Moving from static to dynamic methods. Front end was the target and now it is the back end.
We collect tons of data and have however know one uses this data to date. (wordsmith)Phil Nigh of IBM. TI, Qualcomm etc. are participating.-Statistical Data AnalysisFeed data forward and Feed backwards Elaborate on the fly (real time)Surprised that the ITRS working group has defined essentially what OT is delivering to the market. (e.g., Parts Average Testing algorithms) Analysis driving Adaptive Test will occur both real-time (in parallel with testing), near-time (at the end of sample testing and at the end of wafer test and lot test) and off-line. “near-time” (e.g., end of wafer) Show circle speaking about forward and backward.
Data feed forward and feed back from manufacturing. Not only for semiconductor test. Feeding data forward and backwards! Potential to impact what we learn at test.We have customers today who are aggressiveAutomotive and medical not only semiconductor test.So, we’ve already introduced the phrase Adaptive Test… and I want to take just a minute to talk about what that really is because – if you were at Semicon West last week – you have heard a lot of “buzz” about Adaptive Test. And, in fact, it’s a topic that’s getting a lot of air time just now.In July, 2008– at Semicon West – the ITRS Test Working Group held a meeting and identified Adaptive Test as an area of specific focus for 2009 and 2010… so, there’s a lot of work going on just now. And Optimal Test is working with the ITRS on definitions. In fact this graphic is borrowed from the ITRS working group.So, what is adaptive test? According to the current state of the definition the Test Working Group has adopted (and it is subject to change)… Adaptive Test is a broad term used to describe methods that change test conditions, test flow, test content and test limits (potentially at the die/unit or sub-die level) based on manufacturing data and statistical data analysis. This includes feed-forward data from inline and early test steps to later test steps and feed-back of data from post-test statistical analysis that is used to optimize testing of future products. Adaptive Test also includes real-time data analysis that can perform Statistical Process Control (SPC) and adjust test limits and content during product testing on-the-fly. Although some simple applications have been applied for some time, Adaptive Test will increasingly be applied and will require updated software algorithms and improved statistical analysis methods and expanded database infrastructure.Note that last piece very carefully… because what is implied by the ITRS’ definition is a reasonably robust IT infrastructure that will support this “data feed forward” and “data feed backward” across a disaggregated supply chain or a geographically dispersed integrated device manufacturer.We think it’s also important to recognize that the “flow” does not stop with traditional functional ATE but rather extends to eTest, burn-in, System Level Test, Card or Board test, and even to returns from field operations… AND that there is an analysis and optimization loop at each stage. So, this is a highly dynamic environment, requiring a rich set of IT tools to succeed. So, how do OptimalTest solutions address this vision? <click>
Bullet 2 – Additional logging without impacting TTR based on adaptive input learned.Bullet 3- can decrease and increase test. Apply only the needed tests, Identifying. Can augment test. Reference Dice.
Security – SFTP (mention security)This model is completely applicable across a global IDM and disaggregated fabless supply chain.OTDF – description. We can capture information Is OTDF public format, YES.70-90% smaller.Integrity, imposes a standard data structureEncapsulation of the impact of adaptive test. Captures any changes.As you apply changes to Adaptive Test, you change device test.
Take always – end to end solution. We are part of the TAG partnership program.
Tug Ot Prez 2010 050510
Advanced Adaptive Test – TTR for IGXL Platforms<br />Lisa Vallerie, OptimalTest<br />Lisa.Vallerie@OptimalTest.com<br />Itai BenJacob, OptimalTestItai.BenJacob@OptimalTest.com<br />
Agenda<br />Introduction to Advanced Adaptive Test for Test Time Reduction<br />Adaptive Test Time Reduction Methodologies<br />Advantages on Teradyne FLEX IGXL platform<br />2<br />
Evolution of Test Optimization <br />Static Test Optimization requires as much as months based on off-line analysis and is driven manually with point solution tools – or without tools altogether<br />Dynamic Test Optimization is needed for overall manufacturing test efficiency to improve IC production and to increase yield learning<br />Adaptive Test Optimization is a dynamic method of managing process variations while eliminating redundancies in real- or near- realtime; both statistical and automated process control methodologies are used<br />4<br />
Adaptive Test ITRS Definition<br />Adaptive Test is a broad term used to describe methods that change test conditions, test flow, test content and test limits (potentially at the die/unit or sub-die level) based on manufacturing test data and statistical data analysis<br />This includes feed-forward data from in-line test and early test steps to later test steps and feed-back data from post-test statistical analysis that is used to optimize testing of future products<br />Adaptive Test includes realtime & near-realtime data analysis that can perform Statistical Process Control (SPC) and adjust test limits and content during production testing on-the-fly<br />5<br />
Benefitsof Advanced Adaptive Test<br />Lower test costs – Decreased test times due to algorithmic test automation vs. legacy solutions<br />Higher yields – Early detection of yield degradation accelerates root cause isolation and yield improvement on future devices.<br />Better Quality & Reliability – Identification of outlier devices that result in infant mortality, improved test and device quality through enhanced process control<br />7<br />
Advanced Adaptive TTR Methodologiesfor Semiconductor Devices<br />OptimalTest has 2 Advanced Adaptive TTR methodologies<br /><ul><li>Parametric Method – performed on parametric tests while analyzing & evaluating, in real-time, the process stability and making test time reduction decisions
P/F Method – performed on low fallout tests</li></ul>Leverage information from actual & historic device test results<br />Applicable at WS and FT<br />Selection is based on device & test characteristics and needs<br />Can be applied simultaneously or individually<br />Patents pending & issued<br />8<br />
Process Flow for Test Time Reduction <br /> High Level Process Flow:<br />Analyze historical test results to identify “test candidates”<br /><ul><li>Provides the expected test time reduction benefits vs. the Yield/PPM impact</li></ul>Creation of Adaptive TTR recipe<br /><ul><li>Selected tests, number of validation units, sampling rate, etc..</li></ul>Creation of TTR rule and activation at SAT<br />Actual results of execution can be reviewed and analyzed using OTPortal Business Intelligence tool<br /><ul><li>Supports feedback and continuous learning</li></ul>AnalysisHistorical Data<br />Test Candidates<br />TTR Recipe<br />Adaptive TTR Rule<br />TTR Rule in Real Time<br />TTR Results<br />OTPortal<br />9<br />
Parametric TTR Algorithm Flow in RealTime<br />10<br />
Adaptive Parametric TTR Algorithm Terminology<br />“Predicted Test Ranges” – calculated based on the actual parametric test measurement results. TTR is only enabled when within the Spec Limits. <br />“Predicted Test Ranges” are (re)-evaluated after “Validation Units” and “Sampling Units” are tested, using the data from these units. Provide a “safety margin” for product quality.<br />OT’s method imposes a “Safety Coefficient” used in the calculation of the ranges to insure the maximum safety of the TTR process. Customer specifies acceptable DPPM<br />11<br />
Adaptive Parametric TTR Algorithm Terminology<br />Algorithms’ parameters are user according to the desirable TTR level and DPPM “risk”<br />User Configurable Variables are:<br />Validation Unit Quantity<br />Sampling Unit Rate<br />Safety Coefficient (acceptable DPPM)<br />Quantity of units used to evaluate the optimal Predicted Test Ranges<br />12<br />
Customer’s Upper Spec Limit<br />OT’s Upper Predicted Test Range<br />Actual Parametric Test Results<br />(normal lot)<br />Test Value<br />Test Value<br />OT’s Lower Predicted Test Range<br />Customer’s Lower Spec Limit<br />Touch Down Sequence<br />Parametric TTR Simulation 1TTR enabled across an entire lot<br />13<br />
Parametric TTR Simulation 2TTR disabled as measurements cross a threshold (lower spec limit)<br />OT’s Upper Predicted Test Range<br />Actual device “failures” <br />Occur after TTR is disabled<br />Test Value<br />Customer’s Lower Spec Limit<br />TTR is dynamically disabled<br />OT’s Dynamically Calculated<br />Lower Predicted Test Range<br />falls below the Customer’s <br />Lower Spec Limit<br />14<br />
Parametric TTR Simulation 3TTR is not enabled when Validation Unitsfail to achieve Predicted Test Range<br />OT’s Upper Predicted Test Range<br />Actual device “failures” <br />occur after TTR disable decision<br />Test Value<br />Customer’s Lower Spec Limit<br />Validation Units<br />TTR is not enabled for <br />entire lot<br />OT’s Dynamically Calculated<br />Lower Predicted Test Range<br />First Validation Unit falls below<br />Customer’s Lower Spec Limit<br />15<br />
Adaptive Pass / Fail TTR Algorithm Methodology<br />Adaptive Pass / Fail TTR algorithm leverages Pass / Fail data in order to decide, in real-time, whether the “candidate test” can be turned off for Test Time Reduction<br />Employs the same mechanism of “Validation Units” and “Sampling Rate” to insure the user’s confidence with diminished test suite(s)<br />16<br />
In FT, the process is straight forward and is implemented as depicted below:<br />For WS, additional dedicated techniques / algorithms support the special characteristics of Wafer/Sort Test<br />Implementation of Adaptive TTR in WS and FT<br />17<br />
Implementation of Adaptive TTR in WS and FT<br />Reference Die – “the health of the wafer”:<br />Die strategically selected according to various attributes<br />Always tested with the full test-program (not TTR “eligible”)<br />Results used by Adaptive TTR algorithm in real-time to enable / disable TTR on subsequent die of the wafer<br />Applicable at Wafer Sort and Final Test (ULT required at FT)<br />Validation Wafers – “the health of the lot”:<br />Few sample wafers/lot<br />Always tested with the full test-program (not TTR “eligible”)<br />Results used by Adaptive TTR algorithm in real-time based on to enable / disable TTR on the rest of the wafers in the lot<br />Reference Die & Validation Wafers used for “Quality Control” <br />18<br />
Locations of Reference/Baseline Die<br />are selected according to <br />various algorithms:<br />Next to E-test structures (for maximized correlation between test sockets)<br />Spread-out equally in each of the 3 ring areas (for maximized coverage)<br />In areas of different yield signatures <br />In most of the lithography exposure locations<br />In areas corresponding with Fabdefect sampled areas<br />Implementation of Adaptive TTR Reference/Baseline Die<br />19<br />
Adaptive TTR on IG-XL TestersTheory of Operation<br />OptimalTest utilizes a compact code, “OTProxy” that can be installed on IGXL testers<br />OTProxyuses a COM component (ActiveXTTRDLL) which is a VB6 code handling the Excel API in order to perform TTR (OTProxyuses this COM object to toggle test-flow Enable Words)<br />This COM object is being registered to the system automatically as part of the OTProxy installation<br />OTProxy’sactions are based on IG-XL Test Instances<br />Running TTR on IG-XL requires a unique Test ID per test<br />20<br />
OTProxy Server<br />TTR Actions<br />IG-XL Tester<br />OTProxyEngine<br />Test Results<br />TTR COM Object<br />Enable Words and Macro Injection Component<br />OTDF Datalog<br />Test Mapping<br />TTR Actions<br />Skip Test List<br />Adaptive TTR on IG-XL Tester Block Diagram<br />22<br />
IG-XL Test Program Manipulationin Real Time<br />Before injection of Enable Words<br />After injection of Enable Words<br />23<br />
Teradyne & OptimalTest Collaboration<br />In 2009 Optimal Test and Teradyne announced a strategic alliance<br />Teradyne’s Sales Force has been trained & has sales tools for use<br />Ongoing technical exchanges with both R&D departments regarding Advanced Adaptive Test Techniques on the IGXL & Image platforms <br />Business level exchange meetings on a monthly basis to leverage efforts to benefit key accounts<br />Collaborating with Teradyne to Beta test our new test management solution ideas & to maintain roadmap coherence<br />A key Teradyne fabless customer is currently deploying an end-to-end Advanced Adaptive Test solution across the supply chain<br />24<br />
Increase Production Yield<br />Improve Production Yield<br />Speed Time to Entitled Yield<br />Increase Product <br />Reliability<br />Optimize Overall <br />Equipment Efficiency<br />1-4%<br />Increase Product Reliability<br />Through Outlier Detection<br />Optimize Overall Equipment Efficiency (OEE)<br />Speed Time to Actionable Data<br />20-50%<br />10-20%<br />MEASURABLE<br />RESULTS<br />Improve Product &<br />Testing Quality<br />Reduce Test Times<br />Advanced Adaptive Test for TTR and/or<br />Improve Capital Utilization<br />Increase Product & Testing Quality<br />Reduce Customer Returns<br />10-30%<br />50-75%<br />25<br />
Implementation of Adaptive TTR in WS and FT<br />Reference Die are tested first to achieve a measure of <br />overall wafer “Health”<br />OTBox: reference die are tested before other die on the wafer<br />OTProxy:a dedicated probing sequence is used<br />Reference (or Baseline) Die<br />Reference die are covered by US Patent<br />29<br />
Adaptive Parametric TTR Algorithm Terminology<br />“Validation Units” – the first “n” units of the lot / wafer These units are fully tested, TTR is not performed<br />Typically “n” = 200-300<br />“Sampling Units” – after “Validation Units” are tested and TTR is enabled, one out of “x” units is being sampled Sampling Units are fully tested<br />Typically “x” = 10-20 <br />30<br />
Adaptive TTR on IG-XL TestersTheory of Operation<br />Test Mapping – Accomplished online via VB application which creates the test mapping file for the OTProxy<br />Datalog (OTDF) triggers the test mapping tool automatically at the beginning of the lot / wafer<br />Enable Words and macro injection are done online by the VB ActiveX DLL<br />OT’s Enable Words injection method doesn’t overwrite the existing user defined Enable Words -- no change to test programs<br />Those 2 operations are done at the beginning of the lot / wafer, after the Test-Program is loaded to the tester memory<br />31<br />