Research On SoC Test Data Compression Based On Partition Coding

Because of the rapid development of SoC design technology, the chip output increases correspondingly, and the design cycle is greatly shorted. At the same time, due to high integration and complexity of SoC, the test complexity and the test data volume increase with it. On the other hand, the bandwidth and frequency of the ATE are limited, too much time has to be spent on the test of the SoC and the cost increases largely. Therefore, Soc test is faced with huge challenge.At present, test resource partitioning for SoC offers an effective technology to these problems, which is mainly divided into three kinds:test pattern generation compaction, built-in self test and built-out self test. This thesis will mainly study the technology of test data partitioning compression of SOC, it belongs to BOST technology.The idea of test data partition compression is to make better use of the correlation among test vectors. According to the idea, the original test set is divided into several fixed-length blocks. The test fields of each block are compared and encoded along the vertical, The approach will certainly increase the chances of the test field that is related. There are a large number of irrelevant position in the original test set, which provide the idea foundation to improving the coding compression ratio. Based on the idea, two kinds of coding compression schemes are presented in this paper, which are index coding compression scheme based on test data packet merging and marking coding compression scheme based on test data packet compatibility.According to the index coding compression scheme based on test data packet merging, every block is divided into several groups with coloring algorithm of graphs, at the same time the number of the test fields in each group is counted. Then several larger groups are picked out and merged into standard fields, the standard fields are stored in the ROM of decompression architecture for indexing. Finally, the test fields are encoded according to indexing. In fact, the packet statistics method is used to encode in this scheme.According to the marking coding compression scheme based on test data packet compatibility, a reference module is selected in each block. Other test modules are marked and coded according to the relevance between its and the reference module. At the same time, in order to correctly decode, the reference is continuously updated so as to increase certain bits of it. In addition, the reference module must constantly be replaced dynamically in order to improve the coding efficiency.For the two kinds of coding schemes, the corresponding decompression ideas and experimental results are presented respectively. By comparison with other compression schemes, feasibility and validity of the two schemes have been proven. Comparatively speaking, not much extra hardware cost has increased.