Nokia · 10.5.1.1

Nokia proposes reducing UE initial cell selection complexity by limiting PSS sequences to one (studying m-sequence and Zadoff-Chu based designs) and by using PSS bandwidth instead of SSB bandwidth for sync raster definition (Option 4) to make the raster sparser without impacting PBCH/SSB performance or channel allocation flexibility. They present evaluation results showing Zadoff-Chu with u=1/2 achieves lowest complexity, while 1/3rd m-sequence with 5 hypotheses and Zadoff-Chu with u=1 have similar complexity. For network energy saving, they propose studying relaxed SS/PBCH periodicity up to 160ms in clustered transmission patterns, on-demand SIB1 delivery extended from Rel-19 to standalone scenarios, and paging-conditional additional synchronization signals. For mobility measurements, they propose SSS-based measurements across all RRC states to ensure consistent cell coverage, with NZP-CSI-RS studied for CONNECTED mode, and argue L3 CSI-RS are not used in practice in NR.

Nokia proposes studying ZC sequence-based designs for PSS to exploit robustness against frequency offset, contrasting with NR's m-sequence approach and noting ZC sequences require fewer frequency hypotheses for reliable correlation peak strength. They propose studying a single PSS sequence to reduce initial cell selection complexity by a factor equal to the number of frequency hypotheses and synchronization raster points. For SSS, they propose Gold sequences as baseline, while noting these sequences lose their autocorrelation/cross-correlation properties under frequency offset, requiring the UE to have sufficient synchronization for IDLE mode measurements. They propose studying on-demand SIB1 delivery in both stand-alone and non-stand-alone scenarios, extending beyond Rel-19's multi-cell limitation. For FR2-1, they require 240kHz SCS SS/PBCH support to halve the time-domain footprint and increase cell spectral efficiency, with numerical analysis showing 2-12% DL resource gain depending on channel bandwidth and FDM assumptions. They propose studying AI/ML in initial access by fully reusing Rel-19 AIML beam management models for spatial-domain and temporal-domain DL Tx beam prediction.