COTScon 98: upscreening shootout at the McLean Hilton

McLEAN, Va. - Upscreening was the lightning rod at the inaugural COTScon conference May 28-29 in the McLean Hilton sponsored by Military & Aerospace Electronics - and lightning struck repeatedly.

Jul 1st, 1998

By John Rhea

McLEAN, Va. - Upscreening was the lightning rod at the inaugural COTScon conference May 28-29 in the McLean Hilton sponsored by Military & Aerospace Electronics - and lightning struck repeatedly.

Upscreening is a process for qualifying commercial-grade semiconductor components for temperature and other environmental extremes that are common in weapon systems. Upscreening often can be done (and often is) by the semiconductor manufacturers themselves or sometimes at large system houses with in-house testing capability.

Increasingly, however, technicians at independent test laboratories are doing the upscreen work. These test labs are dedicated specifically to this purpose as a service to printed circuit board manufacturers. In fact, a new test lab industry is being born out of former Defense Secretary William Perry`s commercial off-the-shelf (COTS) initiative. This, moreover, is causing concern at virtually all levels of military and aerospace systems design - especially among component producers, board manufacturers, systems integrators, and military program managers.

One of the most vocal critics of this process is John Hartman, corporate defense business development manager at Analog Devices in Wilmington, Mass., who raises the logical question: who is legally liable for a failed upscreened part if the failure leads to a catastrophic accident? Certainly not the parts manufacturer, he contends.

Characteristics of individual parts vary among wafer-fab facilities and even from batch to batch, Hartman told a COTScon `98 session. Who, he asks, is better qualified to screen semiconductors than the original producers? Moreover, Hartman points out, semiconductors are like ripe fruit on grocery shelves; they do not tolerate a lot of handling.

Chip and board makers at odds

The upscreening issue is also creating a difficult situation in the military and aerospace electronics supply chain by pitting chip manufacturers against board suppliers. Regardless of who does the upscreening, this process makes sense for the board producers.

The commercial parts to be upscreened are cheaper than the comparable mil-grade or MIL-PRF-38535 Qualified Manufacturing List (QML) parts and typically closer to the leading edge of technology. In the changing defense environment the board manufacturers are becoming the center of gravity of the industry by default, and that logically leads to fierce competition.

The underlying dynamic is the shrinkage of the defense business at both ends. The system prime contractors are merging into a handful of megaconglomerates, and suppliers of military-grade semiconductors essentially are down to four: Analog Devices; Harris Semiconductor of Melbourne, Fla.; National Semiconductor Corp. of Sunnyvale, Calif.; and Texas Instruments Semiconductor Group in Midland, Texas.

Moreover, the U.S. market for military-grade semiconductors declined from about $900 million to less than $700 million between 1992 and 1997. The modern semiconductor industry, born in the 1960s when Air Force leaders placed large orders with Texas Instruments for the Minuteman ballistic missile, is now too busy supplying the voracious needs of new commercial industries to pay much attention to such a relatively small military market - unless the price is right.

For example, last year the big four suppliers rendered 1,782 commercial-grade plastic parts and 438 ceramic military parts obsolete, explains Robert Kroeger, general manager of TI`s military semiconductor products division in Midland, Texas. Today, the needs of the new generations of personal computers, not military systems, drives the company`s output, he says.

In this environment, obsolescence has become a fact of life. One expedient available to program managers and prime contractors is to stock up on the parts they think they will need for the lifecycles of their weapon systems - or at least enough to serve as a "bridge" until they can find another solution, Kroeger adds. Yet making long-term projections on the multi-year lifespans of many systems often is about as accurate as consulting a crystal ball.

DOD in the middle

The Defense Department is caught in the middle. DOD officials realize they do not have much clout in the semiconductor market any more, pointed out Gregory Saunders, director of the Defense Standardization Office, in his keynote address to COTScon `98 attendees. Moreover, DOD officials must find alternative solutions if they are to keep COTS viable.

This does not mean abandoning COTS, Saunders emphasized. He points out that COTS parts actually are more reliable than mil-spec parts in the avionics systems of such older aircraft as the F-111, B-52, and F-16. COTS is also expected to play a key role in the F/A-18 fighter-bomber avionics, Apache Longbow attack helicopter, and the extended-range SLAM missile, he adds.

But the situation does not look promising unless everybody concerned does some "thinking outside the box." On the Multiple Launch Rocket System, for example, Saunders notes not only that the number of obsolete parts doubled from 5 percent to 10 percent from 1993 to 1995, but also that the parts available only through sole sources nearly tripled from 20 percent to 55 percent.

And there is still the question of who will take responsibility for a parts failure.

Some military suppliers are clear on who will take responsibility. The board manufacturers will, affirms Gorky Chin, vice president of advanced technology at VME board designer Vista Controls Corp. in Santa Clarita, Calif. His fellow COTScon `98 panelists, Rodger Hosking, vice president of Pentek Inc. in Upper Saddle River, N.J.; Robert McKee on behalf of the VME International Trade Association in Scottsdale, Ariz.; and Duncan Young, president of DY 4 Systems of Kanata, Ontario, assented.

Chin notes that it is essential to match the part to the application. As evidence of this, he cites the Vista Controls VME boards on the integrated mission management computers for the Global Hawk unmanned aerial vehicle. There are also new techniques that designers can apply early in a program, such as the highly accelerated stress testing method, better known as HAST. This method, when combined with sophisticated analyses, can help designers predict reliability in the field and enable board producers to find the right match.

The key to the upscreening process, Chin says, is to achieve satisfactory yields of the upscreened parts. He says he wants to see yield percentages in the high 90s.

Common-sense approach

Despite the inherent risks of upscreening, there is no reason that common sense cannot prevail, Hosking says. Use QML parts when the program managers demand them, he told COTScon `98, and do not use components beyond the manufacturers` ratings. McKee of VITA also encourages common sense. Screen all parts to be sure they will work in temperature ranges of at least -40 to 85 degrees Celsius, he says, and use an open-systems approach to create a reliable parts supplier base.

Systems manufacturers are beginning to cite successful track records using COTS components. Lewis Johnston, vice president for advanced technology at Racal Communications Inc. in Rockville, Md., says he is enthusiastic about COTS components for his company`s line of tactical radios.

MIL-STD-883 components, he says, are between three and six times as expensive as commercial-grade plastic microcircuits, and Racal designers have successfully used COTS parts in a 3-pound version of the U.S. Army`s 22-pound AN/PRC-119 Single Channel Ground and Airborne Radio System called Leprechaun. Even using plastic encapsulated microcircuits, Racal designers got better reliability than did designers of the 22-pound version of SINCGARS, Johnston says.

Applications of liquid crystal displays (LCDs) also hold promise for using COTS components, says Mike Forde, product manager for rugged displays at BARCO North America in Tucker, Ga. BARCO designers, who provide displays for the U.S. Navy AN/UYQ-70 shipboard computer and display system, do not use expensive custom displays unless it is absolutely essential, Forde says.

There are two commercial grades of LCDs available that Forde estimates would meet 80 percent of military requirements (aircraft cockpits are not one of them) if properly packaged: office-automation LCDs and factory-automation LCDs (also used in automotive applications). Packaging is critical, Forde notes, not just for the temperature range, but also for structural integrity and resistance to humidity.

Rise of the test labs

In the middle of this fray are the test labs, whose officials were trying to tell everybody visiting their COTScon `98 exhibits that they are aware of the problems of upscreening and are taking steps to address them. Among the upscreeners exhibiting at COTScon `98 were Lucent Technologies Component Evaluation Technology of Wichita, Kan., Bell Technologies Testing Division of Orlando, Fla., and Pikes Peak Test Labs Inc. of Colorado Springs, Colo. Lucent, coincidentally, was bought from the parent company on the last day of the conference by five of the organization`s executives and renamed Integra Technologies.

Speaking on behalf of the test labs, Richard Guy, vice president for marketing at Integra, summarizes the pitfalls of upscreening - particularly as the semiconductor manufacturers see them:

- why upscreen at all when mil-spec or QML parts are available?;

- rescreened parts are "used parts," and the extra handling can reduce performance;

- the third-party testers cannot do as good a job as the manufacturers; and

- if the parts are used beyond data sheet limits the chip manufacturers will not take responsibility.

Guy points out that test labs simply provide a service, typically to the board manufacturers but also to the system prime contractors, and only verify that the parts pass tests ordered by the customers. The semiconductor manufacturers, Guy says, are not eager to do special screening, particularly for the small lots involved in military programs.

The reality, he claims, is that they are far better off if they can encourage the users to stick with the more expensive mil-spec parts. Given the small number of suppliers and limited demand, the prices of these parts are not likely to decline much.

New upscreening industry

The test labs really do represent a new industry with all the accompanying startup problems of determining how they relate to their customers. The test labs cannot test the parts exactly the way the semiconductor companies do because much of the needed information is proprietary to the manufacturers. They do not try to, says Guy; that would duplicate what the manufacturers already do and would not help the end users. Instead, he adds, the test labs test the devices for specific applications. They generally use the same kinds of test equipment and procedures as the manufacturers.

As a result, the board and system companies have to make informed cost/performance tradeoffs. "We do not recommend that they upscreen 100 percent of the device types they use without first analyzing their application needs," Guy says. "It is today, and always will be, a risk decision vs. total-cost-of-ownership to be made by the end user whether to use military parts, QML, COTS, or upscreened parts."

There is a halfway house here, and several COTScon `98 speakers, including McKee from VITA, touched on it: ruggedize the enclosure rather than, or in addition to, the parts. Designers can apply this process from the components to the modules, chassis, rack, or entire system or platform, McKee says.

Given the design constraints, this approach is not universally applicable. Yet a success story is the Navy`s New Attack Submarine, in which the critical electronics are protected in a ruggedized COTS compartment about 60 feet high and 30 feet wide. This has cut hardware costs 75 percent below mil-spec parts, Saunders of DOD points out, and enabled the new submarine to have 25 times more signal processing capacity and 50 times more data processing capacity than its predecessor, the Seawolf.

The idea appears to be catching on elsewhere in the Navy. Capt. J. Talbot Manvel Jr. is program executive officer for the Future Carrier Program called CVX. He told COTScon `98 that he expects to implement COTS on his new carrier, which he called "the COTS platform of the future," when it becomes operational in 2013. It will cost $21 billion and should last 50 years.

Manvel says he welcomes ideas from contractors and notes that officials of Newport News Shipbuilding in Newport News, Va., the builder of the aircraft carrier USS Ronald Reagan (CVN-76), initiated the idea for a fiber optic internal communications system. What he wants most of all is an open-systems, standards-based approach, but he says the COTS components for the CVX have not even been designed yet.

As a subset to this idea of ruggedizing the package rather than its contents, officials of a COTScon `98 exhibitor, Isothermal Systems Research Inc. of Clarkston, Wash., reminded participants that it is still possible to use coolant fluids to meet military temperature specifications. The legendary computer designer Seymour Cray did that 20 years ago to cool his top-of-the-line supercomputers at the old Control Data Corp., and Isothermal designers are still using the same fluid, Fluorinert from 3M Co.

The process is simple, maintains Isothermal president Donald Tilton. The spray cooling (which is approved by the Environmental Protection Agency and the Occupational Safety and Health Administration) is readily available and can protect COTS parts in harsh environments.

The company has chassis available in standard ARINC air transportable rack configurations, either as standard products or on a custom basis. It takes about a liter or two per ATR to keep the insides cool, and the fluid costs about $250 a gallon, Tilton notes.

If the diverse and frequently contentious messages of the conference could be synthesized into a central theme, it is that there is no lack of technical solutions to the problems of achieving adequate performance for COTS in critical systems. The catalyst in the process will be common sense. There should be plenty to talk about at COTScon `99.

More in Computers