Automatic design of high-order SC filter circuits

Abstract

The manual design of Switched-Capacitor (SC) filters can be a strenuous process. This task becomes even more complex when the high gain amplifier is replaced by a low gain amplifier due to the loss of the virtual ground node, increasing the complexity of the filter's transfer function and requiring the compensation of the parasitic capacitances during the design phase. This paper proposes an automatic procedure for the design of high order SC filters using low gain amplifiers. The design methodology is based on a Genetic Algorithm (GA) using hybrid cost functions with varying goal specifications. The cost function first uses equations for the estimation of the filter's transfer function and, once the specifications are met, the filter is further optimized in order to increase its robustness to random variations. Afterwards, the gain and settling time of the amplifier is also estimated using equations and optimized against several process corners. The use of equation-based cost functions reduces the computation time, allowing the use of larger populations to cover the entire design space. Once all specifications are met, the GA uses transient electrical simulations of the circuit in the cost functions, resulting in the accurate determination of the filter's transfer function, and obtaining the final design solution within a reasonable amount of computation time.