Memory Arrays for FMCW Radar Signal Processing:A Feasibility Analysis

This paper demonstrates that in-memory computing or memory arrays (MA) are feasible for low-latency radar signal processing, which is critical for modern radar applications. For frequency-modulated continuous wave (FMCW) radars, this work shows that the operating range (distance) of the radar is limited by the size of the MA and the radar's bandwidth. While larger arrays extend the operating range, parasitics deteriorate the accuracy of the measurement. Further, programming errors at lower conductance values have a greater impact on the accuracy of beat signal estimation and therefore memory devices that exhibit lower error at low conductance will result in better performance in radars. Finally, an architecture for calculating the analytic signal is proposed, and simulations demonstrate reasonable ($\sim 12 \%$) accuracy for the selected device parameters with nearly instantaneous theoretical latency.