By J.R. Wilson
Defense and aerospace systems rely heavily on 5-volt power systems, while commercial electronics have already moved beyond 3.3- and 1.8-volt systems. This disparity in commercial and military power standards exacerbates an already critical obsolescence problem in the military world
Imagine a billion cellular telephones, the vast majority of which the owners will replace in less than three years. Now imagine a billion consumers who demand their cell phones to be small, lightweight, multifunctional, clear sounding, and long lasting between battery charges.
Meeting such demand is not an easy task, and involves some prickly issues related to how systems integrators provide electrical power. To meet the burgeoning demand from cell phone users — and similar demands from hundreds of millions of other consumers with electronic toys, appliances, gadgets, and personal computers —integrated circuit (IC) designers must push the voltage of digital signal processors (DSPs) constantly downward.
"Five-volt, 3.3-volt and 2.5-volt technology has peaked, 1.8-volt is here now, 1-volt has been demonstrated in DSPs, and sub-1-volt is coming," sums up Jack Stradley, central and southwestern U.S. manager for Rochester Electronics of Newburyport, Mass. "The DSP market is primarily being driven by the cell phone market. To cell phones, power is precious, so the less power consumption, the better. They want to go to sub-1-volt and it looks like they'll be there by 2005, perhaps sooner."
This raises serious issues for the military, aviation, and satellite communities. Equipment in these fields is built to exacting standards and designs that have been years in the making. These systems depend on systems with proven reliability under intense operating conditions, which designers expect to operate flawlessly for years, if not decades.
Unfortunately, from a technology development standpoint, they also continue to operate with old — or "legacy" — 5-volt systems. For military-quality electronics, the new low-power standards, while in some cases desirable, present serious problems as far as interface and environmental requirements are concerned.
"One of the problems in avionics is finding products that will meet the temperature range required by avionics. When you get down to that low voltage range, you won't be operating at minus -55 to 125 degrees Celsius. You'll be at between 0 and 50 C," Stradley warns. "These new technologies won't even make 0 to 70 C, which is the current commercial range. They don't need to. Cell phones work at ambient temperatures. And it's a throwaway item. You don't get those repaired; you replace them. But you have legacy (aviation) systems already 20 or 30 years old that are supposed to operate another 20 or 30 years and have to be repaired. But how?
"Most of the major weapons systems have some kind of DSP," Stradley explains. "If they are going to use state-of-the-art technology, they will have to use DSPs operating at 1-volt or less."
The U.S. Air Force B-52 bomber, which has been in service since the 1950s, is a case in point, Stradley says. "If you have a B-52 that is supposed to fly 'til 2040, at 2010 you will have a B-52 still operating at 5-volt — and that's a dirty 5-volt, with lots of spikes, perhaps up to 2-volt," he says. "If you hang a smart missile on that wing that operates at 1-volt, are you going to have enough electronic interface between those two systems to allow them to function together?"
One solution is to ignore the new low-power components and continue operating with the old technologies. The problem is finding replacements.
Using old technology
To address that from one angle, the government has allocated some $13 million to dismantle electronic equipment for spare parts. The program is called "demanufacturing of electronic equipment for reuse and recycling" — better known as DEER2. The money is going to Concurrent Technologies Corp. (CTC) of Largo, Fla., an independent non-profit organization created to assist industry and government find solutions to technology problems. Under DEER2, CTC engineers recycle public and private sector electronic equipment and, where applicable, pulls out components that may be used elsewhere to deal with obsolescence issues.
In 1998, CTC technicians recycled some 275 million pounds (9.7 million units) of electronic products and reused approximately 33 million pounds of electronics parts, sub-assemblies, and materials that are compatible with the old 5-volt power standards. Some in industry, however, question the reliability of recycled components.
Another alternative is the Defense Microelectronics Activity (DMEA) in Sacramento, Calif., created in May 1997 from elements of the Sacramento Air Logistics Center at McClellen Air Force Base. When McClellen was placed on the closure list for 2001, the U.S. Department of Defense decided to retain its microelectronics fabrication capability, given the government's rapidly declining clout with the commercial electronics industry.
DMEA officials have acquired licenses to commercial microcircuit processes, many of which the commercial world is no longer using. DMEA experts then use what they call their "Flexible Foundry" to meet low-volume military — and commercial aviation — requirements, and to ensure long-term availability of obsolescent microcircuits that that work well with old 5-volt power systems.
"While industry has reduced or eliminated its focus on military microelectronics, this does not at all diminish DOD's reliance on this critical technology," says DMEA technical advisor Dr. Gary Gaugler. "In reality, military systems rely more on microelectronics than ever before with an increasing emphasis on 'smart' and 'brilliant' systems — especially fire-and-forget — integrated microcircuits are the key enabling technology. On the other end of the technology spectrum is the need to keep existing systems running."
Yet the challenge of keeping old 5-volt systems running becomes more serious with each passing month. According to industry reports, 1998 was the first year in which production of relatively low-voltage microcircuits exceeded that of 5-volt parts. This trend is continuing apace.
"One can anticipate that within a short time, the availability of 5-volt parts from industry will cease," Gaugler warns. "The cost to re-design systems for 3.3-volt operation is high. Furthermore, since many of these systems are software-based, the cost to re-host and re-qualify the software makes the prospect that much more bleak. Clearly, the most cost-effective solution for DOD is to be able to obtain 5-volt parts for as long as they are needed.
That is where the strengths of DMEA come to bear. "We believe the best model is to have us be the original makers of the silicon because, by the time they get through designing, prototyping and flight testing and get into production, it's no longer made," Gaugler says. "We're doing that now on the F-22, which needs about 600 chips a year and the foundry they were using no longer supports them."
The option of converting all 5-volt military systems to lower-voltage standards could lose the performance of the inherent gate speeds of native 3.3-volt devices.
"As you reduce the voltage, the performance decreases at a given feature size," he explains. "At one micron, we can handle 5-volt and 3.3-volt, but if you reduce that operating voltage, your speed goes down. So if you need to have a higher performance at the lower voltage, you need to reduce your feature size, say to 0.6 micron, which has inherently higher performance. If the performance of the 0.6-micron process at 3.3-volt is poorer than that of the native 3.3-volt process devices, then we do have to engage a different process. Then we would likely move to something at a higher performance, like 0.35 micron."
In considering all these elements, systems designers must consider what, exactly, they really need these components to do and whether it is even necessary to incorporate state-of- the-art technology.
"In the military, you have huge power capacity, so you really don't care if it is 5-volt versus 3.3-volt," Gaugler points out. "If you look strictly at performance, such as 'smart' weapons driven by microelectronics, what is really required? You're talking about processors, sensors, and memory. The weapons don't move that fast and are really using finite, fixed algorithms. It is not a computing-sensitive environment. Of course, being able to operate in a low-power environment can help in mission duration, but there is a trade-off to consider," Gaugler says.
"Does the military really need the equivalent of 900 MHz Pentium IIIs with a gigabyte of memory in its airplanes in one place? No," Gaugler says. "All of that requirement is distributed among multiple processors with individual functions assigned to them throughout the airplane — and some of those are redundant. You don't need a lot of computing power when everything is distributed like that."
Software is another factor. Old systems have old software that runs on old computers, This software has been proven and flight tested on an old system. It would cost millions of dollars — and no small investment in time — to prove it out on a newer system.
"In some situations, you have a 1,000-to-1 difference in cost of software to hardware," Gaugler says. "If the military had to take all their software and repost it on something else, it would cost $25 billion for the software alone, according to a Boeing study about five years ago. The longer you can keep the hardware support infrastructure in place, which in the military is predominantly 5-volt, you don't impact the software. And that is really important."
The problem is finding a source for those older technologies long after they have been abandoned by the fast-moving, high-volume commercial world.
"Our approach is to be able to provide just the right technology at the right time in the right volume," Gaugler says of DMEA. This, he says, "means having the ability to produce relatively low volumes of parts very cost effectively, very efficiently, within a short time span of when there is a requirement for them in order to sustain existing systems.
"There really are no multiple sources for anything any more," Gaugler continues. "The days of five or six sources for chips or other elements are over. But as long as DMEA is around and we have the Flexible Foundry, we will make chips. If there is a requirement, we'll supply them. That's our mission."
While DMEA is a non-profit government operation, it is not always a perfect solution, according to some in the industry — even if it may sometimes be the only one available.
"To have a part specially made by DMEA is not a cheap solution," notes Paul Wakefield, marketing director for the National Semiconductor Corp. Enhanced Solutions Group in Santa Clara, Calif. "But if you have a legacy system and need a part, the important thing is to have a solution and the cost to redesign the whole system would be even more expensive."
National is among a handful of companies whose leaders vow to continue supporting the military and other customers who have a long-term need for 5-volt and 3.3-volt systems. Part of that involves distributed power supplies with onboard regulators for local clusters of low voltage — special regulators that stand between 5-volt and 3.3-volt components.
"At the moment, the market for us has predominantly been 5-volt down to 3.3-volt," Wakefield says. "3.3-volt has taken a long time to get into the military compared to the commercial world because of the legacy systems and the difficulty in dealing with mixed systems. Translations down result in burning a lot of extra power and, for designers today, everything has to be the lowest power use possible."
Analog technology is easier to switch because "because you can scale signals to account for differences as you go through the system," Wakefield says. "But with digital, if you have 5-volt signals and 3.3-volt ASICs (application-specific integrated circuits), you have a bigger problem. In those situations, you need a range of products, such as translators."
Experts at the Texas Instruments Military Products Group in Sherman, Texas, are adapting switchmode power supplies TI designed for very small devices, such as handheld games, to meet some of the military need.
"If they were attempting to link to one of our 1.5-volt products, there would be additional capacitors local to the regulator that could help smooth out transients," says Tom Tarr, strategic marketing and applications engineer at TI Military Semiconductor. "We also have a large selection of linear regulators, although if you are going from 5-volt down to 1.8-volt, you will dissipate more power across the pass transistor of the regulator than you will be consuming at your end device. That's one of the advantages of using a switchmode."
I has offers other alternatives as well. "We also have adjustable regulators, where the regulation voltage is set by external resistors," Tarr says. "In effect, those external resistors become a voltage divider and you create a reference voltage. By using two resistors, you can determine what the output voltage will be on an adjustable regardless of the input voltage. In some cases, we may have a 30-volt input going down to below 5-volt output."
While point-of-use regulators and similar devices can handle a variety of different voltages, they, too, are far from a perfect solution, says John Fink, staff engineer at Honeywell Commercial Aviation Products in Minneapolis.
"We're finding obsolescence is making us replace a lot of stuff nobody had planned to replace in the first place because you can't maintain it any longer," he says. "The solution they're starting to come to, on the commercial side, is to do functional card replacements — a card that does the same job as the chip or some other function in the box, using current parts. But that only buys you a little time, because it will happen to you again. And that can be very expensive. It's not something you do lightly."
Fink says commercial systems designers finally are beginning to band together to address the issue, with such traditional competitors as Honeywell, Collins, Airbus and Boeing meeting to seek common solutions.
"We have figured out we can't stand alone anymore, but standing together we're only a little better off. But if we address it as an industry, we will be working the same problem toward a common solution. I don't know what else any of us can do. Ben Franklin's old comment applies perfectly to this case — if we don't hang together, we surely will hang separately."
Cell phone technology
IBM's next-generation silicon germanium (SiGe) chip technology, announced in July for high frequency and low-power communications applications, is an example of cell phone-driven technology with strong military implications.
Based on an ASIC-compatible, 0.25-micron BiCMOS manufacturing process, the new SiGe technology offers 47 GHz performance with a 2.5-volt power supply. IBM engineers are finalizing specifications for the 0.18 micron SiGe technology generation, featuring the addition of copper interconnect, which has been used to produce SiGe test devices in the range of 90 to 130 GHz.
"IBM is not driving the voltages for some of these applications down on its own. A lot of this is determined by the cell phone and things probably are moving slower than that community is interested in doing in terms of reducing power requirements," according to Dave Ahlgren, senior engineering manager for SiGe technology development at the IBM Microelectronics Division in Fishkill, N.Y.
"SiGe was not designed to lead the pack on power requirements, but rather follow the CMOS evolution, which is offset six to nine months behind even the standard CMOS offerings from IBM," Ahlgren says. "CMOS does not make good analog circuits, so we are adding a bipolar device that makes a very high performance analog design space, which is necessary in a lot of telecommunications applications."
The Defense Advanced Research Projects Agency (DARPA) in Arlington, Va., has helped pay for some of the developments on SiGe because of the military-level high performance of the bipolar transistor. IBM officials report producing devices that switch at between 50 and 100 billion times a second — faster than any CMOS device and pushing the limits of gallium arsenide technology.
"Most of the applications where we work directly with the designers are not military but civilian, where you are replacing a whole module," Ahlgren says. "A new cell phone will be developed around a 2.5-volt power supply. The GPS receivers being built also will be designed for it and won't really contain any legacy elements, so you don't have a mixture of semiconductors."
SiGe also is being adapted to wireless communications for aircraft, but its relatively low power requirements and cost savings of silicon versus gallium arsenide that make it attractive to the cell phone industry are of considerably less importance to military and aviation users.
"The cost of redesigning far outweighs any savings in cost in upgrading the generations of technology," Ahlgren says. "One of our customers working on designing parts for aviation is less concerned about reducing the cost of materials by 12 cents than in getting the best possible products."
Power supplies are inexpensive and represent simple designs, so systems designers have had little incentive to set standards that ensure commonality among these devices, Ahlgren points out. "You might have 10 different power supplies of different voltages on a B-52," he says. "A lot of the things we are designing today are powered by batteries. Is that a reasonable approach for an aircraft — a thousand components, each powered by a different battery? You certainly can't replace every component on an aircraft just because the new communications system is designed to operate at 2.5-volt."
Satellites, which their designers do not expect to operate for decades, nonetheless share many of the power supply problems associated with military equipment and commercial aviation. Every extra watt used in space means bigger power supplies, which means more weight — and more cost — to launch.
"Low voltage is of paramount importance to satellite designers, but there are trade-offs," says National's Wakefield. "If they know a product very well and how it performs at various temperature ranges and radiation levels, they will stick with it. There are areas where they need to use new technology, but they don't want to take risks if they can avoid it, so they have been much slower to move to lower voltages."
To drive this point home, Wakefield says designers tend to dust off an old design using 10-volt power supplies about once a year or so for a new satellite. That way, he says, the user can be certain it will work, regardless of technology lag. In such cases, designers trade off between size, performance, and reliability.
"As the geometry size comes down on CMOS parts, the radiation performance from a total dose point of view, as a rule of thumb, tends to improve, however single-event effects tend to increase," Wakefield says. "For satellite designers, going down in voltage gives them a power benefit and can improve total dose performance, but might increase the likelihood of catastrophic events."
For that reason, such critical satellite functions as data processing largely have remained the purview of 5-volt microprocessors even as the commercial world migrated to 3.3-volt and below.
"The critical pieces of new satellite bus electronics are still locked into legacy designs. The computations and signals processing components still have to be balanced against a tradeoff of the most powerful technology out there versus ensuring it can be rated so the mission won't fail," says Tom Marshall, manager for rad-hard space marketing at Intersil Corp. Power Management Products in Melbourne, Fla.
Less critical areas have adopted the newer technologies, however, leading to a mix-and-match design in which 3.3-volt and below elements have been integrated into the system over time. This itself can present design challenges. "As these digital voltages continue to drop, more consideration has to be taken to design everything to work properly," Marshall says. "The trend is toward every board coming with a regulator built in. What we are doing on the satellite side wouldn't be applicable [to military or commercial aviation] because, with radiation hardness and space screening requirements, we have overspeced parts for aviation applications."
The military and aviation communities also must consider the relative value of tradeoffs when considering new technologies. A case in point involves the question of how to keep the aging B-52 bomber compatible with sub-1-volt microelectronic systems that will become operational over the next decade.
"They will have to make a decision on the relative options — develop special translators or make parts specifically for the needs [the DMEA model] or rip out the old systems and replace them with new," Wakefield says. "That will have to be judged at that time and it's impossible to state today what the solution will be then." The best way to avoid similar problems in future military systems, he says, is to invoke a sea change in military thinking.
"It requires a new design philosophy that allows you to look at systems in a more modular way," Wakefield says. "Defense contractors are looking more at being integrators of standard building blocks than designing systems from the component level up and finding ways of incorporating from a standard function point of view. The industry is still trying to find the best solution — and that solution will depend on the customer it is serving. One shoe doesn't fit everyone."
The main problem for the military, from the suppliers' viewpoint, is sheer market volume. The military is no longer the technology definer, but has become a minor player in an increasingly electronicized commercial world. But Wakefield and others in the industry reject claims they are abandoning the military.
"It is always interesting when you are blamed for obsoleting a device, but the basis for obsolescence generally is no one is buying it," Wakefield explains. "Families of products are based on a process and sometimes a process will obsolesce because it is too expensive to keep it in place for the number of products being produced. In many cases, there will be an upgrade part offered with different devices that meet the same specifications, but that is not always the case."
However, he adds, the criteria for a military product are actually very low compared to commercial products in terms of what constitutes a low level of business.
"There are still suppliers interested in supplying for the military market and you do see new products coming out to support those markets," he says. "The military has been having to make hard choices for a number of years as some suppliers have pulled out.
"Acquisition reform also changed the view of how defense contractors can buy products and really changed the way they looked at military products," Wakefield says, adding the military actually is a "small" market only in comparison to the massive size of the commercial demand.
"From our point of view, we still see this as a market we have a commitment to support," Wakefield says.