IC Design Verification Using Stimulus-Response Model Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The high cost and time-consuming process of fabricating physical prototype integrated circuit (IC) designs, coupled with the need for efficient design verification to detect bugs and improve conformance to design requirements, are not adequately addressed by existing methods.
Innovation Solution
A method for IC design verification that involves selecting and adjusting stimulus-response data to train a model, using both known-good and IC-specific data to improve model accuracy, and applying correlation measures to verify accuracy, thereby reducing the need for extensive review and increasing efficiency in detecting design errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If physical prototype IC design is fabricated for verification, then design bugs can be detected and conformance to requirements can be improved, but the process becomes extremely expensive and time consuming
Solution Approach 1:
The patent creates a behavioral model (copy) of the IC design that simulates its response to stimulus waveforms. This model is trained using stimulus-response data from both the design under verification and known-good designs, allowing verification without physical fabrication. The model captures the essential behavior of the circuit while avoiding the costly and time-consuming mask set fabrication process.
Solution Approach 2:
The patent replaces the physical mechanical fabrication process with a computational modeling approach. Instead of physically manufacturing mask sets and fabricating IC prototypes, the system uses software-based behavioral modeling with machine learning to predict circuit behavior and detect design bugs, substituting physical fabrication with digital simulation and analysis.
2Measurement precision
If more stimulus-response data is used to train the model, then model accuracy improves, but the complexity of data selection and processing increases
Solution Approach 1:
The patent merges stimulus-response data from two different sources: the design under verification and known-good designs. This combination allows the model to learn both the specific behavior of the design being verified and the general expected behavior from proven designs, improving accuracy while the systematic data selection process manages complexity.
Solution Approach 2:
The patent implements an iterative feedback process where the model is trained, evaluated using correlation measures against held-out stimulus-response data, and then retrained with adjusted data fractions. The correlation measure feedback guides the selection of optimal data fractions from known-good and DUV designs, systematically improving model accuracy while managing processing complexity through automated evaluation.
3Reliability
If correlation measures are applied to verify model accuracy, then design errors can be detected with higher probability, but the review scope and processing requirements increase
Solution Approach 1:
The patent extracts and compares only the essential features of stimulus-response behavior through correlation measures. Rather than reviewing entire waveform datasets, the system calculates correlation coefficients that capture the essential similarity between model predictions and actual behavior, enabling efficient bug detection by focusing on deviations from expected patterns rather than exhaustive review.
Data Source
AI summary
In described examples, a method of testing an integrated circuit design under verification (DUV) includes selecting first and second stimulus-response data to generate a model, and adjusting model training data in response to model accuracy. The first stimulus-response data is selected from stimulus-response data for a known-good design similar to the DUV. The second stimulus-response data is selected from stimulus-response data for the DUV. The model is trained using the first and second stimulus-response data. A first correlation measure verifies model accuracy with respect to trained DUV stimulus-response data. A second correlation measure verifies model accuracy with respect to untrained DUV stimulus-response data. A fraction of trained DUV stimulus-response datasets in the second stimulus-response data is increased if the first correlation measure is greater than a first threshold, and a fraction of untrained DUV stimulus-response datasets is added if the second correlation measure is less than a second threshold.