Design and test: Getting a handle in productivity
03/01/1998
Design and test: Getting a handle on productivity
Richard Howard, Lucent Technologies,
Murray Hill, New Jersey
Since the 1994 Roadmap, there has been a further realization of how fast the gap between what we can design and what we can build is growing.
Everyone agrees that the cost of testing is exploding. Testing, testers, and design are becoming a major cost of manufacturing a product. In all the other areas of the Roadmap, there are good metrics and benchmarks for performance and goals. In the design area, however, the metrics and benchmarks are still rudimentary. Only now are we beginning to get a handle on how to measure design productivity in a useful way so we can improve it. In lithography, everyone knows the line rules, yields, and depth of focus. But the design area is more complex; many companies are not as willing to give out design costs or design time estimates. So, this part of the Roadmap is still looking for improvements. It is not as quantitative as the processing area, even though it is becoming a greater fraction of the cost.
Since 1994, there has also been a growing emphasis on analog, mixed signal, and RF issues. The original Roadmap was essentially only a digital Roadmap. With the growth in wireless and systems- on-a-chip, a large fraction of chips are devoted on some level to analog and RF devices. As clock frequencies approach 1 GHz, they contain frequencies higher than those used in cellular radio systems. The design issues are more analog and less digital at levels we have not encountered before.
In comparing design to the processing area, many innovations and improvements in design can be added at any point in the timeline. When putting together finer line rule technology, all the steps that go into the IC process - etching, lithography, film deposition, etc. - are required before achieving the next generation. In design, a collection of tools, techniques, and methodologies come in from many directions, each one improving the overall efficiency of the process. This year with the Roadmap, we have tried to be more efficient in setting priorities for the different design and test paths. They are all needed - a more efficient layout tool, a more efficient high-level description language, etc. - and can all be used immediately. We have put more effort this time into understanding which ones will have the biggest impact on the business if they come the soonest. This is probably the biggest difference between the 1997 and 1994 Roadmaps.
All in all, the goal of test is to put test out of business. Manufacturers would like to make the IC perfect the first time, and omit testing, which is a major part of the cost. Testing does not add value; it just makes sure that the IC is functioning. We would like to be able to design the IC so it can be manufactured as reliably as possible.
RICHARD HOWARD received his BSC degree in physics from Cal Tech and his PhD degree in applied physics from Stanford University. In 1978, he joined Bell Laboratories, where he is now the director of the Wireless Research Lab. Lucent Technologies, 700 Mountain Ave., Murray Hill, NJ 07974; e-mail [email protected].