Oh, the Way We Were!
Alexander E. Braun, Associate Editor -- Semiconductor International, 12/1/1998
It was 1978; Jimmy Carter was on the second year of his first and only term; VisiCalc staggered the market by demonstrating that those personal computer things might be good for something else besides games, and Semiconductor International made its appearance.
Inspection, measurement and test were done very differently then. Process decisions took place at a much lower level, and the process engineer was regarded as somewhat of an alchemist. Requirements were not as rigid, and although metrology was done, it was carried out differently with what, by today's standards, would be crude equipment. Many of today's familiar systems did not exist.
Perforce, much of today's inspection, test and measurement systems are highly automated, to make the equipment capable of doing everything that is required accurately and in a timely way. Much of the highly developed judgment, knowledge and skill required of the operator of two decades ago now resides in the equipment itself, allowing his successors and process engineers to concentrate on the process itself and not the metrology.
 Film thickness measurements were done by eyeballing the color of the wafer's oxide.
|
Price and complexity were another difference back then. In the late 1970s the most expensive inspection equipment was an optical metrology station that used a simple compound microscope. The typical cost of thin film metrology equipment was well under $20,000. An ellipsometer or reflectometer cost in the neighborhood of $10,000-$15,000. It would not be until well into the 1980s that automated equipment would really take a foothold.
Over the last 20 years not only the nature, but the amount of inspection, measurement and test carried out have changed; that is, the times a wafer encounters a metrology system during its fabrication life cycle. The number has increased by a factor of 30 or more.
Now, as then, fab managers complain that metrology is not a value-added step and would like to eliminate it. In theory, in an ideal world, this is true. Inspection, test and measurement are not value-added steps. Metrology does not really contribute to the wafer's further fabrication. But people on the process line and ultimately those responsible for the fab's success recognize that with good metrology, yields go up. Without it, they would not know until probing the wafer at the end, whether they were making good chips. Without inspection, measurement and test they would have to guess, and it would be a bad guess at that.
Although the processing grail has still to make metrology redundant by becoming so good at it that it is either not required or can be reduced, processing has yet to attain perfection, and its steps are not yet under that level of control. But even if they were, since it would not be apparent when a process goes out of control, test, inspection and measurement would still be necessary, if for no other reason than to confirm that everything is under control.
Some things never change.... 
The author wishes to thank all the battle-scarred semiconductor industry veterans who took time to share their recollections of the way we were.
