Address Sequences and Backgrounds with Different Hamming Distances for Multiple Run March Tests

Abstract

It is widely known that pattern sensitive faults are the most difficult faults to detect during the RAM testing process. One of the techniques which can be used for effective detection of this kind of faults is the multi-background test technique. According to this technique, multiple-run memory test execution is done. In this case, to achieve a high fault coverage, the structure of the consecutive memory backgrounds and the address sequence are very important. This paper defines requirements which have to be taken into account in the background and address sequence selection process. A set of backgrounds which satisfied those requirements guarantee us to achieve a very high fault coverage for multi-background memory testing.

Analyses of two run march tests with address decimation for BIST procedure

“…As have been shown earlier for the multiple RAM testing with the high fault coverage of UNPSF the optimal set of the address sequences should be used [10]. In a case of two-run RAM testing the pair of two address sequences A j and…”

“…All of these solutions are based on simple algorithms like pseudorandom address generation, address bit negation, address bit shifting, address and address bits permutation and many different counting sequences, like Gray code, anti-Gray code, Maximal average Hamming distance sequences and so on [10,13]. All of these solutions have been proposed and used to keep the low hardware overhead for the BIST implementation, as well as, to achieve the high fault coverage, at least for the UNPSF.…”

“…To achieve the high fault coverage of NPSF two address sequences A j and A k for two-run RAM testing should be different as much as possible [10,13]. As the measure of dissimilarity of any two sequences A j and A k , so called Manhatten city block or it's modification arithmetic distance defined as…”

Analysis of multibackground memory testing techniques

“…, m} as a metric were proposed and experimentally analysed in (Yarmolik and Mrozek, 2007). Based on this metric, the following statement was formulated and experimentally validated (Yarmolik and Mrozek, 2007;Yarmolik, 2008): This statement can be used for selecting the optimal values of background for memory tests generating only one pattern for k neighboring memory cells like the MATS+ and PS (4N) tests. According to this, in the case of multi-run memory testing, memory backgrounds should have the maximal Hamming distance between all pairs of backgrounds.…”

“…However, they can be modified to get detection abilities for NPSFs. Based on traditional March algorithms, various approaches have been proposed to detect NPSFs, such as the tiling method (Goor, 1991;Hayes, 1975), the twogroup method (Goor, 1991;Hayes, 1980), the row-March algorithm (Franklin and Saluja, 1996), transparent testing (Cockburn, 1995;Karpovsky and Yarmolik, 1994;Nicolaidis, 1996;Voyiatzis, 2006), pseudo-exhaustive testing (Karpovsky et al, 1995), testing based on different address sequences (Sokol and Yarmolik, 2006;Yarmolik, 2008) and different address seeds (Yarmolik, 2008), and the multibackground method (Yarmolik and Mrozek, 2007).…”