We introduce a novel method of test generation for microprocessors at the RTL using spectral methods. Test vectors are generated for RTL faults, which are the stuck-at faults on inputs-outputs of the different modules/registers in the circuit, and the vectors are analyzed using Hadamard matrices for Walsh functions and the random noise level at each primary input. This information then helps generate vector sequences. At the gate-level, a fault simulator and an integer linear program (ILP) compact the test sequences. RTL test appraisal also helps reveal the hard-to-test parts of the circuit. An XOR observability tree was used to improve the testability of those parts. We give results for a simple accumulator-based processor named Parwan. The RTL spectral vectors produced higher coverage in shorter CPU times as compared to a gate-level ATPG.

We define N-model tests that target detection of faults belonging to N specified fault models. We provide a method for deriving minimal tests using integer linear programming (ILP) without reducing the individual fault model coverage. Any test sequences, deterministic, random, functional, N-detect, etc., can be minimized for the given set of fault models. Stuck-at, transition, and pseudo stuck-at I DDQ faults are used as illustrations. We generate tests using Mentor Graphics FastScan ATPG tool employing a single fault model at a time. A minimized test set for the three fault models is then obtained by solving the proposed combined ILP problem. For s5378 benchmark circuit we achieved about 50% reduction in the number of vectors and 10% reduction in the I DDQ current measurements compared to the originally generated tests. We also propose a reduced complexity ILP approximation.

We introduce a novel spectral method of delay test generation for microprocessors at the register-transfer level (RTL). Vectors are first generated by an available ATPG tool for transition faults on inputs and outputs of the RTL modules of the circuit. These vectors are analyzed using Hadamard matrices to obtain Walsh function components and random noise levels for each primary input. A large number of vector sequences is then generated such that all sequences have the same Walsh spectrum but they differ due to the random noise in them. At the gate-level, a fault simulator and an integer linear program (ILP) compact these vector sequences. The initial RTL vector generation also reveals the hard-to-test parts of the circuit. An XOR observability tree was used to improve the testability of those parts. We give results for an accumulator-based processor named Parwan. The RTL technique produced higher gate-level transition fault coverage in shorter CPU time as compared to a gate-level transition fault ATPG.

No abstract

ATPG patterns of a digital sequential circuit contain temporally and spatially ordered bits as well as random (or don't care) bits. We synthesize BIST hardware that mimics these characteristics by controlled mixing of spectral components and noise. A Hadamard digital wave generator circuit produces all required spectral sequences and a weighted pseudorandom bit generator provides random bits. While these two blocks serve the entire circuit under test, specific to each primary input are two small blocks, one that combines the required Hadamard sequences in proper proportions and the other a randomizer that adds the required amount of noise if necessary. As an example, the FlexTest ATPG produced 55110 patterns for s38417, detecting 15472 of 31180 stuck-at faults. Sixty four-thousand of our BIST patterns detected 17020 faults as compared to 4244 detected by a previously reported spectral BIST method utilizing similar hardware overhead.