R1-2508729
discussion
Discussion on 6G channel coding
From OPPO
Summary
This document from OPPO contains 4 Proposals and 5 Observations on 6G channel coding, covering LDPC throughput extensions (Options 1-1 to 1-4) and Polar code payload size increase for DCI. The total is 9 numbered items (4 Proposals + 5 Observations).
Position
OPPO proposes that 6G LDPC study focus on extensions that at least double 5G LDPC throughput. For Option 1-1, OPPO prefers increasing lifting size (up to 2*Z for BG1 or 4*Z for BG2) over increasing systematic bit nodes to enlarge code block sizes, citing less specification impact. OPPO presents technical case that Option 1-2 (new protograph for less-iteration decoding) achieves 0.2dB gain at high code rate with better performance-complexity tradeoff, while Option 1-4 (edge reduction) shows no obvious gain over legacy 5G codes. For Option 1-3, OPPO observes 0.2dB gain at 4 iterations but warns of potential error floor increase and complexity growth if the protograph is expanded too large. On Polar coding, OPPO argues the 140-bit DCI payload limitation from the 5G interleaver is a bottleneck likely to be exceeded in 6G (up to 200+ bits with multi-cell scheduling), and proposes studying two interleaver extension schemes where Scheme 2 minimizes hardware and specification changes by only interleaving the last 164 bits.
Key proposals
- Proposal 1 (Sec LDPC coding): The study of 6G LDPC should focus on the extension that is about to at least double the 5G LDPC throughput.
- Proposal 2 (Sec Option 1-1): Study methods to support LDPC code block sizes larger than the maximum in 5G (Option 1-1), with the lifting sizes up to 2*Z for BG1 or 4*Z for BG2, where Z is the maximum lifting size in 5G.
- Observation 1 (Sec Option 1-2 vs Option 1-4): New protograph of LDPC can achieve better performance-complexity tradeoff with less-iteration decoding (Option 1-2) at high code rate.
- Observation 2 (Sec Option 1-2 vs Option 1-4): Our initial investigation does not reveal that edge reducing (Option 1-4) can achieve better performance-complexity tradeoff than 5G legacy code and Option 1-2 code.
- Observation 3 (Sec Option 1-2 vs Option 1-4): For less iterations, the performances of BG1/2 with top-to-bottom scheduling decoding can hardly represent the baseline 5G code performance. So the comparison between 5G code and 6G code still needs to be done on BG1/2 with other specifically enhanced decoding scheduling.
- Observation 4 (Sec Option 1-3): The following are observed for Option 1-3 (increasing the number of systematic bit nodes): It can achieve 0.2dB gain at 4-iterations with larger protograph compared with 5G BG1 in supporting high code rate; It may slightly decrease performance (less than 0.1dB) at large number of iterations; It may raise the error floor compared with larger lifting size solution; It may increase complexity if protograph is expanded too large.
- Observation 5 (Sec Polar coding): By following the RAN1 analysis assumptions for Rel-17 DSS and Rel-18 MC-enhancement, the total number of DCI information bits has a good chance to exceed 140 or even 200.
- Proposal 3 (Sec Polar coding): Study solutions to increase DCI payload size in 6G, where the principle of the solution should hold against the uncertainty of the exact DCI payload size upper-bound.
- Proposal 4 (Sec Polar coding): Study the following schemes of interleaving to enable the maximum DCI payload size to be larger than the one (140+24) in 5G. Scheme 1: Follow the legacy design principle to re-define the interleave table (i.e., defined in TS38.212) to support a maximum Polar code payload size that is larger than (140+24). Scheme 2: Only apply the legacy interleaving over the last (140+24) bits.