You are hereWhat's the lifetime of a digital processor?

What's the lifetime of a digital processor?

By eric.verhulst - Posted on 19 March 2011

Printer-friendly version

Before this question can be answered, we need to refine the question. A simple answer could be that the lifetime of a processor is equal to its MTTF or Mean Time To Failure. This doesn't help us a lot because the MTTF is not a unique number. Leaving out the mean calculation, just like interrupt latency the Time To Failure is not a unique number. It is a histogram and the distribution will depend on the usage pattern.

So, let's restrict the search space. Assume we have an industry grade (-40°C to 85°C) standard microcontroller. Take your favorite brand and assume as well that it has been burned in so that we are in the safe flat bottom of the bath tube curve. We switch it on, use it and now we wait till it fails. We do this a statistically significant number of times, record the numbers and calculate the MTBF.   What can we expect?

If you think that the figure doesn't matter, because the MTTF will be so long, think again. Below 90 nanometer, things start becoming increasingly statistical and the MTTF could be as low as less than 10 years (if no precautionary measures were taken). Should we care? I think about a few applications where less than 10 years becomes problematic. Keep in mind that a MTTF of 10 years also means it can fail a lot earlier. So, your car might stop working, your pacemaker could start skipping a beat. Except throw-away consumer products, a lot of embedded systems would simply not be usable because we wouldn't trust them enough

So, let's take a simple example. Assume you have a control panel with a camera at your front door. Inside is a small 16 or 32bit microcontroller. It reads a small keyboard that activates or deactivates your alarm system but allows as well visitors to ring the bell. To make things better, we integrated some temperature and pressure sensors, so we have a little weather station. In your living room there is a second device. Connected wirelessly, it communicates with the outside control panel and sensors.

So, how long will it take before this set-up fails? What could fail first? The battery? The wireless transceiver? The LCD panel? The camera? The processor? Or the PCB and its connectors? Or will a malicious person hack his way in and gain access to your house in the middle of the night?

To be honest, we don't know the answer. As stated, there are a lot of dependencies. The technology and components used will make a difference. So does the usage pattern. And if the answer is less than 10 years, while hoping for 30 years, what should one change in the design? Can we develop it to last 30 years (99.9% confidence) even if the components have a MTTF of 10 years?

If interested, please let us know your answer. If you or your organisation is interested in researching this  question or in developing such a long live embedded system, contact us.



Syndicate content