246_pdfsam_VLSI TEST PRINCIPLES & ARCHITECTURES

246_pdfsam_VLSI TEST PRINCIPLES & ARCHITECTURES -...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
Test Generation 215 Case 1 stuck-at 0 z h b a c d e 1 0 0 0 1 Case 2 stuck-at 0 z h b a c d e 0 0 0 0 0 ± FIGURE 4.40 Fitness measure on how close a fault is excited. number of additional flip-flops set to logic zero or logic one, as done in [Rudnick 1994a] [Rudnick 1994b]. Note that only logic simulation is needed in this stage. For subsequent stages, the fitness function may take into account the fault-effects that propagate to flip-flops as well, as it may take several time frames in order to propagate the fault-effect to a primary output. As calculating the fitnesses of individuals dominates the computational cost of the GA, care must be taken when designing the fitness functions. Data structures that allow for fast access to the fault-free and faulty values in the circuit, for instance, would be desired. When fitness value calculation becomes prohibitive, one may reduce the cost by estimating the fitness instead of computing the exact fitness. In
Background image of page 1
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: stage one, for example, fault samples may be used instead of simulating all faults. Also, counting the number of events in logic simulation may be used to estimate the number of faults excited; this may eliminate the high cost of fault simulation. When using such fitness estimates, one must be aware of the potential loss in the quality of the derived solution and that the final fault coverage may also be reduced. 4.7.2.4 CASE Studies In the GA-based ATPG by Srinivas and Patnaik [Srinivas 1993], combinational cir-cuits were targeted. Each individual represented a test vector. The fitness function accounted for excitation of the fault and propagation of the fault effect. Depending on the fitness of an individual, different crossover and mutation rates were used. While test sets were large, high coverages were obtained....
View Full Document

This note was uploaded on 05/16/2011 for the course ENGINEERIN mp108 taught by Professor Elbarki during the Spring '08 term at Alexandria University.

Ask a homework question - tutors are online