Error compensation of A/D converters using neural networks

摘要

The paper describes a new technique for the error compensation ofanalog-to-digital converters (ADCs). It can he considered theimprovement of the phase plane compensation technique: the idea is toexploit the generalization capabilities of artificial neural networks(ANNs) to reduce the huge amount of experiments required. The ANNbuilding and setting-up are carried out in an ANN simulation environmentusing an ADC behavioral model, whose errors can be fixed to knownvalues. In this way, a set of ADCs with very different performancescould be simulated in order to investigate about the usefulness of theproposed approach in very different working conditions. Interestingconsiderations are drawn in the analysis of the first results carriedout by comparing the behaviors of uncompensated and compensated ADCoutputs