R1-2601828
discussion
On remaining aspects of channel coding in 6GR
From Nokia
Nokia's prior position on
10.3.1
at
RAN1#124
· AI-synthesized, paraphrased
verify sources →
Proposes defining the boundary between NR-range and beyond-NR-range data rates using the maximum TBS supported in 5G NR Rel-15 per carrier. Requires any new LDPC base graph for beyond-NR-range to preserve the QC-LDPC dual-diagonal structure and sub-matrix construction from 5G, citing encoding efficiency advantages from avoiding matrix B inversion. Presents technical evidence that reusing BG1 with an optimized puncturing pattern (puncturing columns 0 and 26 instead of the first two systematic columns) yields 0.2–0.44 dB gain at low maximum iterations (e.g., 5 iterations) and reduces average normalized decoding iterations. Questions the necessity of UCI payloads beyond NR range for polar codes and proposes that any extension should design segment counts based on code rate and information length, with SCL decoding at list size 8 or 16 as the baseline receiver. Requires that block codes for small block lengths (≤11 bits) remain identical to 5G, arguing minimal gain exists even under ML decoding.
Summary
Nokia presents 8 proposals and 6 observations on 6G channel coding, covering LDPC extension for data channels, Polar codes for control channels, and block codes for small payloads. The document argues for reusing NR designs where possible, proposes specific design constraints if new base graphs are studied, and emphasizes the need for fair comparison methodologies, particularly regarding decoding schedules and complexity evaluation.
Position
Nokia proposes defining the boundary between NR range and beyond-NR range for data rates using the maximum TBS supported in 5G NR Rel-15 per carrier. They propose reusing BG1 with an optimized puncturing pattern (e.g., puncturing columns 0 and 26 instead of the first two systematic columns) to achieve faster convergence at low maximum numbers of iterations, presenting simulation results showing 0.2-0.44 dB gain at MaxIts=5 and reduced average normalized iterations. For any new BG design, Nokia requires keeping the same dual-diagonal QC-LDPC structure as 5G, limiting max code block size to 8448, setting minimum code rate to approximately 2/3, using single-edge design, and puncturing only one column. They oppose relying solely on reverse decoding order for BG1 evaluation comparisons, arguing it causes up to 0.65 dB degradation at high code rates and advocating for near-optimal or top-down schedules instead. For control channels, Nokia questions the necessity of UCI and DCI payloads beyond NR range without further input from other agenda items, and proposes keeping NR RM block codes unchanged for small payloads given their optimized minimum distance and FHT-based ML decoding efficiency.
Key proposals
- Proposal 1 (Sec 2.1): To facilitate the discussion on the data rate beyond NR range for the study of LDPC extension, RAN1 to consider discussing the max TBS supported in 5G NR Rel-15 per carrier.
- Proposal 2 (Sec 2.2.1): If reducing the MaxIts is a design objective for the study of LDPC extension for very high data rate, RAN1 to consider reusing BG1 with optimized puncturing pattern as one candidate for further study.
- Proposal 3 (Sec 2.2.2): For LDPC extensions beyond 5G NR data rate range, in case an additional BG is studied, row orthogonality should be considered together with other criteria for a complete design, and maximum code block size is 8448, the lowest supported code rate should at least be higher than BG1's lowest code rate (e.g., 2/3), puncturing one column, the same BG structure as for 5G BGs should be kept, i.e. using single-edge QC-LDPC with dual-diagonal structure and sub-matrices construction as per 5G design principle.
- Proposal 4 (Sec 2.2.3): For a fair comparison with BG1, RAN1 should also consider alternative decoding schedules (other than reverse schedule) for BG1.
- Proposal 5 (Sec 3.1.1): For the study of UCI segmentation for UCI payload beyond NR range, the number of segments should be designed by taking into account the code rate and information length as well as the encoding and decoding complexity.
- Proposal 6 (Sec 3.1.1): For 6G polar codes evaluation, a baseline receiver should use successive cancellation list (SCL) decoding with a list size of or . The key performance metrics for this evaluation are the overall and undetected error probabilities.
- Proposal 7 (Sec 3.1.2): For DCI payload beyond NR range, proposals should be carefully evaluated. RAN1 can concentrate on a short list that requires only minor changes to the existing framework.
- Proposal 8 (Sec 3.2): For 6G, block codes for small block lengths should be kept the same as in 5G.