Bias-scalable inner-product approximation circuit using analog margin propagation

Bias-scalability is a property of an analog CMOS circuit where its functional response remains invariant with respect to the MOSFET's biasing condition, namely weak-inversion, moderate-inversion and strong-inversion. As a result bias-scalable circuits can seamlessly operate over a wide operating range without the need for re-biasing, re-design or re-calibration. In this paper, we present the design of a bias-scalable analog circuit that computes an inner-product between two vectors. At the core of the proposed design is a margin-propagation based piece-wise approximation technique that relies only on universal conservation principles (like KCL and thresholding) to achieve bias-scalability. The concept is validated using measured results from prototypes fabricated in a 0.5μm standard CMOS process.