Abstract: A speech digitization system including novel encoder and decoder circuits that minimizes the number of resolution bits required to produce a given level of speech quality by optimizing the information content of the digital output signal from the encoder. This is accomplished by providing a companded speech digitization system that includes an amplitude function generator which is adapted to produce an amplitude function signal that maintains substantial duty cycles on the digital output signal over the entire audio amplitude range. Included in the novel amplitude function generator is a unique bias network that serves to center the duty cycle swing of the digital output signal from the encoder around 50% where the information content of the signal is statistically maximized.