Input decompression enables a small number of scan input streams to load stimulus into a much larger set of internal scan chains. Papers have proposed and products have implemented several types of input-decompression architectures. Table 1 lists the most common.
The simplest input decompressor is broadcast scan. This device simply fans out each scan input to multiple internal chains. The main complaint against broadcast scan is that those chains receiving their values from the same scan-in pin have directly correlated values, which may impact fault coverage. In most practical implementations, however, this possibility has not been a problem. Engineers obtain the linear-spreader, space-expansion network by XORing combinations of scan inputs to each internal-chain input. The scan correlations are still there, but they are less direct than with broadcast scan.
Another approach to avoiding the scan-chain dependency associated with broadcast scan is the use of multiplexed scan configurations. In its simplest form, this technique uses several broadcast-scan configurations and provides one or more additional scan inputs to switch between them.
...Alternative approaches for input decompression rely on a sequential linear-feedback and spreading network. The sequential elements are based on LFSRs (linear-feedback-shift registers) or linear automata, but they achieve the same result: Buffer up the input variables from the scan-input streams so that scan cycles not requiring a lot of care bits can defer the use of the variables for later, more demanding cycles. ...
The role of an output compressor is to enable a large number of internal scan-chain output streams to merge to create either a much smaller set of external scan-output streams or to create a signature for each test or set of tests. Table 2 lists the most common compressors, many of which papers have proposed and products have implemented.
<< 1 2 3 4 >> ALL
