Improving X-tolerant combinational output compaction via input rotation

Abstract

Combinational linear compactors can be used to compact the output response for a large number of scan chains into a smaller number of outputs. While some compactor designs can guarantee observation of all scan chains in the presence of a small number of X's, this may not be sufficient for designs with higher X densities. This paper describes an approach for using a combinational rotator between the scan chains and compactor to allow detection of faults even in the presence of high X densities. It is shown that the number of control inputs to the rotator is comparable to the number of control inputs required by conventional X masking approaches, but by not masking, the proposed approach is able to provide higher observability which translates to fewer test patterns, better compression, and better coverage of non-modeled faults. Moreover, the control data for the rotator has many more don't cares than the control data for X masking thereby making it easier and more efficient to compress with a linear decompressor. A heuristic procedure for ordering the inputs to a combinational compactor to increase the probability of observation for a given maximum shift distance is also presented. Experimental results indicate that high observability can be achieved using the proposed method with a relatively small number of control inputs.