A current-mode piecewise-linear function approximation circuit based on fuzzy-logic

This work describes an analog circuit for function approximation working in current-mode. It implements a piecewise-linear fuzzy system: the membership functions are generated by the daisy-chain connection of identical blocks, resulting in a grid placement of the rules, and programming is achieved by binary encoded digital words representing the consequence of the rules. An approximated analytical model is derived, relating the I/O spread with the bias current variation. The current level allows speed and precision to be traded off for power consumption, while maintaining the same normalized I/O characteristic. The circuit is shown to operate correctly both in strong and weak inversion, with a bias current varying by two orders of magnitude