A More General Approach to the Interpolation of Sampled Data by Realizable Filters

This paper derives an expression for root-mean-square (rms) interpolation error in a sampled data system employing lumpedparameter presampling and postsampling filters. The derivation for rms error employs basically a time-domain approach rather than the more usual frequency-domain approach. Plots of rms error versus sampling rate are given for first- through fourth-order Butterworth data using a Butterworth interpolation filter and no presampling filter. It is shown that using 100-percent duty cycle samples (rather than impulsive samples) can greatly improve the rms error.