Supported Features and Configurations

Aerial CUDA-Accelerated RAN 24-1 Download PDF

This release of the Aerial cuBB supports the following configurations and features. These features are verified with test vectors in a simulated environment using TestMAC and RU emulator.

  • Process multiple cells (homogeneous and heterogeneous configurations) using CUDA streams or CUDA graphs

  • Frequency multiplexing

  • BW: 100 MHz

  • Antenna ports: 4

  • SU-MIMO layers: up to 4

  • MU-MIMO layers: up to 4

  • Modulation and coding rates: MCS 0 – MCS 27

  • Optimized LDPC decoder

  • UCI on PUSCH (HARQ up to 11 bits + CSI part 1 + CSI part 2 up to 11 bits)

  • Time-interpolated channel estimation and equalization

  • SINR reporting to L2

  • MMSE-IRC receiver

  • Early HARQ in UCI.indication

  • Format 0 + DTX detection

  • Format 1 + DTX detection

  • Format 2 (unsegmented payload) + DTX detection

  • Format 3 (unsegmented payload) + DTX detection

  • SINR / confidence level reporting to L2

  • Format 0

  • Format B4 (multiple per slot in FDM)

  • Interference level reporting

  • Process multiple cells (homogeneous and heterogeneous configurations) using CUDA streams or CUDA graphs.

  • Frequency multiplexing

  • BW: 100 MHz

  • Antenna ports: 4

  • SU-MIMO layers: up to 4

  • MU-MIMO layers: up to 4

  • Modulation and coding rates: MCS 0 – MCS 27

  • Supports Cat-A O-RAN split and Cat-B O-RAN split. For Cat-A O-RAN split, PDSCH is implemented up to modulation and precoding (identity matrix precoder) For Cat-B O-RAN split, PDSCH is implemented up to the rate matching block.

  • Precoding (4 layers)

  • Process multiple cells (homogeneous and heterogeneous configurations) using CUDA streams or CUDA graphs.

  • Interleaved and non-interleaved mode

  • Aggregation level (AL) 1, 2, 4, 8, 16

  • 1, 2, 3 symbol CORESET

  • Precoding (1 layer)

  • Process multiple cells (homogeneous and heterogeneous configurations) using CUDA streams or CUDA graphs.

  • PSS, SSS generation

  • DMRS and PBCH generation and time-frequency mapping

  • Precoding (1 layer)

  • Process multiple cells (homogeneous and heterogeneous configurations) using CUDA streams or CUDA graphs.

  • NZP-CSI-RS

  • ZP-CSI-RS

  • Precoding (1 layer)

  • Support SRS reporting for normalized 8 bit I/Q samples for upto 32T32R BB Antenna ports.

  • Support SRS reporting according to 5G FAPI 222.10.04 for beamManagemnt, codebook and non-codebook SRS usage.

  • Support SRS reporting according to 5G FAPI 222.10.02 for SINR reporting.

  • Support 32 Transmit and Receive antenna ports

  • Support SRS-based channel estimation, buffering and FAPI-compliant reporting to L2

  • Support PUSCH and PDSCH beamforming weight (BFW) calculation from SRS channel estimates (regularized zero-forcing)

  • Support up to 4 layers multi-user MIMO PUSCH

  • Support up to 8 layers multi-user MIMO PDSCH

  • Limited set of configurations available

  • Standalone LDPC decoder

  • Support for C++ std::format syle logging like std::format("{} {}!", "Hello", "world", "something");

  • Support for C (printf) style formatted strings.

Previous cuPHY Software Mainfest
Next Multicell Capacity
© Copyright 2024, NVIDIA. Last updated on Jun 6, 2024.