Accurate timing, power and noise models for memory instances are required to achieve design closure in today’s system-on-chip (SoC) designs. Memory compilers, which use data-fitting technology through interpolation and extrapolation to generate memory instance models, often fail to meet accuracy requirements. A complete simulation-based re-characterization of such memory-compiler-created memory instances has become the most critical capabilities for an IP characterization tool. The combination of Magma’s SiliconSmart ACE characterization technology and FineSim Pro simulation technology provides an optimized characterization methodology with enough simulation power to acquire measurements for all arcs on all data points and to deliver the required accuracy in the created models.
“SiliconSmart ACE Memory Characterization provides ease of use through automation, efficiency through optimization and accuracy through simulation, significantly reducing the time and effort to model any memory instance,” said Anirudh Devgan, general manager of Magma’s Custom Design Business Unit. “With its simple and user-friendly functional description and multiple automation techniques, an IP characterization engineer can set up the flow quickly and characterize a memory instance without having deep knowledge of how the memory is designed. By integrating Magma’s industry-leading FineSim Pro simulator into the SiliconSmart ACE characterization tool, Magma not only improves accuracy and speed, but also lowers our customers’ overall cost of owning a memory characterization tool.”