Thoughts on Costs, Metrology and ATE

When dinosaurs wandered the earth and I covered the defense industry, there were concerns that, at the rate costs were climbing for aircraft avionics and countermeasures, the entire U.S. gross domestic product would be insufficient to pay for just one fighter by the year 2000.

I was reminded of this at a SEMICON West press conference announcing the formation of the Semiconductor Test Consortium (STC). STC seeks to have the semiconductor test industry sign off on several ATE architectural attributes sought by Intel Corp. for the testing of logic chips, particularly system-on-a-chip (SoC), to keep platform costs down.

At the press conference, Intel argued that, as semiconductor technology progresses, chipmakers cannot continue buying costly platforms useful for only a few technology nodes before having to be replaced by even costlier ones. It was pointed out that platforms sometimes do not talk to one another — even those from the same OEM. STC's initiative states, "To make fundamental improvements in cost, competitive capability, platform support, and performance scaling, component ATE needs to move from proprietary platform architecture to modular solutions that scale and encourage multiple supplier participation." As of this writing, OEMs Advantest (instrumental in STC's creation), National Instruments, Racal Instruments, Wavecrest, and GuideTech have joined STC.

The single-backplane concept has a scalability capable of supporting ATE improvements for more than a decade, to which any OEM product can be added with everything talking to everything else in perfect harmony. Upgrades only require that a module be slipped out and another slipped in. This is attractive. However, it appears to ignore realities such as R&D costs, intellectual property, and return on investment.

STC's viewpoint seems to parallel that of manufacturers that want integrated metrology (IM) but complain that it — and stand-alone platforms — are too expensive. For IM to function the way they want it to, it would have to be almost free and never break down.

An obstacle to inexpensive metrology is that user requirements change constantly: They want to inspect increasingly complex structures and shapes, sidewall liners, etc. These capabilities do not come without effort (read "cost") on an OEM's part. No stand-alone platform provides a 2000 hour MTBF, but those demanding IM want sophisticated technology squeezed into a small box to go into a process tool and provide a 10,000 hour MTBF. This might be attainable with a simple product, but at the cost of functionality. The simplicity vs. functionality trade-off is unavoidable — both for IM and the ATE that STC is focusing on — requiring, in the end, backup stand-alone tools that increase metrology and ATE investments.

Metrology equipment is getting increasingly costly. However, OEMs cannot meet the demand for a $100,000 metrology box if it has $75,000 to $100,000 worth of parts inside. IM, for example, requires careful development and a reduction of moving parts. An ellipsometer, for instance, requires that the optics be folded for it to work. Reliability — lacking robotics — can be three times that of a stand-alone platform. It is not surprising that, by the time the overhead (R&D, support, etc.) is factored in, the solution cannot be provided for under $300,000. Should IM be installed in every track system and etcher? This expense for such a higher-end IM module provides a payback only in specific instances (metrology for logic devices, for example). Here, IM systems in the etcher can measure every wafer that comes through, do a feed-forward gate etch trim to tweak critical dimensions, and go to etch. Wafer-to-wafer variations could be reduced from 20 to less than 2 nm — the economic payback justifying the cost with a first-order correlation to device performance.

So is STC's solution viable? Let us consider: Are metrology and ATE platforms costly? Yes! Could these systems communicate better? Certainly! Are improvements needed? Absolutely! Are OEMs willing to become commodities? No!

Surely, a solution lies somewhere in between — in IM applications and an intelligent combination of platforms to do more than one measurement operation, either by combining technology or through modules to optimize the platform's capabilities and extend its lifetime. Beyond this, GEM compliance is probably the extent of the cooperation between competing ATE as well as other metrology OEMs. None are willing to publicize their IPs to the world. However, they must face and solve both the problem of increasing costs and shorter platform lifecycles.

Inspection and test requirements from processing to system will get increasingly demanding, but chip manufacturers cannot expect OEMs to meet the demands of R&D and productizing by becoming commodities.

For additional information on inspection, measurement and test, go to www.semiconductor.net/imt.