`Somewhere between 50 to 100 times more money is spent to create test issues than is spent on solving them.' For about 17 years I have worked for companies specializing in producing automatic test equipment for the semiconductor industry. Earlier, I spent six years writing ATE test programs at what has become the planet's largest semiconductor company. During that time, one thing that has always amazed me is how much money is devoted to R & D for IC fabrication and design relative to how little is spent on the subsequent test issues. Somewhere between 50 to 100 times more money is spent to create test issues than is spent on solving them. Imbalance The practical result of this imbalance is that creating test capability for many new devices takes months instead of days. Although the quality of test has improved somewhat, it has not kept pace with the complexity of the circuits being developed. That being said, the company that I work for (and its competitors) continue to slug it out and produce incrementally better equipment in an industry whose gross revenue fits neatly inside the total R&D budget for the top five semiconductor companies. Somehow, fortunately, this has worked itself out, and the semiconductor revolution hasn't been halted. More work in test, though, could have made the industry move faster and more smoothly. There is a part of the test-related industry that is in greater need of funding than the ATE itself, and that is the component handling industry. Cohu is the giant of the U.S. industry with $208 million in revenues and $20.5 million in R&D expenditures for 1999. Outside of Advantest's handler division, the rest of the players are fairly small and their R&D budgets are smaller. In fact, the gross revenue of many of the handling companies is on par with Cohu's R&D budget. And all of this happened in a pretty good year. We all remember the previous year. During 1998, the handling industry spent less on R&D in absolute dollars. My rough guess is about 40% less than 1999. A rough estimate is that the results of this upside funding will not result in marketable equipment until the beginning of 2001. Most of this equipment will be in answer to the questions that were raised before the current issues of testing many ultra-low mass devices in parallel. Small Players The actual mileage of the handler R&D dollar is probably worse than the amount of dollars spent would indicate. Remember that the industry has many small players in it, which means that the industry budget is split into a lot of small pieces and redundant efforts. What does this likely under-funding of test handling equipment mean to us? Currently, it means that there are only two operating strip or panel handlers on the market that are delivered and operating. A third one will reportedly become available this year. As I have mentioned in past columns, this is a critical capability for economical testing of CSPs. It also means that the maturity level of that equipment will probably be less than what it could have been. And that means that engineers in the field will be plowing a lot of time and effort into integrating these machines with testers and process flows. It's not surprising then that time to market is suffering, as is the ability of the assembly and test outsource industry to expand at a higher rate. So, when you are wandering the halls of any of the SEMICON shows this year, look at the companies displaying IC fabrication, ATE and handling equipment. See what you can tell about the amount of R&D spending at each of them. It should then become rather obvious where the next production challenge will be.
|
