DARPA's Trust in Integrated Circuits program has been hitting the news on and off over the past few months, ever since they released the details of the program back in December. The program, which aims to verify the integrity of the electronics that will underpin critical military hardware like the F-35 Joint Strike Fighter, could give the military, and defense contractors, a guaranteed way to check if their chips have been compromised.
If the program pans out and produces a real way to verify microprocessors, it’ll be interesting to see how a Trusted chip imprimatur will play with the Trusted Foundries program. (Trusted Foundries were set up to counter what the DoD perceives as a rising threat to defense microelectronics posed by the offshoring bleed in the semiconductor industry.) Every article about the DARPA program (including mine) maintains that there is no conflict between a chip verification method and the Trusted Foundries program.
Will a DARPA Good housekeeping seal of approval become a standard last step in the trusted foundry procedure? Will one of these Trusted Entities get knocked out of the ring, or will the government integrate them into one oversight body?
All of this brings me to the point: On April 9, Northrop Grumman announced that its Advanced Technology Laboratories (ATL) semiconductor plant, in Maryland, had just achieved Trusted Foundry status. In fact, their accreditation is Category 1A which is, as you'd expect, the highest level that can be awarded to a foundry. A scant two weeks later, they won multiple contracts for the F-22 Raptor. Northrop gets $252 million to design and manufacture the F-22’s communications, navigation and identification subsystems. That’s a lot of chips.
The programs themselves are not in conflict. The first question might be: if you have a way to “ensure” that a chip is pure, why do you need your own (more expensive) stateside fabs? Because you don't want someone reverse engineering your most mission critical circuits, like the stuff that goes into an F-35. That’s a no-brainer.
But there are an awful lot of cooks working on the soup.
Trusted Foundries by definition are onshore, and they go through an accreditation process that can only be called grueling; verifying a facility can take months to years.
Then you have your Trusted Designers, like Sandia National Laboratories’ microelectronics center. These guys design the chips, but their Paleolithic .35 micron fab is ill-equipped to produce chips for anything anyone needs these days, so they send the designs to a trusted foundry to produce.
But some people don't even agree that they're safe once they've hit the foundry. "Even domestically there may be problems," HRL’s Charles Henry Field told me (HRL is one of the official TAPO trusted foundries). You need trust all the way through the supply chain. Malicious tampering could happen all the way down the chain—what about the delivery truck?"
Comments (1)
We applaud the IEEE Spectrum for bringing such an important issue as Trusted Electronics to the attention of the larger community. What is lost in the comments on the blog is that the issue of trust goes beyond leading-edge chips.
True, Sandia National Laboratory has a 0.35 micrometer silicon fab that also produces MEMS, optoelectronic integrated circuits and discrete devices (Sandia also operates a III-V device and circuit fab). However, to say that Sandia's silicon capability is a "Paleolithic .35 micron fab" that is "ill-equipped to produce chips for anything anyone needs these days..." is not only gratuitous; it is also ill-informed.
Sandia's silicon fab is optimized for radiation-hardened, analog, and mixed-signal microelectronics, as well as custom digital ASICs. Sandia focuses on performance optimization through design. Analog circuits, unlike digital circuits do not scale in geometry when it comes to optimum performance. A 0.35-micrometer technology offers better analog performance because of better device matching, high supply voltages, and extended signal dynamic range, etc. Properly designed and fabricated, larger devices are more likely to continue to perform in extended operating environments including temperature, shock, and radiation. Sandia's fab has strong defense and space customer pull given the fact that it's one of the few remaining on-shore rad-hard microelectronic foundries and the only one producing chips hardened to survive the nuclear battlefield (radiation environments can be extremely noisy electrically). Again, the focus of Sandia's captive fab is on rad-hard components, as well as a range of other critical components for a variety of national security applications that commercial industry chooses not to produce or that cannot be obtained from trusted on-shore foundries.
The glamour in technical writing comes from describing the leading edge of modern technologies. In contrast, the professional designer asks "what works?" Before asserting what "anyone needs these days" one has to ask: "how many MIPS are needed for the application?" There are some very high-end applications for which absolute processing throughput is essential. For many other applications the extra margin provided by larger-geometry devices with their inherent higher operating margins are more important than the greatest absolute throughput. After all, a large number of commercial and defense systems don't contain leading-edge technologies--have you looked inside your coffeepot or toaster lately? Systems designers don't earn their keep based on how many leading-edge chips they can cram into their products; their challenge is to provide the best performance at the lowest cost at an appropriate level of reliability.
In the macho posturing about whose fab produces the smallest feature sizes, what is lost is that system designers need a wide variety of parts to achieve the functionality their customers require. In defense industries, and increasingly in commercial systems, the issue of trusted components is increasingly coming to the forefront. Theft of intellectual property (http://wps2a.semi.org/wps/portal/_pagr/103/_pa.103/248?&dFormat=application/msword&docName=P043702 ), counterfeit parts, and concern for Trojan Horses increasingly require trusted suppliers. What the nation's defense contractors require, and what the commercial sector increasingly requires, is a coordinated plan to supply all the parts that are needed.
To provide the best performance at the lowest costs for their customers, Sandia engineers design their systems using ASICs from their in-house 0.35 micrometer fab for those parts whose attributes Sandia's technology optimizes. In addition, Sandia engineers use external fabs such as the IBM Trusted Foundry for high-speed/high-density digital ASICs which those fabs optimize. While Sandia's facility has developed and delivered Weapons Reserve microelectronic products for over two decades, Sandia has also applied for DoD Trusted Foundry status which is expected to be completed by the close of 2008. We welcome the opportunity to extend the dialog to the larger arena of all Trusted Electronics.
Dahlon D. Chu, Senior Manager
Micro & Electronic Products
MS 1071, Sandia National Laboratories
Albuquerque, NM 87185
Email: ddchu@sandia.gov, Tel: (505) 844-0173
Dave Myers, Principal Deputy Director
Microsystems Science, Technology, and Components (MSTC) Center
MS 1071, Sandia National Laboratories
Albuquerque, NM 87185
Email: myersdr@sandia.gov, Tel: (505) 845-9563
Posted by Dahlon Chu | May 11, 2008 4:53 PM
Posted on May 11, 2008 16:53