A 0.25 V 460 nW Asynchronous Neural Signal Processor With Inherent Leakage Suppression

Further power and energy reductions via technology and voltage scaling have become extremely difficult due to leakage and variability issues. In this paper, we present a robust and energy-efficient computation architecture exploiting an asynchronous timing strategy to dynamically minimize leakage and to self-adapt to process variations and different operating conditions. Based on a logic topology with built-in leakage suppression, the prototype asynchronous neural signal processor demonstrates robust sub-threshold operation down to 0.25 V, while consuming only 460 nW in 0.03 ${\rm mm}^{2}$ in a 65 nm CMOS technology. These results represent a 4.4 $\times$ reduction in power, a 3.7 $\times$ reduction in energy and a 2.2 $\times$ reduction in power density, when compared to the state-of-the-art processors.