In reading about the Metrorail accident in DC I noticed this bit in the article.
I guess I’m surprised because we often expect 20 years lifetimes at most even in relatively bulletproof electronics like those in cars. It fascinates me that a system, especially a system exposed to the elements, could be spec’d to work reliably for 70 years. Is this commonplace in industrial electronics? How do you deal with issues like corrosion etc.
All depends how much money you want to spend. The electronics in cars can be expected to last ~20 years because that’s typically how long the rest of the car lasts; there’s no point in desiging the electronics to last any longer than that.
I will note that the [url=]Voyager 1 spacecraft* was launched in 1977; radio transmissions were received from it in 2006, when its electronics were almost 30 years old.
The chips in your desktop PC are designed for relatively high performance in a relatively compact package; they’ve got pretty skinny traces which are subject to corrosion and other killers (I vaguely recall hearing that molecular diffusion is a problem on these scales, too). Want a chip to last longer? Make the traces bigger. Provide better cooling. Pot it to seal it from the elements. Suspend it to protect it from vibrations.
All of which costs money. But in the Metro case, it seems they decided it was cheaper to (try to) build a relay that lasts 70 years than it was to pay someone to replace them every 5-10 years.
Relays are rated by how long they can be energized and how often. A solid state relay might be rated for 100,000 hours energized which is over a decade. In the field of course there would be a duty cycle to consider. If it were a solid state relay switching a few times a day without being subjected to extreme temperatures I could believe one would last that long. Personally I’ve never seen such an application and don’t know how often the relays in this case were cycling or at what load or temperature. I have mechanical relays at work that switch every second or two 24/7 and they last for years so, if the design is robust it’s not too far-fetched to expect solid state relays to last a very long time. We would need to know more specs from this application I think to really know if that’s reasonable or not.
My understanding is that as current flows along a wire, the electrons knock loose single atoms from the wire’s structure at random. After enough time and enough current, the wire is swiss cheesed and you will get a short circuit. The thinner the wire, the faster it gets swiss cheesed. Which is why there are still electronics in use today which are 50+ years old (such as nixie tubes, with the originals still in constant use in antique elevators) – they used much thicker wires. Miniaturization comes with a price. So yes, I think it’s possible to realistically design electronics which will last decades or even centuries as long as there are no moving parts and the wires are thick enough to withstand swiss-cheesing.
I was going to mention Voyager as well, both V1 & V2 are still in constant communication with JPL 32 years after launch! Their computers and electronics are still very much alive although the RTGs are slowly dying and they have to ration available power to selected systems as the output decreases.
There’s a difference between having a service life of 70 years and having a service life of 70 years without maintenance.
Lets look at your car example. A car that last 20 years probably has a rebuilt tranny, engine, has gone through hundreds of oil changes, lots of replaced parts, etc. At the end of its lifespan it may only be the same frame, basic electronics, and interior as it was on day one. Sure its the same car, but techinically its a platform that requires a lot of maintenance and replacement parts.
This component probably has a maintenence schedule of replacing x, y, z at times a, b, c. It lasts 70 years, but only because its designed to be taken care of for that long.
Fascinating, I never knew this was an issue.
This is known as electromigration, and it isn’t a serious issue for any current electonic devices. This caused problems with some ICs, but the problem has since been resolved.
As a nitpick, “swiss cheesing” won’t cause a short, but a break, which is pretty much the opposite of a short. A short circuit is when you have a conductive path where you shouldn’t have one; a break is when you don’t have a conductive path where you should.
I’ve never heard of electromigration causing issues in conductors. Only at the junctions in ICs, diodes, etc. Maybe the process you are describing isn’t properly called electromigration, but in any case, I’ve not heard of this.
Do you have any cites I can check out about this causing issues for conductors?
There are thousands of radios in existence that were built in the 1920’s and continue to work. The capacitors and tubes have usually been changed, but most still use the original resistors, inductors, wiring, and speakers.
Having said that, an electronic component is made from a hodge-podge of materials, and the functionality of a component is dependent on the properties of the materials. Unfortunately, many materials degrade over time. The degradation is exacerbated by changes in temperature, humidity, vibration, radiation, etc.
A component that is hermetically sealed and is not subjected to vibration and large changes in temperature can last a very long time, depending on the type.
A couple of links:
Electromigration Failure in Au and Joule Heating Induced Oxidation in Cu Conductors
We and our fab have reliability experts who can predict the reliability of a certain design and process. I’m not one of them - one of my jobs is to figure out what is really happening - but I can give the latest issues if anyone cares.
There is a tradeoff between performance, density, and reliability. For normal ICs you don’t want to have too much reliability, since then you will have a slower, simpler chip, and very few people care about a microprocessor in seven years when it will be obsolete. On the other hand I once did work with Sandia National Labs, which has its own fab for building chips with considerably more conservative design rules. They had to make their own microprocessors, since no commercial company was interested in anything that old. That is where you get the kind of chips that went into Voyager (no relation 
)
I have several piezoelectric quartz crystal units dating from WWII that will work today; quartz has no inherent life span since if has nothing that will wear out. The frequency will change due to leakage of the container resulting in contamination on the quartz but the quartz itself will not change. Given a perfect package (which doesn’t exist) and a perfect power source, a quartz crystal unit will operate at the same frequency forever.
Exactly. My dad made the latter part of his career out of adjusting, cleaning, and repairing railroad signal relays. There are doubtless still many “WEF” tagged relays in the bungalos of the (then) D&RG.
I’m pretty much the only person I know that checks for trains when approaching gated crossings. Takes just one failed relay and the gates won’t function. Dad got called out to help trouble shoot them 2-3 times a year, and that was on a fairly small RR.