Method for reducing the memory required to simulating a circuit on a digital computer

• Main Author: TROYANOVSKY; BORIS
• Format: Patent
• Language: English
• Published: 15.09.1998
• Edition: 6
• Subjects: CALCULATING; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

A method for simulating the response of a circuit to one or more stimulating signals using a digital computer. The circuit is represented by N nodes. Each node is connected to one or more devices. The nodes are held at a set of potentials represented by where Xn,-h=Xn,h*, n runs from 1 to N, and H is an integer greater than 0. The method of the present invention determines the values of the Xn,h by iteratively computing a new estimate for X=[X10, . . . , X1H, X20, . . . , X2H, . . . , XN0, . . . XNH] from a previous estimate for X based on Fnh, for n from 1 to N and h from 0 to H. Here, Fnh is the net current flowing into node n at frequency omega h. The net current at node n includes a non-linear resistive component rn(x(t)) and a non-linear charge storage component qn(x(t)), where x(t) is a vector having components xn(t). The method of the present invention defines a set of mapped frequencies {+E,cir omega +EE k} wherein 0</=k</=S-1, S>/=H+1. There is a integer valued function mu (h) such that, omega h is mapped to +E,cir omega +EE mu (h), where 0</= mu (h)</=S-1 for h=0 to H. Here, +E,cir omega +EE k is an integer, and +E,cir omega +EE k NOTEQUAL +E,cir omega +EE k' if k NOTEQUAL k'. The present invention iteratively solves the vector equation F=[F10, . . . , F1H, F20, . . . , F2H, . . . , FN0, . . . FNH] I=0. In each iteration, the quantities, =0 to 2S-1 are generated. The quantities ( alpha beta LAMBDA 0, alpha beta LAMBDA 1, . . . , alpha beta LAMBDA s) are then generated from ( alpha beta lambda 0, alpha beta lambda 1, . . . , alpha beta lambda 2S-1) by taking the Fourier transform of ( alpha beta lambda 0, alpha beta lambda 1, . . . , alpha beta lambda 2S-1). Similarly, the quantities ( alpha beta PHI 0, alpha beta PHI 1, . . . , alpha beta PHI S) are generated from ( alpha beta phi 0, alpha beta phi 1, . . . , alpha beta phi 2S-1) by taking the Fourier transform of ( alpha beta phi 0, alpha beta phi 1, . . . , alpha beta phi 2S-1). It can be shown that only a subset of { alpha beta LAMBDA xi } and { alpha beta PHI xi } need be stored.