Charge-packet splitting in charge-domain devices

The charge packet splitting operation employed by charge-domain filters to implement fractional multiplication has been described. Several experiments have been designed to determine the accuracy of this operation. The fabricated circuits investigate the dependence of split accuracy on coefficient value, channel width, barrier shape, barrier location, and the extra (dummy) barriers. A computerized data acquisition system has been designed and used in the characterization of the circuits. The dc and pulse transfer curves of these circuits show a good overall linearity. The transfer inefficiency, coefficient value, and coefficient error have been determined as a function of clock frequency. Results show that coefficient error is essentially independent of frequency, and that large coefficients and large channel widths yield higher accuracies. Dummy barriers decrease coefficient error significantly, especially in narrow-channel devices.