This correspondence describes a new quantization technique called hybrid adaptive quantization (HAQ) that uses instantaneous [1], [2] as well as syllabic [3] adaptation of the step size. Two types of instantaneous adaptive algorithms have been used-Jayant's adaptive quantizer (JAQ) and the incremental adaptive quantizer (IAQ). Computer simulations have been performed for a sine-wave, correlated Gaussian signal and digitized speech. Signal-to-noise ratio (SNR) computation for PCM and DPCM coders indicates that the hybrid technique is superior to the normal adaptive quantizer, when both have the same ratio of maximum to minimum step size.

Speech encoding using unity bit coders in the form of delta modulators (DM) has become very popular. Many different designs have been proposed and implemented in monolithic form. The strategy used in all of these coders is to adapt the `step size' of the coder to suit the input signal power. The author examines a class of unity bit coders where a basic linear coder is used while the input to this coder is constrained or compressed to be within suitable limits, so as to have the linear coder operating in its optimal range. Results are presented for delta modulators, delta-sigma modulators, and a modified unit bit coder, and it is found that some of the coders perform as well as the `conventional' adaptive delta modulators.

Adaptive quantizers for PCM and DPCM coding of speech have been of great interest, and several excellent quantization strategies have evolved [1], [2]. In this paper we propose another algorithm for the step-size change in a quantizer. In the present strategy the quantizer is always a linear one except for the fact that the step size can be changed by a fixed amount at each sampling instant. This fixed value is also the minimum allowed step size. Hence, we have the name incremental adaptive quantizer. The quantizer has been simulated on a computer and its performance has been compared with that of Jayant's adaptive quantizer (JAQ), both in PCM and DPCM coders. The figure of merit used in the comparison is the signal-to-noise ratio (SNR) with a correlated sequence as the input. The preliminary results indicate that the new design can perform very well. However, it has the drawback of not being able to respond to very fast changes in the input. It can be easily realized in an integrated form using ROM's.

A multidigit adaptive delta modulation (ADM) system has been proposed where the error signal, between the input and the approximated signal produced by ADM coder, is coded in an auxiliary encoder. The error in the auxiliary coder is processed by another ADM and so on. The bit rate of each of these coders isf_{r}/Nwhere fris the overall transmission rate andNis the number of coders used. The bit streams are interleaved for transmission and at the receiver they are separated and decoded, and these signals are added and filtered. It is shown that for a given transmission rate, each coder operates at a basic sampling rate of frBsuch thatN_{opt} = f_{r}/f_{rB}gives the optimum number of coders to be used for maximum signal-to-noise ratio (SNR). A bound is derived for the maximum SNR of such a system and is compared with the bounds derived for other predictive coders. The experimental results of a two-digit ADM are presented. An average SNR of 30 dB is obtained with a dynamic range of 32 dB at fr= 32 kbits/s for band-limited noise signals. The SNR increases with the sampling rate at 15 dB/octave, as against 9 dB for a single-digit ADM. The frequency response is good and the variation of SNR with the message frequency of the delta coding system has been improved. The effect of channel errors has also been studied and the performance of the system is found satisfactory.

Two adaptive delta modulation (ADM) coders, one for A/D conversion of speech and the other for fiat spectrum signals, have been developed by incorporating a second adaptive feedback loop around the basic delta modulation (DM) coder. The dynamic range of the coder for integrated spectrum has been extended to more than 35 dB with a signal-to-noise ratio (SNR) > 28 dB, for a pulse-repetition frequency (PRF) of 32 kbits/s only, by using an adaptive circuit consisting of digital/analog slope detector and an FET expander. For flat spectrum signals, similar results with constant SNR over the band are obtained at a bit rate of 40 kbits/s. Thus the coders have performances equivalent to those of a 7-digit PCM, using only 32 and 40 kbits/s, respectively.