Answers are grounded in this paper and cite their sources.
Traditional Winner-Take-All (WTA) circuits rely on the matching of an array of N transistors, where N is the number of inputs of the WTA. This implies that when N increases, also increases the size of the WTA and the distance between the transistors that have to match, which yields to transistor matching degradation and loss of precision in the overall WTA circuit performance. Furthermore, when multi-chip WTAs are required, transistor matching is even worse, and the performance is drastically degraded. In this paper we present a WTA circuit technique, based on current-mode principles, that does not rely on matching of a transistor array. The operation of the circuit depends on current transport and replication by local current mirrors with a small number of outputs. Thus N can increase and the precision is not degraded, not even if the WTA is distributed among several chips. Furthermore, the different chips that constitute the WTA can be of different foundries, and the precision of the WTA is not drastically degraded. Simulation results on these facts are provided.>
No takes yet. Share an insight, caveat, or question.
Serrano et al. (1995) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: