In-Field Logic Repair of Deep Sub-Micron CMOS Processors

TL;DR

This paper introduces an in-field repair method for deep sub-micron CMOS processors that enhances reliability and reduces downtime by using spare blocks and partial threshold voltage recovery, with manageable area and power overheads.

Contribution

It presents a novel architectural-level repair technique combining spare block replacement and threshold voltage recovery to improve processor longevity and reliability in ultra deep-sub-micron CMOS chips.

Findings

70% area overhead tolerated

Less than 18% power overhead

Significant reduction in downtime and increased reliability

Abstract

Ultra Deep-Sub-Micron CMOS chips have to function correctly and reliably, not only during their early post-fabrication life, but also for their entire life span. In this paper, we present an architectural-level in-field repair technique. The key idea is to trade area for reliability by adding repair features to the system while keeping the power and the performance overheads as low as possible. In the case of permanent faults, spare blocks will replace the faulty blocks on the fly. Meanwhile by shutting down the main logic blocks, partial threshold voltage recovery can be achieved which will alleviate the ageing-related delays and timing issues. The technique can avoid fatal shut-downs in the system and will decrease the down-time, hence the availability of such a system will be preserved. We have implemented the proposed idea on a pipelined processor core using a conventional ASIC design flow. The simulation results show that by tolerating about 70% area overhead and less than 18% power overhead we can dramatically increase the reliability and decrease the downtime of the processor.