Globalfoundries Inc.

Vineet Wason

Zhi-Yuan Wu

Jung-Suk Goo

Kevin J Yang

Ciby Thuruthiyil

Sriram Balasubramanian

Lingquan Wang

Juhi Bansal

Karthik Chandrasekaran

Varsha Balakrishnan

Arunima Dasgupta

Worst case performance of an SRAM cell may be simulated more accurately with less intensive computations. An embodiment includes determining, by a processor, a process corner G of an SRAM cell, having pull-down, pass-gate, and pull-up devices, process corner G being defined as the worst performance of the cell when only global variations of parameters of the SRAM cell are included, setting each of the pull-down, pass-gate, and pull-up devices at process corner G, performing, on the processor, a number of Monte Carlo simulations of the SRAM cell devices around process corner G with only local variations of the parameters, generating a normal probability distribution for Ibased on the local Monte Carlo simulations around process corner G, extrapolating the worst case Ifrom the normal probability distribution of Ito define a process corner SRM representing a slowest SRAM bit on a chip, and validating an SRAM cell based on the SRM corner. Embodiments further include creating a library of SRM corner values for multiple SRAM cells, and validating an SRAM cell by selecting an SRM corner from the library. Embodiments further include linearly scaling the SRM corner value with global sigma input variations from 0 sigma to 6 sigma and/or with local sigma input variations from 0 sigma to 6 sigma, selecting a scaled SRM corner value at the sigma corresponding to design and memory size requirements for the SRAM cell, simulating the scaled SRM corner by a processor, and employing the simulated scaled SRM corner to validate performance of an SRAM cell.

Model library implementation and methodology for worst case performance modeling for SRAM cells

According to one exemplary embodiment, a method for robust statistical semiconductor device modeling includes building a semiconductor device model using at least one new device parameter variation, constructing a variation library for the semiconductor device model, and verifying the variation library against measured data from physical semiconductor devices. The variation library is constructed by determining variations of the at least one new device parameter variation and standard device parameters as functions of, for example. sizes and locations of semiconductor devices on semiconductor dies.

Method for robust statistical semiconductor device modeling

Growing community of inventors

Santa Clara, CA, United States of America

Vineet Wason