Exclusive Test and its Application to Fault Diagnosis Vishwani D. Agrawal Dong Hyun Baik Yong C. Kim Kewal K. Saluja Overview ► Problem Statement ► Introduction Background on Diagnosis Definitions for Diagnosis ► Main Idea Exclusive Test Example of Exclusive Test Exclusive Test Generation Properties of Exclusive Test ► Diagnosis Method ► Results ► Conclusion Agrawal, Baik, Kim and Saluja: VLSI Design 2003 2 Problem Statement ► Obtain high resolution diagnostic test using a single-fault ATPG. Agrawal, Baik, Kim and Saluja: VLSI Design 2003 3 Introduction: Background on Diagnosis ► Single-fault dictionary approaches Simulation based: Chang et al. Fault Diagnosis of Digital Systems, NY, Wiley-Interscience, 1970 Most common method for diagnosis ► Diagnostic test pattern generation: Specialized ATPGs Implication based: Gruning et al. DIATEST: A Fast Diagnostic Test Pattern Generator for Combinational circuits - ICCAD, 1991 Multiple-pass strategy: Savir et al. Testing for, and Distinguishing between Failures - FTCS, 1982 Agrawal, Baik, Kim and Saluja: VLSI Design 2003 4 Introduction: Definitions for Diagnosis ► Consider CUT on the right All 10 faults are detected by 5 test vectors: T1 = 001, T2 = 010, T3 = 011, T4 = 101, T5 = 111 Diagnostic dictionary a g d b T1T2T3T4T5 T6 a1 10100 0 b1 00010 1 c0 00101 0 c1 01010 0 d1 00010 0 f1 00100 0 g0 00001 0 h0 01000 0 i0 01001 0 i1 10110 1 c i s-a-1 s-a-1 e Add T6 = 000 f h ● DR = 10/9=1.11 • Diagnostic Resolution (DR) . ► 10 Faults, but only 9 ● DR = 10/10=1.00 syndromes: DR = No. of faults (classes) ► 10 syndromes: b1 and d1 cannot be of syndromes b1distingushied and d1 areNo. now distinguished test syndrome for fault g0 A measure of quality of diagnosis Agrawal, Baik, Kim and Saluja: VLSI Design 2003 5 Main Idea ► Exclusive test An Input vector that detects only one fault from a pair of targeted faults at a primary output C0 C1 C0 C2 1 C1 C2 1 C0: A fault free circuit C1: CUT with fault f1 C2: CUT with fault f2 Exclusive test vector CUT fault f1 D CUT fault f2 Agrawal, Baik, Kim and Saluja: VLSI Design 2003 6 Example of Exclusive Test ► Application Generate an additional vector to improve diagnostic resolution: distinguish a pair of faults, b1 and d1. ► Example Diagnostic dictionary a T1T2T3T4T5 T6 a1 10100 0 b1 00010 1 c0 00101 0 c1 01010 0 d1 00010 0 f1 00100 0 g0 00001 0 h0 01000 0 i0 01001 0 i1 10110 1 g d b c e Exclusive test vector “abc” T6 = 000 i s-a-1 s-a-1 f CUT fault b1 CUT fault d1 h D Exclusive Test Generation Kim, Agrawal and Saluja “Multiple Faults: Modeling, Simulation and test” VLD 2002 0 a Exclusive test for (b1,d1), T6 = 000 g d 0b i b1 e d f D s-a-1 0 c h g d1 i h e f Agrawal, Baik, Kim and Saluja: VLSI Design 2003 8 Properties of Exclusive Test ► If there exists an exclusive test two faults then they can be distinguished from each other by using that test. ► If no exclusive test exists then the faults cannot be distinguished; two faults form an equivalent fault set. Agrawal, Baik, Kim and Saluja: VLSI Design 2003 9 Diagnosis Method Start with fault detection tests Make dictionary and isolate undiagnosed fault sets Done Yes Is DR satisfactory? No Generate an exclusive test for an undiagnosed fault pair Append the test Yes Test exists? No Form an equiv. Fault set ATPG aborted Agrawal, Baik, Kim and Saluja: VLSI Design 2003 10 Results: Model ► For illustration, an XOR-tree is added to the output of the circuit under test to make it a single output circuit. We use * to denote a modified circuit with a single output XOR-tree at its outputs. ► General later. multiple-PO case is discussed Agrawal, Baik, Kim and Saluja: VLSI Design 2003 11 Test Generation−ISCAS85 Circuits Circuit names c432* c880* # of fault detection tests c1908* c3540* 82 104 176 239 524 942 1879 3428 # of redundant faults 4 5 4 90 # of aborted faults 0 2 27 81 520 935 1848 3257 Fault coverage (%) 99.24 99.26 98.35 95.01 Fault efficiency (%) 100 99.79 98.56 97.57 # of equiv. collapsed faults # of detected faults Agrawal, Baik, Kim and Saluja: VLSI Design 2003 12 Diagnostic Results−ISCAS85 Circuits Circuit names c432* c880* c1908* c3540* # of faults 520 935 1848 3257 # of syndromes 426 789 1450 2706 # of diagnosed faults 354 686 1121 2351 Diagnostic resolution ( DR) 1.22 1.19 1.27 1.20 5 6 8 12 82 104 176 239 # of syndromes 506 870 1579 2844 # of diagnosed faults 492 808 1331 2559 Diagnostic resolution (DR) 1.03 1.07 1.17 1.14 2 3 8 8 126 152 262 328 44 48 86 89 0 0 0 1 14 79 321 662 Max. faults per syndrome # of fault detection vectors Max. faults per syndrome Total test vectors # of exclusive tests added # of equivalent pairs # of aborted pairs Test Generation−c432 Detection tests c432* c432 c432 82 82 82 524 524 524 # of redundant faults 4 4 4 # of aborted faults 0 0 0 # of detected faults 520 520 520 Fault coverage (%) 99.24 99.24 99.24 Fault efficiency (%) 100 100 100 # of fault detection tests # of equiv. collapsed faults Agrawal, Baik, Kim and Saluja: VLSI Design 2003 14 Diagnostic Results−c432 Circuit names c432* c432 c432 # of faults 520 520 520 # of syndromes 428 495 500 # of diagnosed faults 354 471 479 Diagnostic resolution (DR) 1.22 1.05 1.04 5 5 4 # of fault sets 506 507 507 # of syndromes 506 507 507 # of diagnosed faults 492 494 494 Diagnostic resolution (DR) 1.00 1.00 1.00 1 1 1 131 131 123 # of exclusive tests 49 13 41 # of equivalent pairs 14 13 13 Max. faults per syndrome Max. faults per syndrome Total test vectors # of aborted pairs 0 0 0 C432: Simulated using tests derived with c432*, then targeted only undiagnosed faults Conclusion ► Definition of an exclusive test and an ATPG method are introduced. ► A comprehensive exclusive test based diagnostic method is presented where a conventional single fault ATPG can be used. ► Results for ISAS85 benchmark circuits are presented. Agrawal, Baik, Kim and Saluja: VLSI Design 2003 16 Supplement 1: ► Kim, Multiple Fault Model Agrawal and Saluja - “Multiple Faults: Modeling, Simulation and test” - VLSI Design,2002 Convert multiple fault test generation problem into single fault test generation problem. s-a-1 s-a-1 a A a A B b B C c C D d D s-a-1 b c d s-a-0 s-a-0 Multiple (4) stuck-at fault Equivalent Single Stuck-at fault Agrawal, Baik, Kim and Saluja: VLSI Design 2003 18 Supplement 2: Test Generation−ISCAS85 Circuits Circuit names # of detection tests c17* c432* c499* c880* c1355* c1908* c2670* c3540* 6 82 58 104 104 176 236 239 22 524 758 942 1574 1879 2747 3428 # of redundant faults 0 4 0 5 0 4 84 90 # of aborted faults 0 0 32 2 32 27 884 81 # of detected faults 22 520 726 935 1542 1848 1779 3257 Fault coverage (%) 100 99.24 95.78 99.26 97.97 98.35 64.76 95.01 Fault efficiency (%) 100 100 95.78 99.79 97.97 98.56 66.8 97.57 # of equiv. faults Agrawal, Baik, Kim and Saluja: VLSI Design 2003 19 Supplement 3: Diagnostic Results−ISCAS85 Circuits Circuit names c17* c432* c499* c880* c1355* c1908* c2670* c3540* # of faults 22 520 726 935 1542 1848 1779 3257 # of syndromes 14 426 691 789 873 1450 1285 2706 9 354 661 686 360 1121 972 2351 1.57 1.22 1.05 1.19 1.77 1.27 1.38 1.2 4 5 4 6 11 8 11 12 # of diagnosed faults DR Max. faults per syndrome Diagnosis with detection and exclusive tests # of faults 22 520 726 935 1542 1848 1779 3256 # of syndromes 22 506 710 870 902 1579 1385 2844 # of diagnosed faults 22 492 694 808 366 1331 1097 2559 DR 1 1.03 1.02 1.07 1.71 1.17 1.28 1.14 Max. faults per syndrome 1 2 2 3 3 8 11 8 11 126 72 152 129 262 293 328 # of exclusive tests 5 44 14 48 25 86 57 89 # of equivalent pairs 0 0 0 0 0 0 0 1 # of aborted pairs 0 14 16 79 744 321 630 662 Total test vectors