VarMan Statistical Variation and Yield Analyzer
When designing using the latest semiconductor process technologies, process variations across a silicon wafer or device must be carefully accounted for to achieve targeted specifications and production yields. This may require thousands, millions, or even billions of Monte Carlo simulations, because some devices, such as bit-cells within a memory, may need analysis to go to 7 sigma and beyond.
Traditional Monte Carlo simulation is too slow for these purposes, particularly with tight design schedules. VarMan is a breakthrough technology delivering up to 30X simulation performance compared to traditional Monte Carlo simulation, while guaranteeing the same level of accuracy. Its high sigma analysis capabilities provide very fast and highly accurate yield estimation and the ability to quickly identify sample fails and design weakness.
VarMan is easily integrated into a design flow. It supports common third party and proprietary transistor-level simulators and has an intuitive, easy-to-use interface. It provides true statistical corners extraction and local variability analysis that can handle a huge number of parameters.
VarMan provides seven kinds of statistical analysis targeted analog, library and memory applications:
Fast Monte Carlo analysis:
- The innovative approach of VarMan to Monte-Carlo analysis provides equivalent results to classical Monte Carlo Analysis, with up to 30x+ time speed-up.
High-sigma performance limits:
- Given the required sigma (yield), this analysis will find the design performance limits that correspond to this yield. The analysis is very economical in simulations and robust to multi-failure zones.
High-sigma yield estimation:
- Given design performance limit, this analysis will quickly verify and estimate the yield to 4-6+ sigma with a very limited number of simulations.
- Captures the process configurations that trigger parametric failures, to accurately estimate the yield and to predict extremely rare event
High-sigma spread (HSS):
- “High Sigma Spread” (HSS) generates a complete view of a cell yield in one operation, exploring sigma from very low to high, to ensure a design with best power, performance, and area. HSS is robust for non-linearity and non-Gaussian distributions, commonly seen in advanced process technology nodes.
Variability eXplorer analysis:
- Illuminates design performance-critical zones based on a user-controlled simulation budget, and will identify marginalities from 3-sigma to high-sigma
- Highly cost effective, providing variability induced marginal corners and the most influential parameter.
True corner extraction:
- Investigates the PVT parameters that lead to the performance most likely statistical corners to achieve a given yield. Compared to the long verification time of the huge number of PVT corner combinations, this analysis drastically reduces simulation runs to the essential true corners of the design.
Local variability analysis:
- An advanced sensitivity analysis to help the designer identify hotspots versus local variability sensitivity. With a limited number of simulations, VarMan is able to identify the most sensitive transistors to local variations.
- Breakthrough analysis techniques gives impressive simulation time speed up, significant time saving, and complete control of results accuracy
- Reliable and proven analysis validated on advanced FinFET and FDSOI technology nodes
- Supports all golden SPICE simulators and design environments
- User-friendly GUI, clear presentation of results with access to data and waveform into golden viewers, fast and easy to use ‘load and analyze’ model
- Smart simulation manager takes advantage of LSF/SGE/OGE/Slurm computing clusters.
- Analog, RF, Standard cells, IO, and memories
Machine Learning in the EDA-Specific Domain – 20 Years in the Making
Memory Statistical Characterization Solution with VarMan
Standard Cell Statistical Characterization with VarMan
Statistical Analysis Flow for Analog Design with VarMan
Variation aware design for advanced nodes and low power technologies