older calculators not in regular use - how often and how long to run them

[brouhaha]

(06-27-2024 04:34 PM)Rafa Wrote:  Not trivial at all. I also would be interested in a specific maintenance regiment.

TL;DR: I suggest powering up most electronic devices for a minimum of a week per year to keep electrolytic capacitors from having excessive oxide degradation.

The only components for which any on-time is required are rechargeable batteries, which have been discussed to death elsewhere (TL;DR: don't leave ANY batteries in an electronic device for a long time, rechargeable or otherwise), electrolytic capacitors, and ceramic capacitors.

The issue for electrolytic capacitorrs is that when they are manufactured, they go through a "forming" process, which gradually (by means of current limiting) allows a DC voltage across the capacitor to ramp up to 125% to 150% of the rated voltage. This builds up an oxide layer on one electrode, with thickness proportional to the votlage, which is necessary for the capacitor to work. When an electrolytic capacitor is not used for a long time, that oxide layer gradually breaks down. If allowed to break down to below a minimum level for the in-circuit working voltage, then the next time it is used, there will be "punch-through" due to the working voltage being too much for the reduced oxide thickness, and the capacitor is destroyed and will have to be replaced.

In HP calculators, the electrolytic capacitors are in the power supply circuitry. In general I would suggest that having them powered on for a week per year would probably be sufficient to prevent oxide degradation from becoming a problem. That week could be distributed throughout the year in lesser increments, or it could be all in the last week of December.

It is possible to "reform" electrolytic capacitors that have suffered oxide degradation using essentially the same forming process that was used at the factory. This can only be done if the capacitor has not suffered punch-through, leakage, or any physical damage other than the gradual oxide breakdown. There are some people that make wild claims about reforming capacitors being bad, but some of the capacitor manufacturers actually recommend it for capacitors stored for long times unpowered, and there is even a US Department of Defense standard for capacitor reformation.

The other major problem with electrolytic capacitors is seal leakage, allowing the electrolyte to either physically leak out, or dry out in place. This can happen due to use of the capacitor beyond its ratings, at its ratings for a long time at high temperature and/or high ripple current, or manufacturing error (e.g., the "capacitor plague"). There's no fix for this other than to replace the capacitors.

Ceramic capacitors also have some long-term failure modes, but they are much slower, and there's not really any preventative measure to take.

Of course, all of the components have failure modes over the long term, but the batteries and capacitors are both the fastest to become problematic, and the easiest to take preventative or corrective measures. Generally all electronic components have a design life of five years, regardless of claimed MTBF of of claimed years of data retention of memory chips. The failure rate of components goes through a bathtub curve, where there is a high failure rate during or just after manufacturing (infant mortality), a period of very low failure rate, and then due to aging, a return to high failure rate (the far end of the bathtub curve). Components having a five year design life means that the manufacturer expects that the far end of the bathtup curve is somewhere beyond five years from manufacturing, but that they do not make any legal assurances that the components will operate correctly beyond five years. Obviously all HP calculators made prior to 2019 are mode of components that are already beyond their design life.

The reason that some things have an MTBF rating of hundreds of thousands of hours is that MTBF is a statistical measure (or predicition) of the failure rate of a POPULATION of items, not of any individual item. A 250,000 hour MTBF doesn't mean that there is any expectation that ANY units of the product will work that long. It means that if you operate 1000 units for 1000 hours each, on average you expect four units to fail within that time. The MTBF is only applicable during the design life of the unit, so if the design life is five years, the MTBF might help you understand how reliable you can expect the unit to be within those five years, but it literally tells you NOTHING about the reliability of the unit beyond its design life.

With regard to EPROM/EEPROM/Flash memory that has claims of data retention of 10, 20, or even 100 years, that too is a prediction based on the measured rate of charge leakage, which is an exponential decay function, and which is HIGHLY sensitive to temperature (higher temperature = shorter data retentioin). If you were to call any of the manufacturers and ask them if they warrant that the chip will hold data for the claimed retention time specification, the answer is absolutely not.