The Circle Of Test And EDA Is Complete

Mentor Graphics was founded by ex-tester people and 30+ years later, Mentor releases its first tester.


For those of you who were around and involved with EDA back in the early ’80s, you may remember that chip design was not the focus. It was the board that received most of the attention. Chips were small and did not require much in the way of functional verification. Synthesis had not been invented and so gate-level design was where everything happened, and much of that was manual. The primary focus of the new startups; Daisy Systems Corporation, Mentor Graphics and Valid was on schematic capture, simulation and PCB layout. But they were not the only game in town. There had been other companies who thought that they owned the only piece of EDA worth having – the logic simulator. LASAR, CADAT, Hilo were the three big names of the day and two of them were owned by tester companies. Board level test needed vectors and those vectors had to be optimized. This required fault simulation and automatic test pattern generation in order to find a minimal set of vectors that would detect all of the stuck at faults in the system.

But, if we go back just a few more years, it was ex-Tektronix people who started Mentor Graphics (Gerry Langeler, Tom Bruggere, Dave Moffenbeier). On several occasions, Mentor acquired technology and even land from Tektronix. Hilo (owned by GenRad) went into oblivion, surpassed by Verilog which tackled the newly emerging chip design flow rather than board design, and tester companies also struggled with the new breed of Japanese test companies that were also focusing on chip test rather than board test. The test industry struggled while EDA soared.

EDA technology also reduced the need for large expensive testers by designing in special circuitry, such as scan. This further reduced the market for the tester companies who had become one of the largest expenses in the production of complex chips. Scan compressed the time a chip was on the tester and thus a greater number of chips could be tested using the same amount of equipment.

Given that as a backdrop, I was a little surprised by an announcement from Mentor a short while ago. Mentor now thinks that test is a worthy market for them. Mentor has a tradition of going after markets that other EDA companies haven’t yet considered. A good example of this is embedded software development. This has become a sizable part of Mentor’s business these days and while the connection may seem obvious today, it was a very bold step at the time and few people thought that Mentor would be successful with it.

So, back to test and the Mentor announcement. It all started in the Budapest University of Technology & Economics. In 1997, a group of researchers created a startup called MicRed (Microelectronics Research and Development). Their focus had been on thermal characterization of IC packages, single and arrayed power LEDs, stacked die and other multi-die packages and other high power semiconductor devices. MicRed was acquired by Flomerics Group PLC in 2005, which itself was acquired be Mentor in 2008. MicRed had continued to work in the area of power component testing and diagnosis of failures.

Power devices were once a fairly small market, being confined to things such as locomotive engines, but today, similar power devices are appearing in cars, solar arrays and many other applications where devices could be switching up to 1500Amps. At the same time, companies want 30+ year lifetimes on these devices with reliable operation over 10’s of thousands to millions of cycles. These devices suffer from wire bond degradation, metallization issues, solder fatigue and die and substrate cracking. In the past, a designer of these products would run a device until it failed and then take it to the lab to determine the cause of the failure and then make design improvements.

Mentor now provides a different and automated way of determining a failure before it actually happens. It is called the structure function which shows the heat flowing through a package from junction to ambient. By obtaining the structure function from a known good device, it can then be compared to other devices during power cycling. Deviations identify a specific change in the packaging which indicates what aspect of the device is failing.

So, Mentor now has a tester for power semiconductors that not only finds out if a device is working, but if it has any structural flaws even before it leave the building.

That takes us full circle with Mentor having been formed from ex-tester people and now 30+ years later, Mentor releases their first tester.