Design for testability method at register transfer level

Abstract

The testing of sequential circuit is more complex compared to combinational circuit because it needs a sequence of vectors to detect a fault. Its test cost increases with the complexity of the sequential circuit-under-test (CUT). Thus, design for testability (DFT) concept has been introduced to reduce testing complexity, as well as to improve testing effectiveness and efficiency. Scan technique is one of the mostly used DFT method. However, it has cost overhead in terms of area due to the number of added multiplexers for each flip-flop, and test application time due to shifting of test patterns. This research is motivated to introduce non-scan DFT method at register transfer level (RTL) in order to reduce test cost. DFT at RTL level is done based on functional information of the CUT and the connectivity of CUT registers. The process of chaining a register to another register is more effective in terms of area overhead and test application time. The first contribution of this work is the introduction of a non-scan DFT method at the RTL level that considers the information of controllability and observability of CUT that can be extracted from RTL description. It has been proven through simulation that the proposed method has higher fault coverage of around 90%, shorter test application time, shorter test generation time and 10% reduction in area overhead compared to other methods in literature for most benchmark circuits. The second contribution of this work is the introduction of built-in self-test (BIST) method at the RTL level which uses multiple input signature registers (MISRs) as BIST components instead of concurrent built-in logic block observers (CBILBOs). The selection of MISR as test register is based on extended minimum feedback vertex set algorithm. This new BIST method results in lower area overhead by about 32.9% and achieves similar higher fault coverage compared to concurrent BIST method. The introduction of non-scan DFT at the RTL level is done before logic synthesis process. Thus, the testability violations can be fixed without repeating the logic synthesis process during DFT insertion at the RTL level.