In this thesis the performances of adaptive differential pulse code modulation (ADPCM) and adaptive delta modulation (ADM) systems are compared in dynamic range, noise tolerance and peak signal-to-noise ratio (SNR) at different bit rates. The ADPCM system selected for the study has an adaptive quantizer with a one word memory. For ADM first order constant factor delta modulation (CFDM) has been selected. Also, continuously variable slope delta modulation(CVSD) has been chosen as a reference. Extensive computer simulations have been done using real speech for the comparative study.
According to the simulation results, dynamic ranges of ADPCM and CFDM are about the same (55-58 dB) for a 60 dB compander, but that of CVSD is much narrower (20-25 dB) than the other two. Those three coders studied have significant SNR gains over non-adaptive coders. The performance of ADPCM system is always better than that of CFDM system regardless of bit rate, but in low bit rate (16 Kbits/s) CVSD system in a little (about 1 dB) better than ADPCM system.
In a noisy channel, however, CVSD outperforms the other two coding methods. Comparing CFDM and ADPCM, the performance of the former is slightly better than that of the latter in a noisy channel.
In addition, an analytical study has been done with R-C filtered band limited Gaussian signal as an input. In this analysis overload average step size and granular average step size have been used to modify the firmulas for granular and slope overload noise obtained by Van de Weg and Protonotarios, respectively. The theoretical and computer simulation results agree closely.