Low power current-sense amplifier with single-reference reuse and closed-loop feedback for energy-efficient RRAM computing-in-memory

Abstract

The device mismatch of non-volatile resistive random access memory (nvRRAM) and the performance limitations of conventional analog readout circuits cause low quantization precision and high power consumption in current RRAM-based computation circuits. To address these issues, we propose a Single-Reference Successive Approximation Current-Sense Amplifier (SRSA-CSA) based on single-reference source reuse and a closed-loop feedback architecture. Compared to state-of-the-art RS-CSA designs (Ye et al., 2023 [1]⁾, SRSA-CSA achieves a 2.83× faster readout latency (30 ns vs. 85 ns) and 23.4 % lower power consumption (64.67 μW vs. 84.6 μW) under 180 nm CMOS technology, offering a new paradigm for energy-efficient computation-in-memory chip design.