Savannah: A Real-time Programmable mmWave Baseband Processing Framework
5G new radio (NR) frequency range 2 (FR2) in the millimeter-wave (mmWave) band has a much shorter baseband processing deadline compared to that in the sub-7 GHz FR1 band. This tight deadline requires an efficient real-time system for baseband processing using minimal computational resources. We demonstrate Savannah, a software framework for efficient mmWave baseband processing using minimal and heterogeneous computing resources, including CPU and eASIC. Savannah vectorizes matrix operations and memory access patterns in multi-input multi-output (MIMO) arithmetic, offloads low-density parity-check (LDPC) coding to an eASIC, and enables single-core operation. We demonstrate that Savannah, using a single CPU core and an eASIC, can support a 2×2 MIMO link with 100 MHz bandwidth under full uplink traffic load, yielding a data rate of up to 487 Mbps.
Authors
- Tingjun Chen (ORCID: https://orcid.org/0000-0002-5717-5755)
- Chung-Hsuan Tung (ORCID: https://orcid.org/0009-0005-0578-8153)
- Zhenzhou Qi (ORCID: https://orcid.org/0000-0001-6330-5184)
- Anuj Kalia (ORCID: https://orcid.org/0009-0004-1586-655X)
Institutions
- Microsoft (United States) (US)
- Duke University (US)
Publication Details
- Published
- 2024-12-04
- DOI
- https://doi.org/10.1145/3636534.3698843
- Citations
- 1
- Primary Topic
- Millimeter-Wave Propagation and Modeling
- Type
- article
- Field-Weighted Citation Impact
- 0.21
Funders
- National Science Foundation
- International Business Machines Corporation
- Defense Advanced Research Projects Agency