beam prediction compared to classical methods.

Key proposals

Your notes

Private to your account