Dynamically determining number of simulations required for characterizing intra-circuit incongruent variations

摘要

A method is disclosed comprising using a circuit recognition engine running on a computerized device to detect a number and type of devices in an integrated circuit. The method characterizes device variation by selecting a set of dominant active devices and performing simulation using the set of dominant active devices. Three different options may be used to optimize the number of simulations for any arc/slew/load combination. Aggressive reduction uses a minimal number of simulations at the cost of some accuracy loss, conservative reduction reduces the number of simulations with negligible accuracy loss, and dynamic reduction dynamically determines the minimum number of simulations needed for a given accuracy requirement.

主权项

1. A method comprising:
using a circuit recognition engine running on a computerized device, detecting dominant devices in an arc under consideration; perturbing a device parameter value of each of said dominant devices, simulating said arc and making performance measurements to obtain a measure of sensitivity of said performance measurements to said perturbing said device parameter value of each of said dominant devices, using said computerized device; calculating, using said computerized device, overall sensitivity of said arc using an equationstotal=∑i∈D⁢si2 where stotal is said overall sensitivity,
D comprises a set of said dominant devices, andsi is a measure of sensitivity of said performance measurements to said perturbing said device parameter value of each of said dominant devices; and calculating, using said computerized device, an estimate of sensitivity error using an equation comprising:Error=stotal-stotal2+(Sv-∑i∈D⁢signi·si)2 where Sv is overall sensitivity of said arc when all device parameter values of all devices are perturbed simultaneously, and signi is an estimated positive or negative sign of si based on direction of perturbation.