A new scheme which can greatly reduce the effects due to finite open-loop gain and DC offset and has no limitation on the input is presented. The principle based on work by Nagaraj et al. (1987) is to perform a preliminary charge transfer operation preceding every desired charge transfer operation, thus obtaining a close approximation of the finite gain error which is stored and subsequently used for correction. The precision achievable is quite high. Its major applications will be in precision signal processing circuits. Simulation studies on PSpice verify the effectiveness.