IC Timing Yield Mitigation via Statistical Cell Replacement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional process corner-based timing sign-off methodologies fail to address high-sigma rare local timing events, leading to significant timing yield loss in integrated circuit design due to increased variations in chip manufacturing processes and environments.
Innovation Solution
The method involves determining delay and pin slack distributions for cells in an integrated circuit design, calculating yield loss through convolution operations, and identifying cells with high yield loss for replacement with cells having less parametric variation, thereby mitigating yield loss by optimizing critical paths and reducing rare-event timing degradation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional process corner-based timing sign-off methodologies are used, then design verification is simplified, but timing yield loss occurs due to high-sigma rare local timing events not being captured
Solution Approach 1:
The patent segments the timing analysis problem by identifying and treating high-sigma rare-event cells separately from常规 cells. It divides the cell population into groups based on their yield loss characteristics, allowing targeted analysis and optimization of critical rare-event cells while maintaining efficient processing for other cells.
Solution Approach 2:
The patent applies partial action by focusing computational resources only on cells that contribute significantly to yield loss (high-sigma rare-event cells) rather than performing exhaustive analysis on all cells. This selective approach captures the critical rare events without the prohibitive cost of complete Monte Carlo simulation.
2Measurement precision
If Monte Carlo simulations are used to capture high-sigma rare events, then timing yield accuracy improves, but computational effectiveness decreases
Solution Approach 1:
The patent changes the approach from full probabilistic simulation to a deterministic method that uses statistical parameters (delay distributions, pin slack distributions) to estimate yield loss. By transforming the problem into parameter-based analysis rather than event-based simulation, it achieves accurate rare-event capture with computational efficiency.
Solution Approach 2:
The patent substitutes the mechanical Monte Carlo simulation process with a mathematical convolution-based calculation system. Instead of repeatedly simulating timing paths with random variations, it uses analytical methods combining delay and pin slack distributions to directly compute yield loss, replacing computational simulation with mathematical computation.
3Adaptability or versatility
If cells with high parametric variation are used, then design flexibility and availability increase, but timing yield loss increases due to high-sigma rare events
Solution Approach 1:
The patent applies local quality by allowing different cells to have different characteristics based on their location and function in the circuit. Critical cells on timing-sensitive paths use low-variation cells to ensure yield, while non-critical areas can use high-performance cells with higher variation, optimizing the trade-off between flexibility and reliability locally throughout the design.
Solution Approach 2:
The patent performs preliminary identification and ranking of cells by their potential yield loss contribution before final cell selection. By pre-characterizing cells using delay and pin slack distributions, designers can proactively select replacement cells that minimize yield loss before timing closure, rather than reacting to yield issues after fabrication.
Data Source
AI summary
Embodiments provide for mitigating parametric yield loss of an integrated circuit (IC) design. In certain embodiments, a delay distribution associated with at least one cell disposed in the design is determined. A pin slack distribution associated with paths in which the at least one cell is disposed is determined. A residual distribution is determined based at least in part on the delay distribution and the pin slack distribution. Yield loss associated with the at least one cell is determined based at least in part on the delay distribution and the residual distribution. When it is determined that that the yield loss associated with the at least one cell exceeds a yield loss threshold, the at least one cell may be identified as a candidate for replacement with a replacement cell.