R1-2601829
discussion
On remaining aspects of modulation in 6GR
From Nokia
Nokia's prior position on
10.3.2
at
RAN1#124
· AI-synthesized, paraphrased
verify sources →
Proposes adopting 5G NR uniform QAM schemes as baseline 6G modulation and adding uniform 1024QAM in UL. Opposes uniform 4096QAM in DL due to R-ML MIMO detection complexity and EVM requirements. For geometric constellation shaping, requires down-selection from higher-order uniform QAM with sign symmetry, quantization with limited bits per I/Q, and minimization of shape variants per order. Deprioritizes AI two-sided model approaches while allowing AI receiver evaluation. Presents a technical case against probabilistic constellation shaping, arguing AWGN gains do not carry over to fading channels and highlighting impacts on HARQ retransmissions, PAPR, and MIMO detector complexity. Proposes studying MCS tables that allow multiple modulation-order/code-rate pairs per spectral efficiency point.
Summary
Nokia presents 9 proposals and 5 observations on 6G modulation, advocating for 5G NR equal probability uniform QAM as the baseline while cautiously addressing potential enhancements like higher-order QAM, constellation shaping (geometric and probabilistic), and MCS table optimization. The document emphasizes implementation complexity, specification impact, and the need to decouple MCS table design from shaping techniques.
Position
Nokia proposes that 5G NR equal probability uniform QAM constellations be explicitly adopted as the supported baseline modulation scheme for 6G (not merely 'as basis for study'), while allowing separate discussion of enhancements. They propose supporting uniform 1024QAM in uplink but present a technical case against uniform 4096QAM in downlink, citing prohibitive computation complexity and power consumption of R-ML MIMO detection at the UE and very low required EVM. On constellation shaping, Nokia requires that the specification impact on Rel-18 advanced DL-MU-MIMO receivers be addressed, warning that per-MCS optimized shapes and probabilistic shaping factors would break the co-scheduled UE modulation detection framework. They require decoupling MCS table optimization from geometric/probabilistic shaping decisions, and propose studying spectral efficiency operating points achievable with more than a single modulation order to address MIMO layer imbalance.
Key proposals
- Proposal 1 (Sec 2): Support 5G NR equal probability uniform QAM constellation as the basic modulation scheme for 6G for DL CP-OFDM (up to 1024QAM), UL CP-OFDM (up to 256QAM), and UL DFT-s-OFDM (up to 256QAM plus pi/2 BPSK).
- Proposal 2 (Sec 3.1): Support uniform 1024QAM in Uplink.
- Proposal 3 (Sec 3.2.1): Address the specification impact of geometric shaping and probabilistic shaping, e.g. on DL-MU-MIMO.
- Proposal 4 (Sec 3.2.2): The optimization of MCS tables is an independent topic and shall not be considered as a by-product of geometric/probabilistic shaping.
- Proposal 5 (Sec 3.2.3): If geometrical constellation shaping were to be supported in 6G, following options reduce the implementation costs: down-selection from higher-order uniform QAM, quantization based on higher-order uniform QAM, and deriving lower-order GCS from the highest-order GCS by down-sampling.
- Proposal 6 (Sec 3.2.3): Evaluation of geometrically shaped constellations shall take the quantization into account, assuming a maximum number of bits per I/Q (e.g. 4, 6, or 8 bits).
- Proposal 7 (Sec 3.2.3): If geometrical constellation shaping were to be supported in 6G, strive to reduce the number of different geometrically shaped constellations of the same order.
- Proposal 8 (Sec 3.2.3): If geometrical constellation shaping were to be supported in 6G, the support of GCS can be embedded in a capability framework as a function of the number of spatial layers and modulation order.