AdaAFE-CIM: A Hardware Implementation of Subspace Tracking for Adaptive Radar Data Compression

This paper presents AdaAFE-CIM, a novel architecture that implements subspace tracking for adaptive data compression in array processing hardware. The system combines analog feature extraction with digital subspace tracking to reduce power and hardware requirements in large beamforming systems. Our implementation forms 64 compressed channels on a 1024-element antenna array, tracks the signal subspace using a modification of Oja's algorithm, and updates the compression matrix in a closed loop. AdaAFE-CIM achieves a 95.4% reduction in power consumption compared to fully digitizing. The design uses custom 28 nm CMOS analog back-end circuitry at 1 GHz and digital components running subspace tracking updates at 1 kHz. Simulations with dynamic wireless channels show successful subspace tracking with compression, noise, and quantization.