A temperature stable current reference source with programmable output

A current reference source is presented that achieves stability over temperature with a neural network architecture and floating gate devices. It allows postfabrication adjustment to a user-specified output value. Tuning of floating gate devices is also used to accurately cancel temperature gradients at various temperatures. The circuit allows approximation of arbitrary functions over temperature. A first implementation for a limited temperature range yielded an accuracy of about 1.5%. Simulation results of an improved version are presented