Multi-Corner Static Timing Analysis Sensitivity Method
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional deterministic static timing analysis (DSTA) is inefficient due to its reliance on multiple timing runs for various process corners, leading to excessive computational burden and overly pessimistic timing margins, which are exacerbated by inaccuracies in cell and wire delay modeling and sensitivity to process and environmental variations.
Innovation Solution
A multi-corner static timing analysis method that identifies timing tests with passing and failing slacks, computes sensitivities to variables, and compares them to thresholds to determine failing tests, thereby reducing unnecessary timing margin and improving design robustness by isolating and addressing critical sources of variation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple DSTA timing runs are executed for various process corners, then timing closure is achieved, but computational burden increases excessively
Solution Approach 1:
The patent segments the timing analysis process into two distinct phases: (1) a comprehensive multi-corner DSTA phase to identify all potential timing failures, and (2) a targeted sensitivity analysis phase that focuses only on the identified failing tests. This segmentation allows the method to achieve timing closure while reducing overall computational burden by avoiding redundant analysis of passing tests.
Solution Approach 2:
The patent applies partial action by performing complete multi-corner DSTA analysis only for tests that actually fail, rather than re-running all tests multiple times. The sensitivity analysis is applied selectively to failing tests based on their sensitivity thresholds, eliminating excessive computational action on tests that already pass or are not sensitive to process variations.
2Reliability
If timing margin is increased to account for process variations, then design robustness improves, but timing performance deteriorates
Solution Approach 1:
The patent applies local quality by differentiating the treatment of different timing tests based on their individual sensitivity characteristics. Tests with high sensitivity to process variations receive increased timing margin and robustness checks, while tests with low sensitivity maintain tighter timing margins. This localized approach ensures robustness only where needed, preserving overall timing performance.
Solution Approach 2:
The patent dynamically adjusts timing margin parameters based on the sensitivity analysis results. By changing the timing margin parameter selectively for specific failing tests rather than applying a uniform increase across all tests, the method achieves necessary design robustness while minimizing the impact on overall timing performance.
3Measurement precision
If sensitivity analysis is performed for all timing tests, then critical variations are identified, but computational resources are wasted on non-critical tests
Solution Approach 1:
The patent performs preliminary action by conducting a initial screening of all timing tests to identify failing tests before performing detailed sensitivity analysis. This preliminary identification step allows the computationally intensive sensitivity analysis to be applied only to the subset of failing tests, significantly improving analysis efficiency while maintaining measurement precision for critical variations.
4Reliability
If delay model inaccuracies are addressed by using extreme values, then timing safety is ensured, but timing margin becomes overly pessimistic
Solution Approach 1:
The patent substitutes the traditional mechanical approach of using extreme delay model values for all tests with a statistical sensitivity analysis approach. By replacing the conservative extreme-value method with a sensitivity-based methodology that considers actual process variation impacts, the patent maintains timing safety while achieving more accurate and less pessimistic timing margins.
Data Source
AI summary
Methods for identifying failing timing requirements in a digital design. The method includes identifying at least one timing test in the digital design that has a passing slack in a base process corner and a failing slack in a different process corner. The method further includes computing a sensitivity of the failing slack to each of a plurality of variables and comparing each sensitivity to a respective sensitivity threshold. If the sensitivity of at least one of the variables is greater than the respective sensitivity threshold, then the at least one timing test is considered to fail.