RATS Reliability and Testability Seminar ELF 18 test
RATS (Reliability and Testability Seminar) ELF 18 test status and current results By Intaik Park, RATS, Fall 2005
Outline • • • Introduction Stuck-at tests Transition tests Sequence-dependent defects Future work Intaik Park, RATS, Fall 2005 2
ELF 18 Test Chips • Test chips – Fabricated by Philips – 0. 18µ process – 6 DSP cores (core freq. 20 Mhz) • Test environment – Agilent 93 k tester – Single device cycle (5 Mhz) – Advantek Handler Intaik Park, RATS, Fall 2005 3
Current Test Status • 2 nd round of LOT test finished (10/11/05) • Total: 2, 842 chips tested – 2, 041 good chips – 489 Control Fails – 271 Logic Failures (435 cores) scan chain test fails: 276 cores valid logic fails: 159 cores – 40 Contact Fails Intaik Park, RATS, Fall 2005 4
Tests Applied • Stuck-at Tests – Single Stuck-at test (new conversion flow) – Gate Exhaustive test • Transition Tests – Transition tests (tool A, tool B) – Transition test for Fanout-Free Region (hazard-free and robust transition tests) – Gate super-exhaustive test Intaik Park, RATS, Fall 2005 5
Outline • • • Introduction Stuck-at tests Transition tests Sequence-dependent defects Future work Intaik Park, RATS, Fall 2005 6
Stuck-at tests • Stuck-at test (converted using new flow) – 430 vectors – 98. 53% coverage • Gate Exhaustive Test – 1, 527 vectors – 98. 54% coverage Intaik Park, RATS, Fall 2005 7
New Conversion Flow • Old conversion flow –. vhdl . wgl – Using LTran • New conversion flow Philips ATPG New Flow . PAT Old Flow . WGL –. vhdl . pat – conversion script . SWAV. HP. BINL Tester Intaik Park, RATS, Fall 2005 8
Conversion flow comparison • In. WGL file Idle cycles clk SE 1 1 0 1 0 0 1 1 1 Capture Cycle New flow Old flow Intaik Park, RATS, Fall 2005 9
Stuck-at test escapes • Test Escapes Stuck-at (new flow) Gate Exhaustive 14 17 • 2 cores detected only by Stuck-at test (new flow) (no other test detected) SSF 3 GE 11 6 Escapes Intaik Park, RATS, Fall 2005 10
Outline • • • Introduction Stuck-at tests Transition tests Sequence-dependent defects Future work Intaik Park, RATS, Fall 2005 11
Transition tests Name length cov Test Tr (tool A) 820 98. 22% Transition test by tool A Tr (tool B) 2, 859 99. 65% Transition test by tool B GSE 24, 373 95. 95% Gate Super-Exhaustive GSE tr 19, 504 95. 73% GSE transition on output HF FFR 4, 049 99. 64% Hazard-free tr for FFR RB FFR 4, 671 99. 64% Robust tr for FFR Tr elem 3, 315 99. 65% Tr test elementary gates Hf FFR elem 3, 763 99. 65% HF FFR elementary gates Rb FFR elem 4, 354 99. 62% RB FFR elementary gates Intaik Park, RATS, Fall 2005 12
Transition test for FFR • Restrict transition test to propagate transitions on output of gates • Targets only inputs of Fanout-free Region • Top-off with normal transition test Intaik Park, RATS, Fall 2005 13
Transition test for FFR (cont. ) Str A Normal Tr 01 A X 1 B Stf A Z 01 10 A X 1 B Z X 0 May have no transition or have hazard Hazard-free Tr 01 A X 1 B Robust Tr 01 A 11 B Z 01 10 A 11 B Intaik Park, RATS, Fall 2005 Z 10 14
Transition test for FFR (cont. ) • If output transition is guaranteed, dominance can be used for fanout-free region (same as SSF) • Tests for inputs of FFR tests all the faults inside the FFR Intaik Park, RATS, Fall 2005 15
Transition test for FFR escapes Test Escapes Tr (tool A) 14 Tr (tool B) 14 Tr Elem 13 HF FFR 14 HF FFR Elem 13 RB FFR Elem 12 Intaik Park, RATS, Fall 2005 16
Gate Super-Exhaustive Test • Gate Super-Exhaustive – Extension of gate exhaustive test for transition test – Apply all possible input transition combinations • GSE transition – Subset of GSE patterns producing transitions on output of gates Intaik Park, RATS, Fall 2005 17
GSE test escapes Name Escapes GSE 7 GSE tr 11 GSE 1 Intaik Park, RATS, Fall 2005 GSE tr 6 5 18
GSE tr and HF & RB tr FFR • GSE tr and Robust and hazard-free tr have similar concepts – Restrict on test to produce transition on output • Differences – GSE tr == union of hazard-free tr & robust tr – XOR and MUX • GSE: xor, mux tested exhaustively • HF, RB tr: xor, mux tested non-exhaustively Intaik Park, RATS, Fall 2005 19
GSE tr = Union of HF, RB tr • GSE tr: method 1 AND method 2 • HF tr: method 1 OR method 2 • RB tr: method 2 only A str test 01 A Z 01 01 B 01 A Z 01 11 B method 1 method 2 Intaik Park, RATS, Fall 2005 20
XOR and MUX • GSE tr: method 1 AND method 2 • HF tr: method 1 OR method 2 • RB tr: method 1 OR method 2 A str test 01 A 00 B XOR Z 01 01 A 11 B XOR Z 10 method 2 method 1 Intaik Park, RATS, Fall 2005 21
GSE tr, HF, RB tr escapes Test Escape GSE tr 11 1 Hazard-free Tr 14 10 3 RB Tr HF tr Robust Tr 13 Intaik Park, RATS, Fall 2005 Escapes 22
Test escape summary Test Escapes Stuck-at (new flow) 14 HF tr 14 GE 17 RB tr 13 GSE 7 Tr elem 13 GSE tr 11 HF tr elem 13 Tr (tool A) 14 RB tr elem 12 Tr (tool B) 14 Intaik Park, RATS, Fall 2005 23
Outline • • • Introduction Stuck-at tests Transition tests Sequence-dependent defects Future work Intaik Park, RATS, Fall 2005 24
Sequence-dependent defects • 3 different tests – Test 1 (LOC), Test 2 (Stuck-at), Reference test • if Test 1 fail, Test 2 pass (or vise versa), seq. dep. defect • if ref test fails, invalidate • check all individual vectors Intaik Park, RATS, Fall 2005 25
Test 1 (LOC transition test) Previous state Shift-in Activation vector System clock capture Propagation vector System clock capture Output response Shift-out Intaik Park, RATS, Fall 2005 26
Test 2 (stuck-at test) Previous state 1 bit shifted prop. vector Shift-in Propagation vector System clock capture Output response Shift-out Intaik Park, RATS, Fall 2005 27
Reference test (stuck-at test) Shift-in Activation vector System clock capture Propagation vector (Output response) Intaik Park, RATS, Fall 2005 Shift-out 28
Tester cycle logs • Tester limits 1, 024 failing cycle logs / chip • 6 cores/chip, 184 scan cells, 819 vectors • To log all possible fails, need 904 k cycle logs (882 test rounds) • Assigned 65 cycles for test 1 and test 2 40 cycles for ref. test Intaik Park, RATS, Fall 2005 29
Tester cycle logs (cont. ) • Different failing cycles – Some fail > 65 cycles for one vector – Some fail < 65 cycles for 819 vectors • Iterative method – Apply tests starting from 0, 100, 200 … 700 – If found seq. dep. defect seq. dep. – If finished logging w/o seq. dep. possible seq. indep. Intaik Park, RATS, Fall 2005 30
Sequence-dependent cores • 37 seq. dep. cores – 3 test 1 (loc) fail only – 2 test 2 (stuck-at) fail only – 32 fail mixed • 23 finished logging without seq. dep. – possible seq. indep. cores • 99 need more failing cycle logs Intaik Park, RATS, Fall 2005 31
Outline • • • Introduction Stuck-at tests Transition tests Sequence-dependent defects Future work Intaik Park, RATS, Fall 2005 32
Future work • Apply stuck-at test (using old flow) • New HF & RB tr test (XOR, MUX tested exhaustively) • At-speed transition tests Intaik Park, RATS, Fall 2005 33
Backup slides Intaik Park, RATS, Fall 2005 34
Gate exhaustive test • All the possible input combinations – To each gate • Gate response – To primary output (or scan F/F) From 2004 Fall RATS by Kyoung Youn Cho Intaik Park, RATS, Fall 2005 35
Gate super-exhaustive test • Apply all possible two input combinations – To each gate • Sensitize the effect of the second input – To some observation points From 2005 Summer RATS by Kyoung Youn Cho Intaik Park, RATS, Fall 2005 36
Sequence dependent defect • Defect that makes the output of a combinational circuit depend on the sequence of the input patterns applied • Examples: – Stuck-open fault – Feedback bridging fault From 2005 summer RATS by Intaik Park, RATS, Fall 2005 37
Hazard-free transition test • Transition delay test in which all sensitizing inputs do not change in one of the transitions such that a transition is produced on the output of the gate – AND gate: restrict stf at input – OR gate: restrict str at input Intaik Park, RATS, Fall 2005 38
Robust transition test • Transition delay test in which all sensitizing inputs do not change Intaik Park, RATS, Fall 2005 39
ELF 18 primitives • Total 52, 282 primitives – XOR : 637 – XNOR: 479 – MUX: 3, 324 Intaik Park, RATS, Fall 2005 40
- Slides: 40