Differential pulse code modulation (DPCM) is a lossy coding method in which the difference between two consecutive samples is encoded. In this process, a predictor is calculated from past sample values, which is subtracted from the current sample value and its prediction error is transmitted. On the receiving side, a predictor is also calculated from past reconstructed samples
and corrected with the transmitted prediction error. In contrast to DPCM, Adaptive Delta Pulse Code Modulation (ADPCM) also uses prediction, but adapts the predictive signal
to theinput signal, allowing better prediction. The method is used in speech digitization and has the advantage that much fewer bits are required to encode a sample, resulting in substantial data reduction. With this method, errors
can occur when there are large differences between adjacent samples. These can be corrected by sampling the full amplitude value again after a few samples instead of the difference value. A disadvantage of this method is that a level loss of 6 dB/octave occurs and thus the signal-to-noise ratio (SNR) drops by 6 dB/octave and high tones are transmitted worse.