In this thesis the performance of adaptive delta modulation with hybrid companding (HCDM) has been studied. First, an analytical study on HCDM quantization noise has been done. New results for the signal-to-quantization-noise-ratio (SQNR) of HCDM have been obtained. A formula for granular quantization noise power of HCDM system with feedback estimation has been obtained assuming Gaussian input signal. This formula for granular noise has been combined with the modified formula for constant factor delta modulation (CFDM) slope-overload noise to obtain the analytical expression for overall quantization noise. To verify the analysis computer simulation has been done. Close agreement between the analysis and computer simulation results has been obtained. Second, optimization of system parameters has been done and discussed. With those parameters optimized, computer simulation has been performed using real speech to obtain SQNR as a function of the input signal level in the case of ideal and noisy channels. Performance of HCDM is compared with those of continuously variable slope delta modulation (CVSD) and constant factor delta modulation (DFDM). Among the three systems, HCDM yields the best performance in SQNR and dynamic range regardless the channel is ideal or noisy. Lastly, a hardware of HCDM has been built and experimental measurements have been performed. When an 800 Hz sinusoidal wave was used as an input signal, the measured peak SQNR of HCDM yielded 3 to 4 dB higher than CVSD that is most widely used at present.