April 29, 2008; Page B7
If you think you have problems keeping that five-year-old computer in your house up to date, imagine what life is like for operators of really old technology -- airplanes or communications systems, say, built in the 1960s and '70s. These machines use computer chips that have long since gone off the market, and finding replacements for them is a growing headache.
Peter Sandborn, a professor of mechanical engineering at the University of Maryland, wrote about the problem in a recent issue of IEEE Spectrum magazine, and chatted about it last week.
* * *
The sort of technology obsolescence you describe -- where does it have the biggest impact?
They tend to occur most often in safety-critical systems. The makers of these systems usually pay a huge amount of money to qualify them and certify them in the first place, to people like the [Federal Aviation Administration] or the military. And they are around for a very long time.
The city you live in probably owns a 911 system, and they probably spent a lot of money buying it from someone like Motorola. You as a taxpayer aren't really interested in buying a new 911 system every 18 months because the parts are obsolete.
Power plants, especially nuclear plants, are another example. The people who support these plants are concerned that if they have a failure in their electronics, they have to fix it with the same part, because if they don't, they might lose their certification. And this is true for airplanes too.
What technologies are the most problematic?
Microprocessors and memory chips are the worst. It was once estimated that 3% of these chips go obsolete every month.
Why are chips the problem? Why not brakes, say, or hydraulics?
The issue is how much it costs to put in place the factory that builds the thing. Chip fabs are a billion dollars apiece. But a machine shop that bends metal or makes brake shoes isn't a billion dollars; there are subcontractors who will do that. But there are no subcontractors who will build you a two-million gate microprocessor.
Why not just use a similar part?
The problem is that some of these systems, like airplanes and power plants, are highly certified, which means you can't change anything about them. Innocuous changes that shouldn't do anything at all sometimes, unfortunately, do. So people err on the side of being conservative.
For example, a memory company once sold its production line to another company. After the sale, exactly the same chips were being made by exactly the same equipment in exactly the same place, except with a new company's logo. But that was enough to force a Boeing or a Honeywell to consider requalifying the systems that used the chips, a process that can be enormously expensive.
So are there safety issues with obsolete parts?
We hope there aren't. Part of the problem is that we live in a world that is very risk-averse. If we weren't as risk-averse, we'd just use an alternative part and not worry about it. With some products you can do that, but not with airplanes.
Are there other ways to design systems that avoid this?
If you know a part will only be around for eight years, you commit to a refresh of those parts eight years hence. The alternative is to say a system will last 25 years, and then plan on fighting obsolescence problems as they crop up. The latter is what usually happens, because it is often cheaper in the short run, but not the long run.
It seems like military procurers can't make anyone happy. They are either knocked for $500 hammers, or knocked for using off-the-shelf parts that go obsolete.
Right. When the military got out of the chip business, and started to go to Radio Shack for them, this problem came along with it, and it's gotten expensive. It's the backlash to acquisition reform.