Efficient compression of x-masking control data via dynamic channel allocation

A scheme is presented which uses one sequential linear decompressor to decompress test cubes and another sequential linear decompressor to decompress mask control data for masking unknown X´s in the output response. However, instead of the conventional approach of using a fixed number of tester channels for driving each, a method for dynamically adjusting the number of channels driving each is proposed. The key idea is that by carefully ordering the test cubes such that the channel requirements of the decompressor used for masking the previous output response (currently being scanned out) can be balanced with the channel requirements of the decompressor used for scanning in the current test cube. By matching test cubes with more specified bits with output response with fewer X´s, and test cubes with fewer specified bits with output response with more X´s through careful ordering of the test cubes, the total aggregate worst-case number of tester channels needed with dynamic channel allocation can be reduced considerably compared with conventional fixed channel allocation approaches thereby improving compression. For test cube and output response pairs where there are excess free variables, a method is described for boosting observability by selectively ANDing together outputs of multiple stages of the mask decompressor for driving the masking logic. Experimental results show that compression and observability can be significantly improved with the proposed scheme.