We present a Linear Feedback Shift Register (LFSR) like architecture, because the LFSR can bring a lot of data by using a few bits. We calculate the ATE data by Gauss-Elimination and put the ATE data to our decompression architecture to generate a lot of patterns. And one ATE data will run several times in the architecture. If some faults cannot be detected, we will generate the patterns which are brought by fault simulation. If still a few faults cannot be detected, the faults will be caught by directly modify the bits of ATE data. The less ATE data we use, the less cycles we need. Because the reason, we can get high speed testing. And the less data, the low-cost environment we need. The cycle we can save is up to 61.60x. And the average rate of compression can get 19.33x.
Circuits and Systems (ISCAS), Proceedings of 2010 IEEE International Symposium on, pp.437-440