Title
Biologically inspired hierarchical structure for self-repairing FPGAs
Abstract
In this paper we follow the lead of biology, and utilize modular redundancy to achieve self-test and, especially, self-repair in VLSI systems. We first review dual- and tri-modular redundancy (DMR and TMR) and show why these well-established techniques alone are not sufficient for self-repair. We introduce two enhancements to these: (1) distributed BER test, and (2) self-diagnostics with autonomous repair logic. The first is used to distinguish between common transient errors and less-frequent soft faults. The second is used to distinguish and isolate the failing unit, which in FPGAs can usually be healed by reprogramming configuration memory using dynamic partial reconfiguration. To enable completely autonomous repair of soft and hard faults, we add one more redundant (spare) unit per cell, making a quadruple-redundant system (QMR*). We compare these methods as applied to several benchmark designs in a 28nm Xilinx Kintex-7 FPGA. We show that our QMR* approach extends system reliability to the order of centuries (neglecting wear-out effects).
Year
DOI
Venue
2017
10.1109/RECONFIG.2017.8279780
2017 International Conference on ReConFigurable Computing and FPGAs (ReConFig)
Keywords
Field
DocType
Self-repair,Healing,FPGA,Dynamic Partial Reconfiguration,Fault isolation,Self-diagnosis,Tri-modular redundant (TMR),Dual-modular redundant (DMR),Routeability
Spare part,Computer science,Parallel computing,Field-programmable gate array,Vlsi systems,Redundancy (engineering),Modular design,Control reconfiguration,Maintenance engineering,Embedded system,Built-in self-test
Conference
ISSN
ISBN
Citations 
2325-6532
978-1-5386-3798-2
2
PageRank 
References 
Authors
0.41
0
2
Name
Order
Citations
PageRank
David C. Keezer16817.00
Jingchi Yang220.41