Key points are not available for this paper at this time.
A popular method of time-varying quantization involves a multlplicative change in quantizer overload point after each sample time. Although this technique produces an average loading factor that is invariant with rms input, it is vulnerable to transmission errors. Here, we introduce a related quantizer which has the advantage of disipating the effects of transmission errors at any desired rate. On the other hand, adaptation is not perfect. The loading factor now depends on rms input and in designing the quantizer one must compromise between the goal of constant loading factor and the goal of rapid error dissipation. We provide a design method and apply it to the modification of a quantizer currently used for speech. The new quantizer dissipates transmission errors with a 5 ms time constant while maintaining accurate adaptation over a 40 dB input range.
Goodman et al. (Sat,) studied this question.