Paper on Low-Power mmWave Massive MIMO Equalizer ASIC accepted for publication in IEEE TCAS-II!

Wideband communication at millimeter-wave (mmWave) frequencies and massive multiple-input multiple-output (MIMO) are key technologies in fifth generation (5G) and beyond-5G wireless communication systems. These technologies, however, create enormous amounts of data to be processed per second, resulting in excessive hardware complexity and power consumption. Aiming at reducing the power consumption of uplink data detection, we present a 22nm application-specific integrated circuit (ASIC) of a beamspace equalizer for mmWave massive MIMO systems. The ASIC -- named GARAMON -- implements sparsity-adaptive equalization (SPADE), which is a sparse matrix multiplication technique that reduces power dissipation. SPADE exploits the inherent sparsity of mmWave channels in the beamspace domain in order to dynamically skip multiplications for which the magnitude of both operands are below pre-defined thresholds. ASIC measurement results demonstrate an equalization throughput of 46Gbps and show that SPADE offers up to 38% power savings compared to traditional antenna-domain equalization.

The paper was co-authored by Seyed Hadi Mirfarshbafan and Prof. Christoph Studer. A preprint is available on external page arXiv.