The Research On Test Data Compression Of System-on-a-Chip (SoC)

The rapid progress in micro-electronic technology promotes the advent of System-on-a-Chip (SoC), which brings integrate circuit (IC) into a new period of development. The design of SoC mainly adopts the technique of reusable intellectual-property (IP) cores, so it can accelerate the process of development and improve the system performance. However, with the increase in the number of IP cores integrated, and its function becoming more complex, test data volume grows quickly, test access is also more difficult. All the cases pose the more challenges for SoC test. This dissertation makes research in test data compression of SoC.An important branch of compression method is to encode test data and that is used widely. The dissertation analyses decompression of some typical code schemes are very complicated, although their have high compression ratio. So distance-marking method is proposed in the dissertation, which is a variable-to-fixed-length code scheme. Not only compression ratio of the method is high, but also communication protocol is simple and decompression hardware overhead is low.Distance-marking compression method proposed in this dissertation firstly employs compatible compression of multiple scan chains to pretreat test data based on the character that there are lots of don't care bits in test set. The pretreatment achieves the test data sharing of multiple scan chains. Corresponding decompression cost of compatible compression of multiple scan chains is little, which only needs a set of fan out lines. But compression ratio of the pretreatment is considerable.Algorithm of test pattern generation determines the character that there are a few different bits between adjacent test patterns in the test set. High compression ratio can be achieved if code scheme based on difference sequence is used. Distance-marking method is a code scheme based on difference sequence. So test data can be compressed further if distance-marking method is used to code test data after the pretreatment. Compared to other similar code schemes, the predominance of the distance-marking method is that its codes are fixed, which reduces complexity of decompression greatly.Meanwhile, the dissertation also discusses the problem of concurrent test method for multiple cores. The approach that test sets are merged directly is presented. After using compatible compression of multiple scan chains to pretreat the merged test set, modified distance-marking method is used to compress test data. During test application test data is broadcasted to every core under test by bus, thereby implementing concurrent test of multiple cores. Compared to same type of concurrent schemes, the proposed method has outstanding advantage such as high compression ratio and easy implementation of decompression.