Surrogate Netlists for Fast Tail-Region Variation Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The high computational cost and time required for characterizing the tail regions of a circuit's response distribution due to process variations in integrated circuits, especially in large designs with high standard deviation requirements, hinder efficient yield analysis and design optimization.
Innovation Solution
Utilizing surrogate netlists that simulate faster than the actual target netlist, reducing runtime by performing initial simulations to identify the tail region, followed by accurate simulations of the target netlist to characterize the response distribution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If circuit simulations are performed on the actual target netlist to characterize the tail region distribution, then measurement precision is improved, but loss of time increases significantly
Solution Approach 1:
The patent applies preliminary action by first performing simulations on a surrogate netlist to identify the tail region of the distribution before conducting accurate simulations on the target netlist. This preliminary identification step allows the subsequent characterization to focus only on the relevant tail region, reducing the overall simulation runtime while maintaining accuracy where it matters most.
Solution Approach 2:
The patent uses a surrogate netlist as a copy or approximation of the actual target netlist. This surrogate model allows for rapid initial simulations to locate the tail region, and then a smaller number of accurate simulations on the target netlist can be performed specifically for characterization, significantly reducing total simulation time while preserving measurement precision.
2Measurement precision
If a large number of simulations are performed to characterize the tail region distribution, then measurement precision is improved, but productivity decreases
Solution Approach 1:
By performing preliminary simulations on the surrogate netlist to identify the tail region boundaries before conducting the actual characterization simulations on the target netlist, the patent reduces the number of simulations needed from potentially thousands to a manageable number, thereby improving productivity without sacrificing measurement precision.
Solution Approach 2:
The surrogate netlist serves as a computational proxy that enables rapid preliminary analysis. This copying approach allows the methodology to efficiently locate the tail region and then focus computational resources on accurate characterization, significantly improving yield analysis efficiency while maintaining distribution characterization accuracy.
3Measurement precision
If full-accuracy simulations are performed on the target netlist, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent performs preliminary identification of the tail region using surrogate netlist simulations before conducting full-accuracy simulations on the target netlist. This preliminary step defines the specific parameter ranges and conditions where accurate measurements are needed, allowing full-accuracy simulations to be focused only on those critical regions rather than the entire parameter space.
Solution Approach 2:
By using the surrogate netlist as a copy for preliminary analysis, the patent identifies which regions require full-accuracy simulation. This enables the target netlist simulations to be concentrated on the tail region specifically, reducing the overall time required while maintaining measurement precision where it is most critical for circuit response characterization.
Data Source
AI summary
In some aspects, surrogate netlists are used instead of the actual netlist of interest (the target netlist) in order to speed up the runtime for variation analysis. It is desired to characterize the distribution of a circuit's response as a function of process variation, in a tail region of the distribution. The tail region of the distribution is located by performing a variation analysis based on samples generated by circuit simulations of one or more transistor-level surrogate netlists. The circuit simulations of the surrogate netlists have shorter runtimes than circuit simulations of the target netlist, resulting in a decrease of the overall runtime. The distribution in the tail region is then characterized based on samples generated by circuit simulations of the actual target netlist.