Study On Ternary Delta-Modulation Coding Based On Hopfield Networks

With the development of the multimedia, analog-to-digital (A/D) and digital-to-analog (D/A) are becoming an indispensable technology. In view of the carried information in the signal and the dissipated power of the circuit, as well as the multiple-valued coding circuit, this dissertation presents the multiple-valued delta-modulation and the A/D converter based on the neural networks and that lower dissipated power and higher information density are obtained by using the developed technologies.Firstly, analyzing the theory and the circuit of binary logic delta-modulation, The theory and the related circuit topology of ternary logic delta-modulation are present. It is found that the ternary logic delta-modulation system has more lower power dissipation and lower quantizing noise and especially the enhanced capability in dealing with the slow-going signal by comparing with the binary and ternary logic delta-modulation. Secondly, considering the design of A/D converter as the core of the modulation system, more attention is paid in it and the successive approximation A/D converter due to its wide application. Considering the virtue of more encoded information in ternary than binary logic and analyzing the ternary A/D converter using two comparators, the directive effect of the switch-signal theory in the circuit design is pointed out. Because the design rule of the Integrate Circuits is lower power dissipation and smaller the chip area, a new A/D converter based on ternary logic in which a comparator with "window" characteristic is applied is proposed. This circuit reduces the number of comparator and simplifies the circuital structure. At last, According to the self-study and self-adapting characteristic of Neural Networks and the similarity in structure between the Hopfield network and A/D converter, theHopfield network is applied on the circuit of A/D converter. But a disadvantage exists in the local minimum of the binary logic A/D converter using Hopfield network and it will do recede the coding precision. In order to overcome it, basing on the successive approximation A/D converter and the switch-signal theory, study of the binary logic A/D converter using ahead networks is analyzed and applied the arithmetic in the ternary logic and put forward the ternary logic A/D converter using the ahead network. This proposed A/D converter overcomes the local minimum, heightens the coding precision and shortens the converting time.