A new design strategy for the Tank and Hopfield neural analog-to-digital converter

A neural analog-to-digital converter with resolution enhancement

“…As an example, for an input voltage of 7.5 volts, our network outputs (V3,V2,VI,V0) are equal to -0.5 volts, or halfway between logic "0" (0 volts) and logic "1" (-1 volts). Substituting these values into (6) yields the exact answer, i.e., output equals input. Similar results are obtained at the other switchpoints (integer multiples of 0.5 volts).…”

“…We have presented a new design strategy to eliminate the problems (missing output codes and unequal step widths) associated with the Tank and Hopfield A/D converter [6]. The experimental results of this new method is shown in Fig.…”

“…3, which indicates approximately equal step widths. The improvements are due to the development of a modified Hopfield neuron, which allowed a gain reduction from 160 to 16 [6,7]. Figure 4 shows a block diagram of the modified neuron.…”

“…The previous methods have used or assumed that the neurons were either digital or very high-gain analog amplifiers. Our method uses analog amplifiers for the artificial neurons with gains similar to their biological counterparts [6]. In this paper, we discuss the addition of a resolution enhancement method to a four-bit neural A/D converter.…”

Design and sensitivity analysis of feed-forward neural ADC's

“…After the first and probably the best known neural ADC was proposed by Hopfield [1], several schemes have been proposed to overcome the main shortcoming of the original work, namely the local minima and the limited speed due to multi-loop structure [2,3]. However, the increased complexity and even lower speeds caused both the original work and its extensions to become of less interest during the recent years.…”