R1-2508737
discussion
Channel coding for 6GR air interface
From Huawei
Summary
This Huawei/HiSilicon contribution provides 28 observations and 5 proposals on channel coding evolution for 6G, focusing on LDPC extensions for higher data-rate data channels and Polar code enhancements for control information. The document argues against larger lifting sizes for LDPC, proposes fair comparison methodologies based on area efficiency and equal computational complexity, and recommends reusing NR Polar codes for control information within the NR range while allowing for simple extensions like increased mother code length and more CB segmentations if payloads exceed NR limits.
Position
Huawei opposes increasing the LDPC lifting size beyond current NR values, arguing multi-block parallel decoding with unchanged lifting size Z achieves 20% higher area efficiency than single-block parallel with doubled lifting size 2Z. They require that all LDPC extension evaluations compare BLER performance under identical computational complexity using a unified area efficiency model, rather than reporting throughput alone. They propose reusing NR Polar codes for control information within the NR range and present a technical case against two-stage DCI decoding, citing significantly higher FAR (10⁻² vs required 10⁻⁷) from segmented CRC and increased construction complexity from partial polarization. They argue RNTI-FAR is a system-level issue resolvable by gNB RNTI assignment strategy and split-reduced SCL decoding, not requiring standard changes. They propose scalable DCRC interleaver generation and increasing maximum PDCCH code length to 1024 bits (reusing the existing UCI sequence) as sufficient simple extensions for potential larger DCI payloads.
Key proposals
- Proposal 1 (Sec 2.1.2.2): Do not consider larger lifting value for LDPC extension considering the additional area overhead without any performance gain under the same area efficiency.
- Proposal 2 (Sec 2.3): When evaluating LDPC extension for higher throughput, a fair comparison should be conducted considering BLER performance under the same computational complexity, area efficiency alongside throughput, a unified area evaluation model, and both processing latency and extra waiting delay.
- Proposal 3 (Sec 3): The necessity and motivation of polar code enhancements need to be strongly justified first by identifying the actual source of existing problems and the specific requirements that differentiate 6G from 5G control channels.
- Proposal 4 (Sec 3.1): Confirm the working assumption with revision to reuse NR polar codes for control information within NR range (larger than 11 bits).
- Proposal 5 (Sec 3.1.2): If the maximum UCI payload size would exceed 1706 bits, more than 2 CB segmentations should be considered for UCI with payload size larger than 1706 bits.
- Observation 1 (Sec 2.1.1): For LDPC codes, quasi-cyclic (QC) structure is beneficial for hardware commonality between 5G and 6G.
- Observation 2 (Sec 2.1.2.1): For LDPC codes, fast convergence with reduced iteration number is simple and effective for higher throughput.
- Observation 3 (Sec 2.1.2.2): For LDPC codes, the different types of parallelism, i.e., single-block parallel, multi-block parallel, row parallel, and multi-decoder parallel decoding, can affect not only throughput but also hardware implementation.
- Observation 9 (Sec 2.1.2.8): For LDPC codes, fixed-code-rate BGs cannot support fine-granularity rate matching and IR-HARQ.
- Observation 10 (Sec 2.1.2.9): Multi-edge BGs in LDPC codes are incompatible with practical backward-compatible decoders. While single-edge decoding offers a workaround, it yields performance inferior to the BG1 baseline.
- Observation 11 (Sec 2.3.1): For LDPC codes, multi-block-parallel (e.g., two blocks with unchanged lifting size Z) decoding achieves 20% higher area efficiency (throughput/area) compared with single-block-parallel (e.g., one block with doubled lifting size 2Z) decoding assuming the same area.
- Observation 18 (Sec 3): RNTI-FAR is avoidable by gNB during RNTI assignment.
- Observation 21 (Sec 3.1.1): For potential larger DCI payload size, there exist scalable DCRC interleaver generation procedures that can ensure the NR same sequence for the existing DCI payload size region, and also provide future-proof (support arbitrary large Kmax in 6G and beyond).
- Observation 25 (Sec 3.1.3.1): The partial polarization in the two-stage DCI decoding incurs significantly higher construction complexity. The segmented CRC has a significantly higher FAR than the 6G requirement.
- Observation 27 (Sec 3.1.3.3): The reliability of control channels is the top priority, so sacrificing error correction performance through sequence optimization is not a viable option.
Revision chain
1 versions in this meeting · oldest first
-
R1-2508737 ← you are here discussion revised