In the mid-1990s the U.S. military began to use more commercial off-the-shelf, or COTS, products. The U.S. Department of Defense (DoD) hoped that COTS would slash the cost and reduce the time required to field weapons systems. And so it did in some cases. Although there are as many definitions of COTS as there are experts, Avionics Magazine has adopted the following: COTS includes items--from chips to boards to boxes and the software that runs on them--that can be purchased directly from the manufacturer. They are not developed with government funds and are not customized for each user.
But COTS introduced its own cost and logistics issues. "This whole process of COTS leads to ever faster product cycles and things becoming no longer available," explains John Becker, director of diminishing manufacturing sources and material shortages (DMSMS) management and analysis in the Office of the Assistant Deputy Undersecretary of Defense for Logistics, Plans and Programs.
COTS content can be a mystery. The lack of visibility into weapon systems at the board and chip level and the military's historically federated approach to managing obsolescence translated into reactive decision making. Instead of anticipating a problem before it happened, people reacted to it afterwards. They lacked enough data about what COTS components were coming in, where they were being used, and what environments they were suited for to make efficient decisions about how to handle their obsolescence.
"It's possible to translate a $100 supply problem into a $100-million problem," laments Becker. The Navy, for its part, estimates that obsolescence issues cost up to $750 million annually. It hopes to reduce this figure by half through more aggressive management.
ARINC Inc., in a report prepared for the Defense Microelectronics Activity (DMEA), described how the F-16 program in 1995 spent $500 million to redesign an obsolete radar, how the KC-130 program in 1997 spent $264,000 on a life-of-type buy to resolve the issue of one obsolete logic device, and how the Air Force in 1997 reprogrammed $81 million for the F-22 program to purchase obsolete or soon to be out-of-production parts and to redesign assemblies so that they could accept commercial parts. Typically, the projects that ARINC studied involved replacing obsolete mil-spec parts with newer mil-spec or commercially available, industrial-grade parts. Experts differ about whether to consider mil-spec parts COTS.
Today COTS is on the rise in airborne systems. The venerable F-15 Eagle--designed in the 1960s and 1970s, before COTS arrived on the scene--uses only 10 to 15 percent COTS, says Samuel Calloway, an electronics engineer with the Avionics Hardware Engineering Group at Warner Robins Air Logistics Center in Georgia. Calloway defines COTS broadly to include commercially available military parts. But the F-15 contains purely commercial chips, such as Nvidia graphics processors, too. The program also plans to add a box that contains 50 percent "pure commercial, VME-type boards," he says.
Avionics designers are turning to commercial components for added performance. "State-of-the-art parts are virtually all commercial," says John Kravetz, engineering department manager with Honeywell's Aerospace and Electronics Systems business. He defines "commercial" parts as catalog items that are "sold to the world" and bought by the manufacturer's part number, excluding mil-spec components. Commercial parts, for example, include memories, microprocessors, graphics processors and field programmable gate arrays.
Typically, the piece parts are tested by the suppliers or third parties. Then Honeywell temperature-tests the box, paying particular attention to the parts. He estimates that as much as 70 percent of electronics devices on newer military airplanes are COTS. Experts estimate that memory chips can turn over in 18 months; processors in two to five years. Graphics processors have been known to go out of production in nine months.
Naval aircraft now use 20 percent to 50 percent COTS parts, according to Bob Ernst, head of the aging aircraft program at the Naval Air Systems Command (NavAir) in Patuxent River, Md. Memory chips and processors are almost 100 percent commercial. COTS boxes and boards are much less common, he says. And military-specific functions such as radar warning receivers, fire control systems and electronic surveillance measures, typically use either mil-spec or very ruggedized, industrial-grade parts. Most of the embedded equipment in bomb racks and launchers uses military-grade boards because of shock and vibration, Ernst says. Their components are application-specific.
To combat obsolescence on the F-15, "we can do a bridge buy--buy a number of processors to help us field and support a system until we do a [box-level] technology refresh," Calloway explains. In older aircraft, where a system will not be upgraded again, maintainers can do a life-of-type buy, typically a larger purchase. Since defense design cycles are measured in years or decades, "this causes us all kinds of problems," Calloway says. And sometimes, with pure commercial parts, there is no notice at all--or maybe 30 days' notice--that it's going out of production. That's not enough time for DoD to react, Calloway says.
Calloway also has relied on the Defense Logistics Agency's (DLA's) Generalized Emulation of Microcircuits (GEM) program to reverse engineer obsolete parts. New technology is used to design a form, fit and function replacement. GEM has emulated 400 obsolete F-15 circuits, ranging from standard transistor-transistor logic (TTL) and complementary metal oxide semiconductor (CMOS) devices to AND gates and NOR gates. GEM offers six or seven base wafers that can be "personalized" to the design required. Calloway cites a 95 percent to 96 percent first-pass success rate with these devices.
Emulation, however, is never the first choice. "We always start off looking at a life-of-type buy, and if we can't do that, we look at substitute or alternative parts," says Calloway. The next step--redesigning boards or boxes--is contemplated only if there are reliability, as well as obsolescence issues.
DoD, meanwhile, is trying to enforce a unified approach to obsolescence. We're changing the facts on the ground, Becker says. "This is a big ocean liner and we're pushing on the side of it. We're making it turn very slowly, but it is turning."
A lot has been accomplished already. ARINC, in its DMEA report, established cost metrics, indicating how much, on average, is expended to deal with obsolescence at the chip and board level. These metrics are widely (and voluntarily) used as a benchmark for cost management.
The B-2 program office, for example, applied the cost metrics to its avionics data and projected a 6-to-1 cost avoidance based on proactive obsolescence management. The 542nd Combat Sustainment Wing at the Warner Robins Air Logistics Center last year actually did chalk up a 43-to-1 cost avoidance ratio, using the DMEA guidance. In other words, from a range of possible maintenance solutions, they chose the most efficient and economical approaches, avoiding more than $340 million of possible expenses in solving obsolescence problems.
After studying data from numerous DoD programs, ARINC estimated that the average nonrecurring engineering (NRE) cost of modifying a circuit board can range as high as $162,000 (minor) to $536,000 (major) in 2005 dollars. This includes engineering, program management, integration, testing and update of the technical data package. But it excludes system-level qualification test or flight test, which could double or even quadruple the card redesign costs, explains Walt Tomczykowski, director of ARINC's obsolescence management and reliability department.
Short of such measures, ARINC listed a hierarchy of other steps, from using existing supplies (no cost) to emulation (up to $95,000). Ninety-five percent of the time, military program managers can find a replacement part, he says, but about 5 percent of the time, they may require an expensive board redesign. Planning ahead, he adds, just like planning ahead for disasters, will help programs implement the most economical solution. If they don't plan ahead, programs should know what the operational impacts will be if solutions are not implemented.
Expected any day now is what Becker describes as a DMSMS guide for program managers, a "DoD way." This document will show people how to manage obsolescence down to the lowest level of NRE costs involved in finding a solution. It will show what the hierarchy of costs is and provide a rating scheme that will grade programs according to how proactive they are.
Program manager accountability is the "whole idea," Becker says. Among other things, the guidance will tell managers to find the lowest-cost solution. The Defense Department also is establishing metrics that show the different programs' case resolution results in order to establish an internal benchmark. The new approach "is being integrated into the review process for the 'acquisition category' [major defense acquisition] programs that come through the building," Becker adds. It will be integrated into milestone reviews and extended to programs already in sustainment.
The Navy already has moved on the issue in a recent directive from the assistant secretary for research, development and acquisition. "They've issued that `thou shalt have a bill of materials [BOM] down to the level of the chip for all your systems,'" says Becker. He also notes the development of a "center of excellence" Web site (www.dmsms.org) that will serve as a central repository for information such as course material and tools. Becker's vision: "Find a solution once and use it many times instead solving the same problem 1,000 times in isolation."
The Navy is requiring acquisition programs at least to have a plan and use business case analysis," Ernst says. "We're also trying to understand how to teach our teams to comply with the policy, to create a template or guidebook of best practices." The Navy, DLA and Defense Supply Center Columbus (DSCC) are developing a "program manager's guide for dummies"--a joint course that will be posted on the COE Web site.
The Navy is asking programs to develop and maintain an "indentured parts list" that ties chips back to individual cards and boxes. "Even if we go to a performance-based logistics concept [where a vendor manages support], we still want to retain the data so that we have some insight and force the contractors to maintain some idea of configuration--of what they have," Ernst emphasizes.
The logistics support community also wants to have a shared data warehouse so that program managers can discover what solutions have been found elsewhere to obsolete parts problems. The Air Force's solution for an F-15 issue may not always fit the needs of an F/A-18 maintainer, but at least it would avoid unnecessary duplication of effort.
Data will ensure more informed decisions. You don't just change a component in isolation, says George (Ric) Loeslein, NavAir's deputy aging aircraft program manager for obsolescence. "You may find that the same component is used in several cards, or you may find there are other components that are obsolete, as well." And those facts may lead you to redesign the whole card, he explains.
"All the services have been trying to get indentured parts lists at some level for a number of years," Ernst says. But the Jan. 27, 2005, Navy DMSMS policy guidance mentioned earlier sets a new tone. It requires acquisition managers to "establish a formal DMSMS plan [to] cover all phases of the life cycle, from program initiation through sustainment and disposal." The key elements of such a plan are processes to:
Integrate DMSMS strategies with technology road maps;
Obtain and use configuration data, e.g., a bill of materials;
Identify and forecast piece part obsolescence impacts and mitigations for all configurations;
Use predictive, cost-effective industry solutions;
Identify systems that use the same components and technologies, and establish commodity management and preferred material processes across programs;
Provide and use documentation and metrics that track DMSMS cases and trends, associated solutions and cost;
Assess contractors' DMSMS programs and ensure they meet program needs; and
Use business case analyses.
Part of the reason BOM data has been so difficult to get in the past is that companies thought the government wanted all their manufacturing data, their trade secrets, Ernst says. But all the government needs is a list of parts, a topdown breakdown and any critical tailoring characteristics, such as screening for parameters like temperature. "We don't need the trace layouts on the cards or the firmware," although the latter could be held in escrow, just in case.
NavAir is promoting an effort to increase visibility into commercial parts' ability to perform reliably in the aerospace environment and ensure stability in their supply, Ernst says. In other words, if the service is going to use a specific part, it needs to know what the part does, what its critical characteristics are, and whether it will work in the intended environment. "You can't just stick your head in the sand and grab parts, which is kind of what we're doing right now," he says. These ideas are being developed by a collection of over 15 semiconductor manufacturers and avionics developers known as the Aerospace Qualified Electronic Components (AQEC) group.
The AQEC document, still in draft, will "call out the requirements for COTS products [to be used] in avionics," says a DLA official. AQEC is trying to come up with a set of rules to cover this type of product--not mil-spec parts, but not pure commercial parts either, which are subject to considerable variability. Pure commercial parts may be fabricated in different locations, assembled in other locations, and use different coatings or molding compounds.
AQEC is interested in "modified COTS," like the "enhanced plastic" (EP) products that Texas Instruments is using. "It comes from one fab [fabrication facility] and has one assembly site." However, components like this are not hermetically sealed like a mil-spec part, he says.
AQEC suppliers also will have to provide products for five or more years or provide the information necessary to obtain the product--i.e., through a distributor--if the life cycle will be shorter than that, the official says. Test data also will be part of the deal, he predicts. Normally, it's the testing the manufacturer already does, he notes. AQEC is trying not to add cost, but it aims to provide users more than pure commercial products do. "They will make available testing data on the product, which they typically may not do for small customers."
AQEC will provide DoD better knowledge of the parts being used across all weapon systems, Becker says. That knowledge means there is a larger pool of potential replacements if something goes out of production. "We'll define preferred parts lists and do business with preferred vendors."
AQEC and other groups also are exploring the concept of volume buys to make the most of the Pentagon's minuscule purchasing power in the electronics market. If enough companies agree to buy AQEC parts, suppliers may be more forthcoming with data. AQEC participants have included Boeing, Lockheed Martin, Northrop Grumman, Honeywell, BAE Systems, Smiths Aerospace, Rockwell Collins and Goodrich, as well as part manufacturers such as Motorola, Texas Instruments, Micron, IBM, Intel, Xilinx, National Semiconductor, LSI Logic, Altera and Analog Devices.
DoD, after all, is an ever smaller portion of the electronics market, Becker says, noting that its estimated 0.5 percent market share may be an exaggeration. (Some people put market share as low as 0.1 percent.) "We make it worse because we take that [portion] and split it into a thousand pieces," he says. So DoD is "going to try to roll it back up to 0.5 percent." He envisions a "microcircuits commodity council," for instance, where individual buyers pool their volumes and use preferred suppliers. "We're going to try to start to communicate with industry as one DoD instead of 1,000 little DoDs," Becker states.