Soft error considerations for deep-submicron CMOS circuit applications

The increasing importance of characterizing both memory arrays and core logic when estimating soft error FIT (failure in time) rates has been demonstrated using test-circuits, a 21264 Alpha microprocessor, and simulations. The reduction of operating voltage has been determined to increase the soft error rate exponentially at 2.1-2.2 decades/volt. Based on the SIA roadmap for CMOS scaling trends, meeting FIT rate requirements in the core logic will pose many challenges in the imminent future.

[1]  C. Lage,et al.  Soft error rate and stored charge requirements in advanced high-density SRAMs , 1993, Proceedings of IEEE International Electron Devices Meeting.

[2]  G. R. Srinivasan,et al.  Soft-error Monte Carlo modeling program, SEMM , 1996, IBM J. Res. Dev..

[3]  B. Davari CMOS technology scaling, 0.1 /spl mu/m and beyond , 1996, International Electron Devices Meeting. Technical Digest.

[4]  James L. Walsh,et al.  IBM experiments in soft fails in computer electronics (1978-1994) , 1996, IBM J. Res. Dev..

[5]  Yoshiharu Tosaka,et al.  Measurement and analysis of neutron-induced soft errors in sub-half-micron CMOS circuits , 1998 .