1. Class Design Project - Test Generation 1 Class Design Project Test Generation Hillary Grimes III ELEC7770 - Project Presentation April 26, 2007

2. Class Design Project - Test Generation 2 Outline  Design For Testability (DFT)  Full Scan Design  Fault Models Stuck-At Faults Transition Delay Faults I DDQ Testing  Results FastScan Multiple Fault Model  Conclusion

3. Class Design Project - Test Generation 3 Design For Testability  After 24 hours, FlexTest only achieved 13.5% stuck-at fault coverage for both designs  Solution: Full Scan Design Possible to control & observe memory elements Simplifies testing & ATPG complexity  All flip-flops replaced with scannable flip- flops:

4. Class Design Project - Test Generation 4 Full Scan Design  Scan elements connected to form scan chain  Basic scan test pattern: Load scan chain Apply primary inputs Measure primary outputs Apply clock Unload scan chain  Disadvantages Area Overhead Performance Overhead

5. Class Design Project - Test Generation 5 Stuck-At Fault Model  Two faults per fault site (gate inputs & outputs) Stuck-At 0 Stuck-At 1  Optimized For Area: Uncollapsed: 69,732 Collapsed: 49,871  Optimized For Delay: Uncollapsed: 70,112 Collapsed: 49,921

6. Class Design Project - Test Generation 6 Transition Delay Fault Model  Two faults per fault site (gate inputs & outputs) Slow-To-Rise Slow-To-Fall  Optimized For Area: Uncollapsed: 69,732 Collapsed: 59,242  Optimized For Delay: Uncollapsed: 70,112 Collapsed: 59,292

7. Class Design Project - Test Generation 7 I DDQ Testing – Pseudo Stuck-At Fault Model  I DDQ –Quiescent I DD Current  Measured Through V DD or V SS  Expensive–current measurement takes much longer than voltage measurement  15 test vectors selected from final test set for I DDQ measurement

8. Class Design Project - Test Generation 8 Results - FastScan  15 I DDQ vectors selected from Stuck-At Test Set Area Optimized: 73.56% Test Coverage Delay Optimized: 74.01% Test Coverage

9. Class Design Project - Test Generation 9 Multiple Fault Model  Generate one test set for both stuck-at & transition faults  Procedure Generate transition delay test set Simulate vectors to find stuck-at fault coverage Add additional vectors to improve stuck-at fault coverage  After pattern compression, 15 vectors selected for I DDQ measurement

10. Class Design Project - Test Generation 10 Results - Multiple Fault Model  15 IDDQ vectors selected Area Optimized: 73.54% Test Coverage Delay Optimized: 73.76% Test Coverage

11. Class Design Project - Test Generation 11 Conclusion  Multiple Fault Model ATPG Reduced the number of compressed vectors for both designs No change in stuck-at fault coverage Reduced transition fault coverage  Area optimized: 96.79% to 94.07%  Delay optimized: 96.20% to 94.22%