View: session overviewtalk overview
09:00 | An effective methodology for automated diagnosis of functional pattern failures to support silicon debug ABSTRACT. We present an industrial-grade solution for automated diagnosis of functional patterns failures. The solution has been successfully demonstrated and deployed on multiple products at Intel. |
09:30 | Improving Diagnosis Resolution and Performance at High Compression Ratios SPEAKER: Sameer Chillarige ABSTRACT. Techniques to maintain high diagnosis resolution are proposed based on a study conducted on multiple customer designs by varying compression ratios up to 400X. |
10:00 | SPEAKER: I-De Huang ABSTRACT. In this paper, we present a novel approach to run scan diagnosis either on production testers directly or on a few compute machines in a high volume manufacturing (HVM) environment. |
09:00 | Polynomial Chaos modeling for Jitter estimation in high-speed links SPEAKER: Majid Ahadi Dolatsara ABSTRACT. Traditional approaches of data dependent jitter estimation in nonlinear high-speed links can be extremely time consuming. Therefore, this paper proposes developing a surrogate model by using the Polynomial Chaos theory to expedite this process. |
09:30 | Self-Learning Health-Status Analysis for a Core Router System SPEAKER: Krishnendu Chakrabarty ABSTRACT. Monitoring core routers is essential to ensure high reliability. However, most operational data are unlabeled, necessitating the design of a self-learning framework that iteratively learns and identifies various health status. |
10:00 | A Stressed Eye Testing Module for Production Test of 30-Gbps NRZ Signal Interfaces SPEAKER: Kiyotaka Ichiyama ABSTRACT. This paper introduces a stressed eye testing module for 30-Gbps NRZ signals. It can inject calibrated random jitter, sinusoidal jitter, and sinusoidal interference. Calibration procedures for the module are also presented. |
09:00 | XLBIST: X-Tolerant Logic BIST SPEAKER: Peter Wohl ABSTRACT. We present a high-coverage, X-tolerant LBIST solution which uses compressor/decompressor structures, including X-control logic, that have already been inserted in the design for scan-compression deterministic patterns. |
09:30 | *Distinguished Paper: Deterministic Stellar BIST for In-System Automotive Test SPEAKER: Yingdi Liu ABSTRACT. The paper presents a compression scheme for in-system automotive test. It generates vectors by complementing scan slices of encodable patterns. The scheme provides significant trade-offs between area and time. |
10:00 | Improving Analog Functional Safety Using Data-Driven Anomaly Detection SPEAKER: Fei Su ABSTRACT. We propose a machine learning method using data-driven anomaly detection for functional safety of analog automotive circuits, with mining dynamic time series in-field data in the context of system operation. |
09:00 | AI Engineering Assistants for ATE SPEAKER: Keith Schaub ABSTRACT. It’s common practice today to have Alexa turn on a light or play music and get traffic information from Siri. These simple tasks are only just the beginning. Machine learning and Deep learning have enabled dramatic progress with vision recognition, and NLP. Industries can integrate these technologies into their existing business process and products and over time, enhance them with their own custom IP. 2019 will be a year of development for the AI assistant. In this paper, we outline an AI engineering assistance model with some early prototypes and use cases that could be applied to test systems, which we believe will not only make for a richer experience, it will improve engineering and production efficiency as well as enhance production floor security. By integrating both vision and NLP AI with products and software interfaces enabling simple voice and gesture commands, engineers would be able to tackle more complex engineering tasks. A first step could be enabling engineers to speak the most commonly performed commands as they interact with the test system thus freeing them up to focus on the task at hand. Vision recognition AI could be used to enhance production floor login security, which would be quite valuable during post yield analysis. As the AI learns and improves, it can begin assisting with test debug, content searches, and even test program development. |
09:45 | Is it possible to impact quality of test with machine learning? SPEAKER: Marc Hutner ABSTRACT. Data Analysis is an important tool for Test and Product Engineers to improve coverage, product quality and operational efficiency of the equipment. Automated test equipment generates a vast amount of data with respect to the part being tested and the system doing the testing. In parallel to the test industry, many other technology sectors have been adopting various forms of Artificial Intelligence or Machine Learning to analyze vast data sets for new insights. Is it possible to leverage these new techniques back into a test application? In this presentation we will discuss infrastructure for impacting quality of test and how machine learning fits into the framework. |
10:15 | Moving Adaptive Test to “AI Test” ABSTRACT. For years the industry has been focused on the use of Adaptive Test techniques to streamline and focus our test efforts for maximum value (and minimum test times). With the advent of Neural Network techniques (i.e. AI) new possibilities are coming to light for focusing our vision on areas where improvements can provide value. This presentation provides an overview of Adaptive Test and Neural Network techniques and then shares a vision for merging the two techniques together in order to improve our device quality, reduce our cost of test, and automate the control of functions best left to the computers supporting us. |
Instead of a regular keynote, this plenary session comprises several elevator talks to provide different perspectives on AI in Test. A follow-up panel will be held in the afternoon for further discussion.
The speakers include:
Ken Butler, Founder, Engineering Tools and Analytics Team, Texas Instruments, Dallas, Texas - A Semiconductor Test Perspective
As we all know, semiconductor manufacturing is a complex process with many interacting components, any one of which can negatively impact quality and cost. AI techniques have already shown great promise as a means to detect, diagnose, and correct for these issues quickly in volume production. We will examine what is in use today and aspects that are being investigated for future deployment.
Anne E Gattiker, Principal Research Staff Member, IBM - Deep Learning perspective
Recently Deep Learning, e.g. employing many-layered neural networks, has revolutionized fields such as Computer Vision and Natural Language Processing. What factors have lead to such remarkable advances in these fields and could the test field achieve similar gains?
Ira Leventhal, Vice President, New Concept Product Initiative, Advantest America, Inc. - ATE perspective
I will discuss how Machine Learning, while showing great promise to solve challenging test-related problems, is not a one-size-fits-all solution. Separating the reality from the hype on how this technology can be successfully applied requires a solid understanding of the strengths/weaknesses of machine learning algorithms, and which applications are the best fit to the strengths. I’ll present real-world successes with applying AI-based semiconductor data analytics and show how these algorithms can outperform other approaches, when applied to the right class of problems.
Xinli Gu, Huawei - AI from a system builder perspective
AI/ML (Machine Learning) is a tool that promotes formal data collection and value discovery. In a large telecom company with thousands of products, end-to-end data collection and analysis using ML technology is proven to be huge values and gradually becomes a “MUST” in the company process. This talk will use an example to demonstrate the framework in one of the applications.
Cheng-Wen Wu, Tsing Hua Distinguished Chair Professor with the EE Dept., NTHU, Hsinchu, Taiwan. - AI on Taiwan Semiconductor Industry
The financial tsunami a decade ago had somewhat slowed down the global semiconductor business, until late 2016 when AI suddenly gave everybody new hope. In the same period, however, the semiconductor business in Taiwan continued its growing trend, though on a slightly different track as the industry had expected. Like a chameleon, sort of, the industry has to quickly adjust itself to survive in the global market trend of cloud, IOT, and AI. In my speech, I will give my perspective on the strengths and surviving skills of the semiconductor industry in Taiwan, and give the outlook of the industry in the AI age.
13:30 | Defect Injection, Fault Modeling and Test Algorithm Generation Methodology for STT-MRAM SPEAKER: Mehdi Tahoori |
14:15 | Testing Resistive Memories: Where we are standing and what is still missing? SPEAKER: Said Hamdioui |
Remember the 1968 film "2001: A Space Odyssey?" Did you know ITC was born in 1969? As the film industry is celebrating the 50-year-old birthday of "2001: A Space Odyssey," ITC will be entering its 50th year in 2019. What has changed in the last 50 years in ITC and in our industry? Moving forward, what shall we expect? How ITC and the test community as a whole will change in the next 25 years?
Do you remember the fun in the past ITC? Will the machine HAL become the general chair for ITC 75?
Check out the MP4 promotion video for AI program this year: http://www.itctestweek.org/AI/
Let's bring back our fun memories and use a little bit imagination in this panel.
The 2019 ITC-50 committee will be there to hear from you!
13:30 | Modeling and Testing of Aging Faults in FinFET Memories for Automotive Applications SPEAKER: Grigor Tshagharyan |
14:15 | ECC-Based FIT Rate Mitigation Technique for Automotive SoCs SPEAKER: Gurgen Harutyunyan |
Following the AI mini-keynotes in the morning, this afternoon panel provides a forum to have more in-depth discussion.
AI has become an inevitable force to affect many industries. How do we see AI might transforms our industry? The most extreme question perhaps is: Will AI eventually eliminate test engineering? If not 100%, to what extent and when?
There are many questions surrounding AI and most importantly, do we really understand what AI is?
After hearing different perspectives in the morning, come and share your thoughts in this panel.