Lecture 19alt IDDQ Testing (Alternative for Lectures 21 and 22)Basic Principle of IDDQ TestingNAND Open Circuit Defect – Floating gateFloating Gate DefectsDelay FaultsWeak FaultsWeak Fault DetectionLeakage FaultLeakage Fault TableIDDQ Vector SelectionHP and Sandia Lab DataFailure Distribution in Hewlett-Packard ChipSematech StudySematech ResultsSematech Conclusions% Functional Failures After 100 Hours Life TestCurrent Limit SettingDifference in HistogramsDelta IDDQ Testing (Thibeault)|IDDQ| and |DIDDQ|Setting ThresholdIDDQ Built-in Current Testing – Maly and NighConceptual BIC SensorSummaryCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 1Lecture 19altIDDQ Testing(Alternative for Lectures 21 and 22)Lecture 19altIDDQ Testing(Alternative for Lectures 21 and 22) Definition Faults detected by IDDQ tests Weak fault Leakage fault Sematech and other studies Delta IDDQ testing Built-in current (BIC) sensor SummaryCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 2Basic Principle of IDDQ TestingBasic Principle of IDDQ TestingMeasure IDDQ current through Vss busCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 3NAND Open Circuit Defect – Floating gateNAND Open Circuit Defect – Floating gateCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 4Floating Gate DefectsFloating Gate DefectsSmall break in logic gate inputs (100 – 200 Angstroms) lets wires couple by electron tunnelingDelay fault and IDDQ faultLarge open results in stuck-at fault – not detectable by IDDQ testIf Vtn < Vfn < VDD - | Vtp | then detectable by IDDQ testCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 5Delay FaultsDelay FaultsMany random CMOS defects cause a timing delay fault, not catastrophic failureSome delay faults detected by IDDQ test – late switching of logic gates keeps IDDQ elevatedDelay faults not detected by IDDQ testResistive via fault in interconnectIncreased transistor threshold voltage faultCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 6Weak FaultsWeak FaultsnFET passes logic 1 as 5 V – VtnpFET passes logic 0 as 0 V + |Vtp|Weak fault – one device in C-switch does not turn onCauses logic value degradation in C-switchCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 7Weak Fault Detection Weak Fault Detection Fault not detected unless I3 = 1Copyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 8Leakage FaultLeakage FaultLeakage between bulk (B), gate (G), source (S) and drain (D)Leakage fault table for an MOS component:k = number of component I/O pinsn = number of component transistors m = 2k (number of I/O combinations)m x n matrix M represents the tableEach I/O combination is a matrix rowEntry mi j = octal leakage fault information:Flags fBG fBD fBS fSD fGD fGSSub-entry mi j = 1 if leakage fault detectedCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 9Leakage Fault TableLeakage Fault TableCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 10IDDQ Vector SelectionIDDQ Vector SelectionCharacterize each logic component using switch-level simulation – relate input/output logic values & internal states to: leakage fault detectionweak fault sensitization and propagationStore information in leakage and weak fault tablesGenerate complete stuck-at fault testsLogic simulate stuck-at fault tests – use tables to find faults detected by each vector to select vectors for current measurementCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 11Com-panyHPSan-diaReject ratio(%)Without IDDQWith IDDQWithout IDDQWith IDDQNeither16.460.80OnlyFunct.6.360.09OnlyScan6.040.11Both5.800.00Functional Tests5.5620Scan and Functional TestsHP and Sandia Lab DataHP and Sandia Lab DataHP – static CMOS standard cell, 8577 gates, 436 FFSandia Laboratories – 5000 static RAM testsReject ratio (%) for various tests:Copyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 12Failure Distribution in Hewlett-Packard ChipFailure Distribution in Hewlett-Packard ChipCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 13Sematech StudySematech StudyIBM Graphics controller chip – CMOS ASIC, 166,000 standard cells0.8 m static CMOS, 0.45 m Lines (Lef), 40 to 50 MHz Clock, 3 metal layers, 2 clocksFull boundary scan on chipTests:Scan flush – 25 ns latch-to-latch delay test99.7 % scan-based stuck-at faults (slow 400 ns rate)52 % SAF coverage functional tests (manually created)90 % transition delay fault coverage tests96 % pseudo-stuck-at fault cov. IDDQ TestsCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 14Sematech ResultsSematech ResultsTest process: Wafer Test Package Test Burn-In & Retest Characterize & Failure AnalysisData for devices failing some, but not all, tests.passpassfailfailpass14652passpass60136failfail146334131251passfail718failpassfailpassfailScan-based Stuck-atIDDQ (5 A limit)FunctionalScan-based delayCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 15Sematech ConclusionsSematech ConclusionsHard to find point differentiating good and bad devices for IDDQ & delay testsHigh # passed functional test, failed all othersHigh # passed all tests, failed IDDQ > 5 ALarge # passed stuck-at and functional testsFailed delay & IDDQ testsLarge # failed stuck-at & delay testsPassed IDDQ & functional testsDelay test caught failures in chips at higher temperature burn-in – chips passed at lower temperatureCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 16% Functional Failures After 100 Hours Life Test% Functional Failures After 100 Hours Life TestWork of McEuen at Ford MicroelectronicsCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 17Current Limit SettingCurrent Limit SettingShould try to get it < 1 AHistogram for 32 bit microprocessorCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 18Difference in HistogramsDifference in HistogramsA – test escapes, B – yield lossCopyright 2005, Agrawal & BushnellVLSI Test: Lecture 19alt 19Delta IDDQ Testing (Thibeault)Delta IDDQ Testing (Thibeault)Use derivative of IDDQ at test vector i as current signatureΔIDDQ (i) = IDDQ (i) – IDDQ (i – 1)Leads to a narrower histogramEliminates variation between chips and between wafersSelect decision threshold Δdef to minimize probability of false
View Full Document