Test Data Compression Using Selective Encoding of Scan Slices
Summary (2 min read)
Introduction
- In the best case, the authors can achieve compression by a factor of using only one tester clock cycle per slice.
- Recent growth in design complexity and the integration of embedded cores in system-on-chip (SoC) ICs has led to a tremendous growth in test data volume; industry experts predict that this trend will continue over the next few years [5].
- The number of ATE channels that can directly drive scan chains is limited due to pin count constraints.
- These compression schemes exploit the fact that scan test vectors typically contain a large fraction of unspecified bits even after compaction.
II. PROPOSED APPROACH
- As shown in Fig. 1, the proposed approach encodes the slices of test data (scan slices) that are fed to the internal scan chains.
- The first data code refers to bit 0 (first bit of group 0), and the second data code carries the content of group 0.
- Control codes 00, 01, and 10 indicate that the current slice code is a single-bit mode slice code, and control code 11 indicates the current slice code is a groupcopy mode slice code.
- As discussed before, two group-copy mode slice codes are needed to encode a given group, one to indicate the starting bit of the group, and the other to carry the content of the group.
A. Upper and Lower Bounds on Compression
- By deriving the upper and lower bounds on compression, this section provides more insights into the proposed compression method.
- Such methods, however, require structural information about the circuit.
- The lower bound can be expressed in terms of care-bit density.
- The number of target symbols in any scan slice can therefore be viewed as a random variable that follows a binomial distribution, i.e., Each target symbol in a scan slice must be mapped to one slice code, hence, the average number of slice codes for a given scan slice is .
B. ATE Pattern Repeat
- A property of the proposed compression method is that consecutive -bit compressed slices fed by the ATE are often identical or compatible.
- Therefore, ATE pattern repeat can be used to further reduce test data volume after selective encoding of scan slices.
- With ATE pattern repeat, these slice codes can be further compacted.
- The example shows Authorized licensed use limited to: DUKE UNIVERSITY.
III. DECOMPRESSION ARCHITECTURE
- Fig. 7 shows the state transition diagram of the decoder.
- The signal , when asserted to 0, resets the FSM to its initial state.
- TABLE II FIVE GROUPS OF OPERATIONS FOR THE DECODER Fig. Fig. 9. One bit of the buffer. may contain less bits).
- Therefore, in the group-copy mode, additional combinational logic is needed to address the other bits together with the first bit.
IV. EXPERIMENTAL RESULTS
- The authors apply the proposed approach to eight representative industrial circuits.
- Table III describes these circuits and the corresponding test sets.
- Table IV shows the compression results obtained using the proposed method for the different test cases.
- Therefore, the proposed method can achieve significant reduction in data volume over ATPG-compacted test sets.
- The authors compared the two methods for the same values of , the number of ATE channels.
V. CONCLUSION
- The authors have presented a test data compression technique for designs with multiple scan chains.
- This method does not require detailed structural information about the CUT, and utilizes a generic on-chip decoder that is independent of the CUT and the test set.
- While the hardware overhead depends on the number of internal scan chains, the authors have seen that for an industrial circuit with over 1-M gates, the overhead is only 1% for as many as 1024 internal scan chains.
- The clock inputs of these scan cells need to be appropriately gated so that they can be triggered separately from other cells in the same scan chain.
- Experimental results for eight industrial circuits show that compared to dynamically compacted test sets, up to 28 reduction in test data volume and 20 reduction in test application time can be obtained.
Did you find this useful? Give us your feedback
Citations
25 citations
Cites background from "Test Data Compression Using Selecti..."
...Therefore we have four categories: fixed-to-fixed [2], [3], fixed-to-variable [4]–[13], variable-to-fixed [14]–[17], and variable-to-variable [18]–[32]....
[...]
...They are based on the fact that even in compacted test sets the percentage of unspecified bits is large [1], [17]....
[...]
22 citations
Cites methods or result from "Test Data Compression Using Selecti..."
...We also compare our method with the Illinois scan [Shah and Patel 2001] and selective encoding [Wang and Chakrabarty 2008] on stimulus test data volume for the IWLS2005 circuits....
[...]
...…1.32 vga 32 17079 9211 44947048 8297 23459759 9500385 10907846 6.04 6.89 21.14 40.50 10.16 In Tables VI and VIII, we present a comparison of the proposed method with the Illinois scan [Shah and Patel 2001] and selective encoding [Wang and Chakrabarty 2008] on stimulus data volume compression....
[...]
...As shown in Tables VII and VIII, FFs and faults represent the number of scan .ip-.ops and the number of transition faults....
[...]
...It is found that the proposed method obtains much less stimulus data volume, as compared to the Illinois scan [Shah and Patel 2001] and selective encoding [Wang and Chakrabarty 2008] for the ISCAS89 circuits....
[...]
...Performance of Scan Flip-Flop Grouping Scheme in Xiang et al. [2007a] for LOC Transition Delay Fault Testing cir. faults FFs FC FC new untestable R1 R2 gs s13207 15602 669 82.53 65.48 2661 28.54 30.85 4 s15850 19046 597 78.8 63.1 2991 19.32 22.54 6 s35932 62798 1728 87.2 78.35 5558 0.98 1.14 158 s38417 49738 1636 96.8 84.04 6347 7.11 8.33 20 s38584 61254 1452 90.76 81.27 5814 7.49 9.65 16 usb 76917 1746 96.65 86.43 7861 8.24 8.91 17 pci 118107 3359 97.64 80.26 20527 6.04 7.14 22 des 448789 8808 99.99 92.3 34512 0.55 0.82 327 vga 700141 17079 98.62 85.99 88428 3.32 3.71 40 Table X. Performance of the Folded Scan Forest with Different Numbers of Scan-In Pins circuits case 1 case 2 case 3 Sin R1 R2 TA Sin R1 R2 TA Sin R1 R2 TA s13207 8 27.81 53....
[...]
12 citations
11 citations
Cites methods from "Test Data Compression Using Selecti..."
...Scan forest was combined with selective encoding (Wang and Chakrabarty 2008) to compress test data of cores, but the new article uses only scan forest to compress test data inside cores. Therefore, due to the use of different DFT architectures, test data and the test delivery time for the new method are more than that in Xiang (2016a). The cores assigned to the 3D stacked NOCs are different from the ones in this article....
[...]
...Scan forest was combined with selective encoding (Wang and Chakrabarty 2008) to compress test data of cores, but the new article uses only scan forest to compress test data inside cores. Therefore, due to the use of different DFT architectures, test data and the test delivery time for the new method are more than that in Xiang (2016a). The cores assigned to the 3D stacked NOCs are different from the ones in this article. In the NOCs with four core classes, three core classes, and a single core class, circuits (b19, des, ethernet, vga), (b19, des, vga), and (b19) are randomly assigned, respectively. In this article, des is replaced by netcard for the first two cases, (vga, netcard), (netcard) are randomly assigned for two core classes and a single core class. The above two reasons contribute to the difference in the core test cost between the new method and the one in Xiang (2016a). The fault-tolerant multicast approach for core testing is not affected much by the number of link/router failures....
[...]
...Scan forest was combined with selective encoding (Wang and Chakrabarty 2008) to compress test data of cores, but the new article uses only scan forest to compress test data inside cores. Therefore, due to the use of different DFT architectures, test data and the test delivery time for the new method are more than that in Xiang (2016a). The cores assigned to the 3D stacked NOCs are different from the ones in this article. In the NOCs with four core classes, three core classes, and a single core class, circuits (b19, des, ethernet, vga), (b19, des, vga), and (b19) are randomly assigned, respectively. In this article, des is replaced by netcard for the first two cases, (vga, netcard), (netcard) are randomly assigned for two core classes and a single core class. The above two reasons contribute to the difference in the core test cost between the new method and the one in Xiang (2016a). The fault-tolerant multicast approach for core testing is not affected much by the number of link/router failures. However, the router testing approach is very sensitive to these failures, which may increases the number of internal router test sessions. As the size of an NOC increases, the new router testing approach becomes less sensitive to the number of failures. The router test cost for the new method is much less than that in Xiang (2016b), because concurrent test delivery and test application for router testing are utilized....
[...]
...Scan forest was combined with selective encoding (Wang and Chakrabarty 2008) to compress test data of cores, but the new article uses only scan forest to compress test data inside cores....
[...]
5 citations
References
178 citations
Additional excerpts
...Another category of compression methods uses statistical coding, variants of run-length coding, dictionary-based coding, and hybrid techniques [17]–[22]....
[...]
175 citations
155 citations
"Test Data Compression Using Selecti..." refers methods in this paper
...The proposed input compression method can be used with recent output compaction methods such as X-compact [2], convolutional compaction [25], and i-compact [26] to further reduce test data volume....
[...]
139 citations
136 citations
"Test Data Compression Using Selecti..." refers methods in this paper
...The proposed input compression method can be used with recent output compaction methods such as X-compact [2], convolutional compaction [25], and i-compact [26] to further reduce test data volume....
[...]