In this paper, the dependence of sensing currents on various device parameters is comprehensively studied by simulating the complete crossbar array rather than its equivalent analytical model. The worst-case scenario for read operation is strictly analyzed and defined in terms of selected location and data pattern, respectively, based on the effect of parasitic sneak paths and interconnection resistance. It is shown that the worst-case data pattern depends on the trade-off between the shunting effect of the parasitic sneak paths and the current injection effect of the parasitic sneak leakage, thus requiring specific analysis in practical simulations. In dealing with that, we propose a concept of the threshold array size incorporating the trade-off to define the parameter-dependent worst-case data pattern. This figure-of-merit provides guidelines for the worst-case scenario analysis of the crossbar array read operations.
Kernel Application of the Stacked Crossbar Array Composed of Self‐Rectifying Resistive Switching Memory for Convolutional Neural Networks
