Analytical Model for Predicting Current Mismatch in MOS Arrays
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current mismatch in CMOS devices due to poly density gradient effects is not accurately predicted by conventional SPICE methodologies, leading to variations in device behavior and performance, especially in smaller technology nodes like 28 nm and 20 nm, which affects IC design and yield.
Innovation Solution
An analytical tool and method are provided to predict current mismatch in MOS arrays by modeling the impact of poly density gradient effects using standard deviations from central rows and columns, reducing the need for extensive silicon area and measurement points, and incorporating these variations into SPICE macro-models for more accurate IC design.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional SPICE methodologies are used to model current mismatch, then the modeling process is simple, but the prediction accuracy of current mismatch due to poly density gradient effects is poor
Solution Approach 1:
The patent introduces new parameters (standard deviations of poly density gradient effects in x and y directions) to the existing SPICE model. These parameters quantify the variation effects that conventional models miss, allowing accurate prediction of current mismatch while maintaining compatibility with standard modeling frameworks.
Solution Approach 2:
The patent uses standard deviation as an intermediary parameter that bridges the gap between physical poly density gradient effects and electrical current mismatch. By measuring and modeling this intermediate statistical parameter, the system can predict final device behavior without directly simulating complex physical variations.
2Measurement precision
If extensive silicon area and measurement points are used for characterization, then the accuracy of current mismatch prediction improves, but the cost and time required for characterization increase
Solution Approach 1:
The patent measures standard deviations at only two central rows and two central columns of the MOS array, which is a partial sampling approach. This partial measurement strategy is sufficient to capture the poly density gradient effects and predict current mismatch across the entire array, avoiding the need for exhaustive measurement of all cells.
Solution Approach 2:
The patent divides the large MOS array into manageable segments (central rows and columns) for measurement. By focusing on these representative segments, the system can infer characteristics of the entire array without measuring every cell, thus reducing characterization time and resource requirements.
3Measurement precision
If standard deviations are determined for all cells in the MOS array, then the prediction accuracy improves, but the silicon area required for characterization increases
Solution Approach 1:
The patent uses partial sampling by measuring only central rows and columns rather than all cells. This partial action provides sufficient statistical information to model poly density gradient effects across the entire array, achieving accurate predictions with minimal silicon area dedicated to characterization structures.
Data Source
AI summary
A system and method for designing integrated circuits and predicting current mismatch in a metal oxide semiconductor (MOS) array. A first subset of cells in the MOS array is selected and current measured for each of these cells. Standard deviation of current for each cell in the first subset of cells is determined with respect to current of a reference cell. Standard deviation of local variation can be determined using the determined standard deviation of current for one or more cells in the first subset. Standard deviations of variation induced by, for example, poly density gradient effects, in the x and/or y direction of the array can then be determined and current mismatch for any cell in the array determined therefrom.