Signal processing with charge-transfer devices

Abstract

NALOG signal processing has been extensively stimulated A by the recent developments in charge-transfer devices (CTD’S). The current availability of distributed analog storage and contiguous tapping structures has made direct realization of transversal filters possible in silicon large-scale integrated (LSI) technology. Once a transversal filter is realizable, then many other signal-processing functions can be implemented. First in importance are the time-invariant finite impulse response (FIR) filters. These can be constructed for preemphasis, reemphasis, or bandpass operation and can be applied to a wide variety of applications since the ability to control the clock frequency and. consequently, the sampling rate of the CTD’S allows the frequency response of these filters to be determined by the application in which they are used. In addition, FIR filters of a more general type can be constructed which are quite difficult to realize by other technologies. Arbitrary FIR filters with bounded impulse responses can be constructed directly by sampling the desired impulse response at the Nyquist rate and implementing these samples as the tap weights of a charge-coupled device (CCD) transversal filter. The quadrature filter which shifts the phase of all frequencies present in the input signal by 7r/2 rad is particularly useful in communication for providing single sideband (SSB) modulation, and in signal processing for providing the analytic signal representation of an arbitrary signal. This filter is directly implementable as a Hilbert transform of the input signal and, as such, can be considered a convolution of the original signal with a function of the form I/t. In addition to the implementation of time-invariant linear filters, contiguously tapped CTD’S can be used to implement some important time-varying filters. In particular, the discrete Fourier transform (DFT) can be implemented for all values of the block size N by means of fixed weight FIR filters and point-by-point multiplication. This implementation of the DFT is called the chirp-Z-transform (CZT) and allows direct computation of a DFT in real time. By simple modification, the corresponding discrete cosine transform (DCT) and discrete sine transform (DST) can also be computed. Although CTD’S offer significant advantages in signal processing, there are some limitations. Unlike digital signal representation, the analog signal can suffer degradation during processing. Some of these limitations are caused by dark current, charge-transfer inefficiency, and clock noise. In addition, there are limitations imposed by the limited dynamic range, tap weight inaccuracy, and output signal amplifiers of the