Abstract: An ADPCM encoder and decoder includes an adaptive quantizer and an adaptive predictor. A quantizing-step-size adaptation circuit determines a quantizing-step-size for the adaptive quantizer in dependence upon the ADPCM signal. When the quantizing-step-size is small, updating of predictor coefficients, produced in the adaptive predictor, in dependence upon a reconstructed signal and a reconstructed error signal derived from the ADPCM signal, is inhibited, the predictor coefficients being updated only in dependence upon a leakage function. This enables improved tracking to be achieved by the ADPCM arrangement.

Abstract: In an ADPCM transmission system, signalling frames of an incoming multiplexed PCM signal are identified and signalling bits extracted therefrom. During each signalling frame, an ADPCM encoder/decoder is controlled by the signalling frame signal to operate entirely for 3-bit ADPCM, instead of the regular 4-bit ADPCM, to which 3-bit ADPCM the extracted signalling bits are added for transmission. The deliberate alteration from 4-bit to 3-bit ADPCM at the encoder and decoder of the system in each signalling frame ensures proper adaptation and consequently no loss of tracking between the encoder and the decoder.

Abstract: An ADPCM encoding/decoding arrangement produces n-bit ADPCM codes which have 2.sup.n -1 possible codes, excluding the all-zeroes code. A range of error signals about zero is quantized by the encoder as a non-zero ADPCM code. In order to avoid such non-zero codes resulting in undesired changes of predictor coefficients of an adaptive predictor of the arrangement, updating of the predictor coefficients in dependence upon a reconstructed error signal derived from the ADPCM signal is inhibited when each such non-zero code occurs.