Collectors and restorers of vintage automobiles place a high value on factory-original parts. Most cherished are parts that are still in perfect condition and still installed on the original auto. One tick down the desirability hierarchy are identical original parts taken from another vehicle – not as cool as the “matching serial number” perfection of the original unit, but very close. Restorers scour salvage yards and wrecks behind barns, sometimes even resorting to theft in order to obtain the one missing link that would complete their masterpiece.
If original iron is absolutely not available, restorers often fall back on what are called “New, Original Specification” (NOS) parts. As the name implies, these parts are generally built in a modern factory with modern equipment and materials, but to the original specifications. While they often lack the old-world charm of the true vintage article, they are often more functional than the original, built to better tolerances, and factory-fresh. NOS parts are particularly good for internal components that cannot be seen from the outside. No concours d’elegance judge is going to break down your engine, find a connecting rod attached to a piston that is NOS instead of original, and dock you points.
Electronics, as we know, are different. First, most of our semiconductor components don’t wear out over time, so there isn’t a big “replacement part” market for aging devices in the field. Second, we engineers have done such a superb job improving our products with each passing year and generation that nobody would want to go back and restore an old piece of equipment. It would be like wanting to spend time and energy to restore a 1932 Duesenberg SJ when one could own a far technically superior 2007 Toyota Prius instead.
OK, maybe not.
Are there, however, situations when it makes sense to have NOS semiconductor parts? Certainly there isn’t a big rush of people wanting to build factory-new TRS-80 replicas. (Oops, now that one guy in Australia is going to be offended and write us a flaming e-mail, telling us that his “Retro-80” project is almost complete – as soon as he can get the cassette tape interface working correctly. NOTE TO HIM: You’re done. The cassette interface never worked on the original units, either.”
What if you wanted to order, say, a few more units of a satellite that was built a decade or so ago instead of spending the extra tens of millions of dollars it would take to design a new, flight-qualified one. You’d call up your supplier, they’d determine that they didn’t have any more lying around in inventory, and they’d set about trying to build you a new one from the old specifications. If you were a government or military organization, there would be a lot of specifications. One of the things in those specifications might be an Actel space-qualified FPGA from several years back, like an RH1020 or RH1280.
These devices have been discontinued by Actel – officially becoming obsolete in 2006, and new design customers have moved on to their much more capable (and contemporary) families like RTSX-SU and RTAX-S. The “vintage” satellites, however, that are currently designed with RH1020 or RH1280 devices would require major re-design and re-qualification work to move up to a newer-generation FPGA. The total cost would likely run into the millions. The operational benefits would be basically nil.
Since there are an estimated 10,000 of these devices flying, it is quite common for a re-order to arise that requires these parts. Recently, BAE Systems announced that they will be once again producing these parts to meet that demand. For the systems integrators charged with building new copies of older-design satellites, it is a godsend. The industry-wide savings could run into the tens or even hundreds of millions of dollars. For BAE Systems, it is a good business opportunity, cranking some large-geometry, old-school, .8 micron devices through their 150nm capable fabs, and selling them for astronomical prices – somewhere in the range of $10K each. While that may sound like a big ticket for a small device, the economics of tooling up don’t work with smaller numbers. Also, this is really about the same price that these devices originally commanded from Actel. Practically none of the cost of a space-qualified device is in the silicon. The testing, qualification, and certification make up most of the tab.
“This is a case of ‘everything old is new again’,” says Timothy Scott, National Sales Manager for Space Components at BAE Systems. “These are small, legacy devices of 8K-12K-equivalent gates that have been widely used in space applications. BAE Systems was fabricating them for Actel when they were previously in production. Now, we’ve made an agreement with Actel that allows us to produce and market them to people building legacy systems – bridging the obsolescence gap.”
BAE Systems tested the idea of fabricating the old parts on their new line with a 256K Prom. When the proms came back working in first silicon, the company decided to pursue producing the old FPGAs – the industry’s closest thing to a NOS FPGA. “The parts are exactly like the old ones – we use the same masks, only now we’re producing a much better looking chip,” Scott continues. With the newer fab, we get nice, crisp lines. In testing, the parts perform exactly like the originals – with even slightly better radiation tolerance in some cases.”
To make life easier for systems integrators, the new parts (the RH1280 has been announced at this point) will have the same SMD (Standard Microcircuit Drawing) number as the originals. This means that companies don’t have to treat it as a new substitute part from a new supplier. The company says flight-qualified versions of the parts will be delivered in the second quarter of 2008 – in the same 172-pin ceramic QFP as before, or in bare die form.
While space-qualified FPGAs are a corner-case for most of the supply-and-demand economics of the industry, the situation does raise the question of obsolescence and ongoing support of devices designed into systems with long field lives and with long design lives. Hopefully, the kind of thinking that brings us this “new” FPGA family from BAE systems will prevail in those situations as well.