It’s Always the Electrolytics

A friend asked me a few years ago: “You are an electronics engineer aren’t you?” Somewhat hesitant to reply, I answered that I was. This is normally greeted with questions such as: “Can you fix my PC?” or “Can you unlock my mobile phone that the FBI has given up on?”. In an ever-electronic world, us electronics engineers seem to hold the keys to heaven. Believe me, if I possessed the divine superpower of turning water into wine, I would be sat next to Jacob’s Creek with a wine glass in hand, drinking it dry.

“It is my HiFi Amp. It is not working properly” he said. Me being an analogue expert, my friend had struck gold. “It makes a sound like someone running his fingernail up a metal comb and only one channel works”.

“It is bound to be the electrolytic capacitors” I casually said. They always go wrong. In fact I know that much of the consumer electronics industry thrives on these devices going wrong after a while. The kit fails to work so you buy a new one.

His HiFi was exchanged for a few tins of beer and within hours I had the top off it. It was indeed an electrolytic capacitor that had gone leaky – all over the PCB. This particular capacitor smoothed the voltage to some logic circuitry that switched a relay to connect the output stage to the speaker. The circuitry detected when the output stage had settled to within a few millivolts of 0V then connected it to the speakers through a relay. The faulty capacitor was smoothing the supply to this logic circuitry, so when open circuit, the logic was oscillating the relay at 50Hz – hence the metal toothed comb sound.

Soldering a large can-capacitor in parallel with the offending part and sticking it to the PCB with some No More Nails (other adhesives are available), the problem was solved and the beer was just beginning to chill.

A few months later my heating system stopped working. I could hear the boiler in the garage trying to ignite, unsuccessfully, then eventually giving up. “It is bound to be the electrolytic capacitors” I casually said to myself. I replaced the 3 electrolytic capacitors (for the grandiose cost of £4 including postage), slotted the PCB back into place, retreated 50 paces and restarted the boiler. Perfect ignition and heating soon followed.

Compare this with silicon. I looked up the reliability report of the LTC3891 a few months ago. It has a FIT (Failures in Time) of 0.32. This is failures per billion device hours. So if I purchased one LTC3891 and ran it for a billion hours it might fail. Moreover, this is at 55 degrees centigrade, not room temperature.

Another way of expressing FIT is MTBF (mean time between failures). Invert the FIT rate to give billion device hours per failure, then divide by (365.25 x 24) and this gives lifetime in years.

A FIT rate of 0.32 is an MTBF of 356 thousand years… and this is at a temperature of 55 degrees centigrade. If the temperature goes down, the mean time between failures goes up (yes it gets longer). This is outstandingly good.

Now, the Arrhenius equation states that failure rates go up as temperature goes up and it is exponential. As we know from COVID, exponential things get really bad, really quickly. However, with silicon we are starting with such a phenomenally low number that even at very high temperatures, the failure rates are still excellent.

Putting the FIT numbers for the LTC3891 into the Arrhenius equation shows that the MTBF at 125 degrees centigrade is still 4571 years. I appreciate this is a statistical number and statistics can be manipulated to make any numbers look good; however, this is still a very respectable lifetime.

There used to be a rule of thumb that stated it was bad design practice to operate a component above 100 degrees centigrade. However, this rule dates from the 1990’s when reliability was 1000x worse. Running components from the 1990’s at 125 degrees centigrade, using the above data, implies that these parts would last just 4.57 years, which is a completely difference scenario.

So we can see that, as long as the silicon is being operated within its specifications, it should be the last thing on the board to go wrong. The fault is far more likely to be caused by mechanical issues (dry joints etc) or, indeed, the electrolytic capacitors malfunctioning.

As I write, I notice my external hard disk has started to make strange noises and is slow to copy data. Its supply voltage is 12V, 1.5A coming from a wall cube. Putting an 8 Ohm load on the wall cube output shows a far from flat output voltage on my oscilloscope. I bet I know where the fault lies.