• Article
[ Printer Friendly ]
Mark Maurer is the Manager of the Government Business Division of Silvaco Data Systems based in San Diego, CA. He is an alumni of Central College in Pella, Iowa.
Published in August/September 2006 issue of Chip Design Magazine
Outsourcing Poses Unique Challenges for the U.S. Military-Electronics Community
Defense study details growing microchip-supply crisis as conversion continues from traditional U.S. dominated IDM apporach to a globally distributed fabless model
There is growing recognition among high-level U.S. government officials that our domestic capabilities for the manufacture and design of advanced electronics are in jeopardy. The semiconductor industry's dramatic conversion over the last decade from a U.S.-dominated integrated-devicemanufacturer (IDM) model to the globally distributed fabless model is creating a microchip-supply crisis. That crisis was recently detailed in a February 2005 Defense Science Board Report [1]. It also has been cited in preceding reports [2, 3]. The DSB report concludes with a recommendation to Congress to make the supply of electronics a national priority. In this article, we highlight the particular challenges faced by the military and aerospace electronics community and some attempts to address those issues. We also recommend future action based on increased acceptance of the fabless semiconductor model in the military community.
ORIGIN OF THE PROBLEM
From the 1950s--when the modern semiconductor was invented--until well into the 1970s, the U.S. government was the major customer for the entire electronics industry. The earliest computers and memory chips were used by NASA. In addition, imaging technology was advanced by our intelligence communities. In these and many other cases, the technologies went on to create massive consumer markets. Nearly every major electronic advance from the solid-state radio to the most well- documented military electronic invention of all--the Internet-can trace a significant part of its lineage to federally funded research of some type. The 1980s brought a period of transition in which consumer electronics emerged as the dominant market force. With the end of the Cold War in the early '90s and the subsequent "dot-com" and electronics booms, the military went from being the driving market force in the industry to less than 1% of the U.S. marketplace. Along the way, nearly all of the integrated device manufacturers that supplied the DoD with advanced electronics left the military/aerospace marketplace. It's a simple question of economics. When DoD funding dried up, commercial, publicly held companies had a diffi cult time making a case to serve the military market.
Following the demise of military-specific IDMs came the rapid decline of the IDM business model altogether. Nearly every segment of the electronics industry has closed down dedicated IDM manufacturing facilities in favor of the fabless model. This new manufacturing paradigm is driven by the tremendous cost associated with modern state-of-the-art manufacturing facilities (billions of dollars). This cost is unsustainable for all but the highest-volume segments of the industry. This fact has led to the creation of specialized companies that have no finished products themselves, but serve as the manufacturer for ever-widening ranges of "fabless" consumer-electronics design companies. The aggregate volume generates enough business to support the huge capital and R&D expenses associated with (smaller numbers of ) semiconductor manufacturing facilities. Cost didn't just drive this model; it drove it offshore. Taiwan and the Peoples Republic of China control well over 70% of all fabless semiconductor manufacturing worldwide. This fact creates additional security- related problems for the DoD.
FIRST RESPONSE: COTS
Commercial consumer-product demands drive foundries to develop process technologies that are smaller, faster, and consume less power. The U.S. military, aerospace, and intelligence communities need many of these same traits in their systems. The concept of leveraging commercial-electronics infrastructure for DoD applications became part of the commercial-off -the-shelf (COTS) movement, which is a DoD attempt to meet system requirements using purely commercial parts. This approach is appropriate in cases where the commercial and military demands for performance are the same. Although this is true in some cases (e.g., hammers), electronics is another story. The military requirements for electronic performance and reliability often exceed what is available commercially. Even in cases where these technical requirements can be met, other requirements for non-obsolescence and security cannot be guaranteed in the rapidly changing, globally distributed commercial supply chain.
DOD PROCUREMENT METHODS AND POLICIES
DoD prime contractors and systems houses have historically relied on a simple but effective method of simply"buying parts" (one reason the COTS approach was initially attractive). They passed requirements down to an IDM provider of specialized piece-parts and had little involvement in the details of R&D, simulation, design, test, and qualification. Virtually all of the higher-level system-integration activities depended on this underlying infrastructure of specialized IDM parts suppliers.
Figure 1 illustrates how this model breaks down in the fabless model.
DoD program offices still specify requirements in much the same way that they did 20 years ago. Yet the approved standards (a must-have for the DoD) for doing business in the mode of Figure 1(b) aren't well developed in the big system houses. How does one carry out critical tasks, such as lot or part acceptance, when the basic elements of a part design flow can be diff erent corporate entities or -- in some cases -- individual consultants? How can system lifetimes of decades be supported by a supplier base that is changing daily? The increased complexity of the fabless model alone creates significant challenges for the systems integrators, who are accustomed to "qualified parts lists." The problem is made much worse for the DoD by the growing commercial practice of fabricating and packaging integrated circuits at offshore foundries. When security requirements demand U.S.-citizen participation only, even many U.S.-based companies and universities have become diffi cult environments for DoD electronics research and development.
A central issue for the procurement of military electronics is the ideal of trust. Our current policies dictate that for vital systems, the internal components must be designed and manufactured by domestic organizations. The solution that we advocate in this article involves increased acceptance of the pure-play semiconductor model within the military-electronics community. It also recommends the nationally directed preservation of onshore elements of the associated supply chain (see Figure 1).
TOWARD A SOLUTION
The radiation-hardened electronics community, which develops specialized electronics for missile and space applications, produced its own version of the 2005 DSB report in 1995. It was called the Hardened Electronics Availability Problem (HEAP) report [4]. Since that time, the community has been seeking ways to address the decline in the industrial base. It is instructive to examine some of the paths taken by the radhard community to address its own encounter with the diminished supply chain.
During the 1990s, most of the IDMs that served the niche "radhard" electronics business exited the market. The cost of maintaining special radhard fabrication processes far exceeded the return off ered by the post-Cold-War, low-volume requirements of radhard DoD customers. In response to the diminishing supplier base for radiation-hardened electronics, a Congressional-mandated Radiation Hardened Oversight Council (RHOC) was established. The RHOC collects system requirements for radiation-hardened electronics and manages the R&D investments for developing and producing these electronics in order to meet system roadmaps. One eff ort that's been identified and implemented through the RHOC has been to retain radhard IDM-like capability through government- funded (Title III and DTRA) capitalization of two CMOS foundries.
This approach is partially rooted in a long-held conviction that the DoD's most demanding radiation survivability requirements (relating to nuclear threats) can only be met through very specialized processing techniques, which the onshore commercial foundries are unwilling (unable) to support without large government investment. Government-subsidized fabrication capability is a costly approach. In this case, however, the DoD end users are dealing with inescapably serious issues. Although the fall of the Soviet Union dramatically reduced the large-scale, strategic nuclear threat, even a single high-altitude nuclear detonation could severely impact both commercial and military space assets [5, 6]. Such a detonation would impact critical infrastructure ranging from banking and finance to intelligence and conventional military infrastructure. Even if it is low-probability, this type of high-consequence event places irreducible radiation survivability requirements on certain DoD assets. It's understandable that large investments have been made in order to assure a supply of strategic (nuclear-hardened) technology. When one considers the economic forces that have driven commercial foundries offshore, however, the concept of maintaining on-shore fabrication solely at the government's expense seems unrealistic as a long-term solution for all DoD electronics needs.
The radhard community recognizes this fact. It is pursuing a parallel path, which attempts to use layout and design techniques to achieve specialized performance in (any) purely commercial process technology. The idea behind this Harden by Design (HBD) approach is that many--if not all--technology, performance, and harsh-military-environment survivability requirements can be met by primarily using the commercial supply chain. These HBD techniques have been proven technically sound for many applications and diff erent process nodes. They've been explored recently in sizeable government programs. They also have been applied for many years by radhard designers, who look at "sizeable programs" and wonder what all the fuss is about. ("Of course edgeless devices don't have edge leakage.")
One unfortunate diversion in the HBD movement has been the association of technology feature size, such as 90 nm, as a standard by which to judge our capabilities in electronics. DoD electronics initiatives require figures of merit that are derived from diverse military system requirements. Advanced systems win battles--not advanced lithography. But the attention that design hardening is receiving at DARPA and other organizations is promising. It may represent signs that the DoD is starting to consider such approaches as the basis for a modernized DoD electronics procurement methodology that's compatible with the fabless or pure-play foundry model.
A step further in the fables-model direction is found in an Air Force program called Systematic Hierarchical Approach to Radiation Effects (SHARE). SHARE brings together a DoD system integrator (L-3 Communications), an EDA tool vendor (Silvaco Data Systems), and two domestic foundries ( Jazz Semiconductor and American Semiconductor). Together, they develop infrastructure consisting of process design kits, libraries, and affordable tool flows that target the process technologies that are of particular interest to the military-electronics community. This program is demonstrating that DoD system integrators and key elements of the modern supply chain (i.e., tool vendors and foundries) can provide solutions to very specialized DoD electronics needs. The SHARE program asserts that the DoD can and should operate according to Figure 1(b).
The radiation-hardened electronics community is observing with interest the recent Trusted Foundry Program. Th is program, which was jointly funded by the DoD and NSA, established take-or-pay access to IBM technology over a 10year timeframe. It leverages IBM's commercially supported manufacturing capabilities and essentially pays for wafer runs. Although it's a step in the right direction in terms of leveraging commercial manufacturing infrastructure, the trusted foundry program must be supplemented with trusted design kits, tool flows, design libraries, packaging, and assembly (i.e., a trusted supply chain that incorporates multiple on-shore foundries, encourages design expertise in those foundries, and forms a basis for standards in system integration that adhere as much as possible to a pure-play business model while accommodating a wide range of DoD requirements, such as radiation, temperature, and security).
The DoD has invested in a range of approaches to preserve on-shore military-electronics capability. The results are that radiation-hardened technology is available at BAE and Honeywell.Wafer runs in advanced IBM technologies have been provided through the Trusted Foundry Program. HBD programs have (re)validated numerous basic technical approaches to using commercial process technology for military applications. Finally, the SHARE program has combined military-systems expertise with simulation tools, military-specific design fl ows and PDKs, and commercial foundries. Solutions to the larger DoD electronics supply-chain problems can be found by combining these programs.
BAE, Honeywell, IBM, and other foundries must be included in the trusted-foundry concept. Affordable tool fl ows and specialized design kits must be developed and maintained for commercial as well as military processes at trusted foundries. In order for this infrastructure to be effectively used by systems integrators (DoD primes), these organizations must learn to more effectively manage outsourced operations. Th is means understanding process technology, design kits, design reuse, and new methodologies for specifying part performance requirements. Unlike past performance specifi cations, new part requirements must include details relating to technology, design methodology and simulation, design tools, fabrication, qualifi cation, and controls.
Requirements for continuing government investment will be minimized if the fabless model is followed and allowed to influence standards in how systems are procured. DoD use of the fabless model means changing not just the design process, but also the piece-part test and qualifi cation, requirements flow-down, and systems integration. The government needs to encourage this approach through the creation of system- performance acceptance criteria that are compatible with outsourcing design, layout, and other technical activities. In order for the radhard community and other specialized DoD end users to benefit from this in the long term, we must maintain an on-shore supply. This issue is not only technical; it is economic and political. It impacts all military and aerospace electronics. This goal can only be achieved if electronics is made a national priority.
"Assured supply of trusted microelectronics components for defense systems use requires actions well beyond the scope and magnitude of those that can be mounted by a single defense supplier, or by the entire defense contractor base. Addressing this problem is a uniquely government function. DoD is charged with the defense of the United States, a mission which depends heavily on microelectronics. The task force considers DoD the logical steward to lead and foster a national solution to this critical problem, regardless of which arm of government must act." These assertions in the February 2005 DSB report that electronics become a national priority are essential for even a progressive approach that leverages the fabless model to be eff ective. Without this type of United States government involvement, offshore migration of the semiconductor industry will extend from manufacturing to other elements of the electronics supply chain including design, simulation, and education. The DoD electronics community will then be left with little domestic infrastructure to leverage.
References:
[1] http://www.acq.osd.mil/dsb/reports/2005-02-HPMS_Report_Final.pdf
[2] http://lieberman.senate.gov/documents/whitepapers/semiconductor.pdf
[3] http://www.defenselink.mil/pubs/space20010111.html
[4] Sheehy, J., Sampson, S., and Hardt, H., "The Hardened Electronics Availability Problem (HEAP): An Evaluation of the Impact of the Defense Drawdown on the Viability of the Radiation Hardened Electronics Industrial Base," 1994.
[5] Dupont, D., "Nuclear Explosions in Orbit: The Spread of Nuclear Weapons and Ballistic Missiles Raises Fears of Atomic Attacks on the Global Satellite System," Scientifi c American, p. 100, June 2004.
[6] http://www.fas.org/spp/military/program/asat/haleos.pdf
Randall Milanowski is a senior scientist with L-3 Communications Jaycor in San Diego, CA. ORIGIN OF THE PROBLEM
From the 1950s--when the modern semiconductor was invented--until well into the 1970s, the U.S. government was the major customer for the entire electronics industry. The earliest computers and memory chips were used by NASA. In addition, imaging technology was advanced by our intelligence communities. In these and many other cases, the technologies went on to create massive consumer markets. Nearly every major electronic advance from the solid-state radio to the most well- documented military electronic invention of all--the Internet-can trace a significant part of its lineage to federally funded research of some type. The 1980s brought a period of transition in which consumer electronics emerged as the dominant market force. With the end of the Cold War in the early '90s and the subsequent "dot-com" and electronics booms, the military went from being the driving market force in the industry to less than 1% of the U.S. marketplace. Along the way, nearly all of the integrated device manufacturers that supplied the DoD with advanced electronics left the military/aerospace marketplace. It's a simple question of economics. When DoD funding dried up, commercial, publicly held companies had a diffi cult time making a case to serve the military market.
Following the demise of military-specific IDMs came the rapid decline of the IDM business model altogether. Nearly every segment of the electronics industry has closed down dedicated IDM manufacturing facilities in favor of the fabless model. This new manufacturing paradigm is driven by the tremendous cost associated with modern state-of-the-art manufacturing facilities (billions of dollars). This cost is unsustainable for all but the highest-volume segments of the industry. This fact has led to the creation of specialized companies that have no finished products themselves, but serve as the manufacturer for ever-widening ranges of "fabless" consumer-electronics design companies. The aggregate volume generates enough business to support the huge capital and R&D expenses associated with (smaller numbers of ) semiconductor manufacturing facilities. Cost didn't just drive this model; it drove it offshore. Taiwan and the Peoples Republic of China control well over 70% of all fabless semiconductor manufacturing worldwide. This fact creates additional security- related problems for the DoD.
FIRST RESPONSE: COTS
Commercial consumer-product demands drive foundries to develop process technologies that are smaller, faster, and consume less power. The U.S. military, aerospace, and intelligence communities need many of these same traits in their systems. The concept of leveraging commercial-electronics infrastructure for DoD applications became part of the commercial-off -the-shelf (COTS) movement, which is a DoD attempt to meet system requirements using purely commercial parts. This approach is appropriate in cases where the commercial and military demands for performance are the same. Although this is true in some cases (e.g., hammers), electronics is another story. The military requirements for electronic performance and reliability often exceed what is available commercially. Even in cases where these technical requirements can be met, other requirements for non-obsolescence and security cannot be guaranteed in the rapidly changing, globally distributed commercial supply chain.
DOD PROCUREMENT METHODS AND POLICIES
DoD prime contractors and systems houses have historically relied on a simple but effective method of simply"buying parts" (one reason the COTS approach was initially attractive). They passed requirements down to an IDM provider of specialized piece-parts and had little involvement in the details of R&D, simulation, design, test, and qualification. Virtually all of the higher-level system-integration activities depended on this underlying infrastructure of specialized IDM parts suppliers.
Figure 1 illustrates how this model breaks down in the fabless model.
DoD program offices still specify requirements in much the same way that they did 20 years ago. Yet the approved standards (a must-have for the DoD) for doing business in the mode of Figure 1(b) aren't well developed in the big system houses. How does one carry out critical tasks, such as lot or part acceptance, when the basic elements of a part design flow can be diff erent corporate entities or -- in some cases -- individual consultants? How can system lifetimes of decades be supported by a supplier base that is changing daily? The increased complexity of the fabless model alone creates significant challenges for the systems integrators, who are accustomed to "qualified parts lists." The problem is made much worse for the DoD by the growing commercial practice of fabricating and packaging integrated circuits at offshore foundries. When security requirements demand U.S.-citizen participation only, even many U.S.-based companies and universities have become diffi cult environments for DoD electronics research and development.
A central issue for the procurement of military electronics is the ideal of trust. Our current policies dictate that for vital systems, the internal components must be designed and manufactured by domestic organizations. The solution that we advocate in this article involves increased acceptance of the pure-play semiconductor model within the military-electronics community. It also recommends the nationally directed preservation of onshore elements of the associated supply chain (see Figure 1).
TOWARD A SOLUTION
The radiation-hardened electronics community, which develops specialized electronics for missile and space applications, produced its own version of the 2005 DSB report in 1995. It was called the Hardened Electronics Availability Problem (HEAP) report [4]. Since that time, the community has been seeking ways to address the decline in the industrial base. It is instructive to examine some of the paths taken by the radhard community to address its own encounter with the diminished supply chain.
During the 1990s, most of the IDMs that served the niche "radhard" electronics business exited the market. The cost of maintaining special radhard fabrication processes far exceeded the return off ered by the post-Cold-War, low-volume requirements of radhard DoD customers. In response to the diminishing supplier base for radiation-hardened electronics, a Congressional-mandated Radiation Hardened Oversight Council (RHOC) was established. The RHOC collects system requirements for radiation-hardened electronics and manages the R&D investments for developing and producing these electronics in order to meet system roadmaps. One eff ort that's been identified and implemented through the RHOC has been to retain radhard IDM-like capability through government- funded (Title III and DTRA) capitalization of two CMOS foundries.
This approach is partially rooted in a long-held conviction that the DoD's most demanding radiation survivability requirements (relating to nuclear threats) can only be met through very specialized processing techniques, which the onshore commercial foundries are unwilling (unable) to support without large government investment. Government-subsidized fabrication capability is a costly approach. In this case, however, the DoD end users are dealing with inescapably serious issues. Although the fall of the Soviet Union dramatically reduced the large-scale, strategic nuclear threat, even a single high-altitude nuclear detonation could severely impact both commercial and military space assets [5, 6]. Such a detonation would impact critical infrastructure ranging from banking and finance to intelligence and conventional military infrastructure. Even if it is low-probability, this type of high-consequence event places irreducible radiation survivability requirements on certain DoD assets. It's understandable that large investments have been made in order to assure a supply of strategic (nuclear-hardened) technology. When one considers the economic forces that have driven commercial foundries offshore, however, the concept of maintaining on-shore fabrication solely at the government's expense seems unrealistic as a long-term solution for all DoD electronics needs.
The radhard community recognizes this fact. It is pursuing a parallel path, which attempts to use layout and design techniques to achieve specialized performance in (any) purely commercial process technology. The idea behind this Harden by Design (HBD) approach is that many--if not all--technology, performance, and harsh-military-environment survivability requirements can be met by primarily using the commercial supply chain. These HBD techniques have been proven technically sound for many applications and diff erent process nodes. They've been explored recently in sizeable government programs. They also have been applied for many years by radhard designers, who look at "sizeable programs" and wonder what all the fuss is about. ("Of course edgeless devices don't have edge leakage.")
One unfortunate diversion in the HBD movement has been the association of technology feature size, such as 90 nm, as a standard by which to judge our capabilities in electronics. DoD electronics initiatives require figures of merit that are derived from diverse military system requirements. Advanced systems win battles--not advanced lithography. But the attention that design hardening is receiving at DARPA and other organizations is promising. It may represent signs that the DoD is starting to consider such approaches as the basis for a modernized DoD electronics procurement methodology that's compatible with the fabless or pure-play foundry model.
A step further in the fables-model direction is found in an Air Force program called Systematic Hierarchical Approach to Radiation Effects (SHARE). SHARE brings together a DoD system integrator (L-3 Communications), an EDA tool vendor (Silvaco Data Systems), and two domestic foundries ( Jazz Semiconductor and American Semiconductor). Together, they develop infrastructure consisting of process design kits, libraries, and affordable tool flows that target the process technologies that are of particular interest to the military-electronics community. This program is demonstrating that DoD system integrators and key elements of the modern supply chain (i.e., tool vendors and foundries) can provide solutions to very specialized DoD electronics needs. The SHARE program asserts that the DoD can and should operate according to Figure 1(b).
The radiation-hardened electronics community is observing with interest the recent Trusted Foundry Program. Th is program, which was jointly funded by the DoD and NSA, established take-or-pay access to IBM technology over a 10year timeframe. It leverages IBM's commercially supported manufacturing capabilities and essentially pays for wafer runs. Although it's a step in the right direction in terms of leveraging commercial manufacturing infrastructure, the trusted foundry program must be supplemented with trusted design kits, tool flows, design libraries, packaging, and assembly (i.e., a trusted supply chain that incorporates multiple on-shore foundries, encourages design expertise in those foundries, and forms a basis for standards in system integration that adhere as much as possible to a pure-play business model while accommodating a wide range of DoD requirements, such as radiation, temperature, and security).
The DoD has invested in a range of approaches to preserve on-shore military-electronics capability. The results are that radiation-hardened technology is available at BAE and Honeywell.Wafer runs in advanced IBM technologies have been provided through the Trusted Foundry Program. HBD programs have (re)validated numerous basic technical approaches to using commercial process technology for military applications. Finally, the SHARE program has combined military-systems expertise with simulation tools, military-specific design fl ows and PDKs, and commercial foundries. Solutions to the larger DoD electronics supply-chain problems can be found by combining these programs.
BAE, Honeywell, IBM, and other foundries must be included in the trusted-foundry concept. Affordable tool fl ows and specialized design kits must be developed and maintained for commercial as well as military processes at trusted foundries. In order for this infrastructure to be effectively used by systems integrators (DoD primes), these organizations must learn to more effectively manage outsourced operations. Th is means understanding process technology, design kits, design reuse, and new methodologies for specifying part performance requirements. Unlike past performance specifi cations, new part requirements must include details relating to technology, design methodology and simulation, design tools, fabrication, qualifi cation, and controls.
Requirements for continuing government investment will be minimized if the fabless model is followed and allowed to influence standards in how systems are procured. DoD use of the fabless model means changing not just the design process, but also the piece-part test and qualifi cation, requirements flow-down, and systems integration. The government needs to encourage this approach through the creation of system- performance acceptance criteria that are compatible with outsourcing design, layout, and other technical activities. In order for the radhard community and other specialized DoD end users to benefit from this in the long term, we must maintain an on-shore supply. This issue is not only technical; it is economic and political. It impacts all military and aerospace electronics. This goal can only be achieved if electronics is made a national priority.
"Assured supply of trusted microelectronics components for defense systems use requires actions well beyond the scope and magnitude of those that can be mounted by a single defense supplier, or by the entire defense contractor base. Addressing this problem is a uniquely government function. DoD is charged with the defense of the United States, a mission which depends heavily on microelectronics. The task force considers DoD the logical steward to lead and foster a national solution to this critical problem, regardless of which arm of government must act." These assertions in the February 2005 DSB report that electronics become a national priority are essential for even a progressive approach that leverages the fabless model to be eff ective. Without this type of United States government involvement, offshore migration of the semiconductor industry will extend from manufacturing to other elements of the electronics supply chain including design, simulation, and education. The DoD electronics community will then be left with little domestic infrastructure to leverage.
References:
[1] http://www.acq.osd.mil/dsb/reports/2005-02-HPMS_Report_Final.pdf
[2] http://lieberman.senate.gov/documents/whitepapers/semiconductor.pdf
[3] http://www.defenselink.mil/pubs/space20010111.html
[4] Sheehy, J., Sampson, S., and Hardt, H., "The Hardened Electronics Availability Problem (HEAP): An Evaluation of the Impact of the Defense Drawdown on the Viability of the Radiation Hardened Electronics Industrial Base," 1994.
[5] Dupont, D., "Nuclear Explosions in Orbit: The Spread of Nuclear Weapons and Ballistic Missiles Raises Fears of Atomic Attacks on the Global Satellite System," Scientifi c American, p. 100, June 2004.
[6] http://www.fas.org/spp/military/program/asat/haleos.pdf
Mark Maurer is the Manager of the Government Business Division of Silvaco Data Systems based in San Diego, CA. He is an alumni of Central College in Pella, Iowa.
......................................................................
