A Self Learning based Diagnosis of Faulty Configurable Logic Blocks (CLBs) in Field Programmable Gate Arrays (FPGA) Using Reconfiguration

Abstract

In many areas of digital systems Field programmable gate arrays (FPGAs) are most important for designing. The main usesof FPGAs are, these are programmable, and faults can be easily diagnosed, once faulty locations are identified. The locationand identification of faults in FPGA has not yet been explored much. A methodology for the testing and diagnosis of faultsin FPGAs is presented based on automatic circuit reconfiguration. The proposed method imposes no hardware overhead.This method can also be used in fault-tolerant systems, in which a good functional circuit can be still mapped to a FPGAwith faulty elements, as long as the fault sites are known. The logic synthesis software assigns the Configurable Logic Block(CLB) resources without system designer intervention. It is very advantageous for the designer to understand certain CLBdetails, including the varying capabilities of the look-up tables (LUTs), the physical direction of the carry propagation, thenumber and distribution of the available flip-flops. FPGA consists of 25 Configurable Logic Blocks (CLB). Each CLB isassigned with an application. The inputs for CLB are applied from a file. There is also a fault file in which error CLBs arepresent. If there is error CLBs, those CLBs are replaced by the spare CLBs. Finally, the errors CLBs are corrected withproper inputs and modified bits are displayed. So efficiency is not reduced and configurability is done without replacing thefaulty components. This FPGA can tolerate not only single faults but also for multiple faults. The power analysis resultsprovided for fault free, stuck-at-1, stuck-at-0 faults in digital circuits validate the point that faulty circuits dissipates moreand hence draw more power.Key words: Configurable Logic Block (CLB), Power Dissipation, Fault Tolerance, Fault Diagnosis, Faults, Full adder (FA).