Test specs monitor automotive semiconductors—complete interview

Jeff Bibbee of Pintail Technologies answered our questions about testing automotive ICs.

Greg Reed, Contributing Technical Editor -- Test & Measurement World, 2/1/2005 9:32:00 AM

In the February issue of "Automotive & Aerospace Test Report," we published part of an interview we conducted with Jeff Bibbee of Pintail Technologies.  Here is the complete interview.  Note, in particular, the expanded comments regarding SPC and PAT techniques and Jeff's final thoughts about specifications.

Jeff Bibbee, a founder and the chief technology officer of Pintail Technologies, is responsible for the company’s technology roadmap and strategy. He is a 20-year veteran of the automated test equipment industry with stints at Teradyne, Advantest, and Schlumberger. He holds four patents and graduated with a BSEE from Akron University. I recently asked him about test specifications for semiconductor parts designed for automotive applications.

Q Automotive test specifications on semiconductor parts have been under development for several years. What are some goals of these specifications?

A The Automotive Electronics Council (AEC; www.aecouncil.com) and semiconductor companies have developed specifications targeted at automotive parts to standardize or recommend statistical methodologies, increase overall quality and reliability, and standardize or recommend world-class procedures.

Q What is the intention of AEC-Q001?

A The AEC-Q001 specification recommends a general method for using Part Average Testing (PAT) to remove abnormal parts and thus improve the quality and reliability of parts from a supplier. PAT is a methodology that prescribes finding test results that fall outside six sigma from the population mean for a given wafer, lot, or group of parts being tested. Any test result outside the six-sigma limit for a given device is considered an outlier and removed from the population—parts that fail the PAT limits do not get shipped to a customer, thereby improving quality and reliability.

Q What are the ramifications of real-time SPC or PAT during testing for semiconductors designed for automotive applications?

A Real-time SPC is used to monitor the results of each test.  Thresholds on fail count, Cpk and other statistics can then be used to alert operations or engineering personnel in the event that a wafer or lot of packaged devices is out of statistical control (maverick lot). Using real-time SPC can reduce scrap and re-test costs, while also providing a statistical basis for dynamic test time reduction via sampling. The added benefit of a statistical approach to testing is that it further improves quality and reliability while opening doors for test process improvement and streamlining.

PAT actually lowers the yield by a small percentage since it removes outlier parts that would normally pass the specification limits.  However, the most important metric in the automotive world, defects per million (DPM), can be dramatically reduced using PAT—the goal is zero.  Reducing DPM is not only an automotive goal, however, as the industry is seeking ways to improve overall quality across product lines. Use of PAT on automotive parts appears to be an industry mandate for 2005—semiconductor users like Visteon, Delco, and Renesas are pushing hard on suppliers like TI and ST to deliver parts that meet the AEC specification, so the intense pressure is really coming from the market itself.

The concept of real-time application is important to understand, too. Because all decisions are made in real-time, the impact to the manufacturing process itself is minimal. Parts that meet SPC or PAT metrics continue down the normal manufacturing chain without interruption or alternate steps. The impact to a company’s IT infrastructure is also minimized, since the data from the test process is correct and complete after the end of each wafer or lot. There is no need to retain mountains of data to prove that devices were tested using PAT or a sampling methodology because the real-time process generates comprehensive traceability and lot summary reports. Likewise, the data is not needed for any kind of statistical post processing because the process occurs in real time. Many suppliers do not realize the amount of data required to perform statistical post processing and the requisite impact on their IT budgets, nor do they fully comprehend what generating all that data does to the performance of their existing test equipment.

Q Are there quantifiable failure rate reductions through implementation of semiconductor test specifications?

A Yes.  According to the AEC, many studies have shown that the outlier parts cause most of the field failures.  While devices that meet specification limits are technically “passing” devices, those devices that perform on the fringes of the specification limits are typically where most field returns occur, according to the companies who use chips in their sub-assemblies, powertrain, engine management, and infotainment systems. Automotive parts endure extreme temperature and environmental conditions and the potential for failure increases in these conditions, so there is simply no place for marginally performing devices. As electronic content increases in automobiles, the potential for device failure dramatically increases liability for manufacturers.

Q Any final thoughts about automotive test specifications regarding semiconductors?

A Suppliers are scrambling to meet these specifications, and their users are telling them that they must meet these specifications to continue doing business with them. There is intense pressure to improve reliability and bring the defect rate down, especially now that many extremely important functions like braking, traction control, dynamic or active stability control, and safety measures are being controlled by semiconductors.

Suppliers are trying to minimize the impact of applying these specifications to their yields, as well.

Since manufacturing costs continue to drop and test costs remain relatively flat, the margins on devices are getting smaller as the cost of test becomes a greater component of the cost of manufacturing. Major yield hits simply cannot be tolerated, which means that suppliers must thoroughly evaluate their test process in order to find candidate tests and iteratively refine the candidate list until they zero in on a good target. Without sophisticated analysis and simulation tools, suppliers will be applying these specifications without a good understanding of what it means to their supply chain—or worse yet, applying them blindly and missing critical tests, which would result in shipping at the same DPM rates with the guarantee that the devices were tested using a specification like PAT. In that case, the guarantee is meaningless and reliability suffers.