Analog neuro-chips with on-chip learning capability for adaptive nonlinear equalizers

In this paper, a modular analog neuro-chip set with on-chip learning capability is proposed for adaptive nonlinear equalizers in mobile communications. The analog neuro-chip set incorporates error backpropagation learning rule for practical applications. For modularity synapse cells and neuron cells we fabricated in separate chips to expand the size of network easily. Using these chip sets, an analog neuro-system is constructed and applied to adaptive nonlinear equalizers for mobile communication receivers. Using the decision feedback architecture, it successfully equalizes non-minimum phase channels as well as minimum phase channels. Unlike other systems requiring external DSP chips for adaptive learning, the analog neuro-system is self-contained and potentially capable of much faster adaptation