A current-mode two-quadrant multiplier for analog array-based neural systems

In this paper an analogue two-quadrant multiplier suited for the implementation of large arrays of multipliers, as in the case of analogue VLSI implementation of artificial neural networks, is presented. To pursue low power consumption and to minimize the silicon area a translinear current mode approach has been adopted. Moreover, we coded the synaptic weight value as the absolute value and sign, thus splitting the circuit into a simple one-quadrant multiplier and a weight sign management. Experimental results are reported and discussed.