Sphere478 wrote on 2024-07-09, 16:54:
The truth is that electrolytic caps just don’t hold up even if they are the good ones sometimes, they all have a shelf life.
While true that electrolytic caps do have a limited shelf life (and also endurance), it's not nearly as short as some people on the internet (and here on Vogons) make it out to be. I don't know why this myth keeps on spreading that all caps die after X many years. That's not true at all.
The real truth is: high quality Japanese caps can last for at least 2 decades and often surpass 3 decades as well... sometimes even longer if they're not in a stressful application. My experience from restoring old Japanese audio gear is that most large(r) caps were actually close to pristine spec when pulled out for a test. It's the really small ones that tend to drift more in terms of impedance and capacity over time due to drying out faster. But even then, I found most of these to be OK. Only ones that seem to suffer more often are those very close to high heat sources. In CRT TVs and monitors, for example, it's often the small caps on the neck board that dry out and go high ESR. And on motherboards, so far I have not seen Japanese caps go bad, provided they weren't from a series known to have issues pertaining to heat sensitivity or long storage life. With that said here's a full list of the brands and series of Japanese caps known to suffer from this:
Rubycon MCZ (heat sensitive)
Rubycon MFZ (custom series for Xbox 360; 2700 uF parts in older series known to go bad after extended periods without power / electrolytic layer breakdown)
Nichicon HM, HN, and HZ series with H01xx through H05xx date codes (6.3V parts known to go bad with or without use; H05xx parts are from a transition year - some tend to be OK, while others aren't; 16V parts from H05 datecode seem to be OK overall)
United Chemicon KZG (don't like long periods of storage without use... by long periods, I mean a few years; 6.3V 3300 uF, 6.3V 1500/1800 uF, and 6.3V 820 uF parts all known to go bad with or without use, though they do seem to last a little longer with use; 16V parts seem to fare better/a-OK, but occasional failure still possible.)
United Chemicon KZJ (same as with KZG above, but seem to do a little better overall)
United Chemicon TMV (go bad after a while, no matter what, so AVOID this series)
Panasonic FL (soft rubber bungs on bottom, making them prone to leaking if physically abused with force or excessive heat from soldering)
Sanyo WF (go bad after a while, no matter what, so AVOID this series too)
And that's about all of them off top of my head.
Note the pattern: all of the above listed series are ultra-low ESR series, most of them emerging around or after the Pentium 4 / Athlon XP era. Anything Pentium II/3 and older will NOT use these series as they came after. Thus older gear with Japanese caps is typically safe from cap-related failures.
Sphere478 wrote on 2024-07-09, 16:54:but for what we use them for, I’ve often wondered about making pcb stacks that had a bunch of tantrums on them and maybe a low ohm resistor to chill out the resistance as a replacement for retro mobos to make them not need to be recapped so many times over the coming decades.
You won't need the low-Ohm resistors for this. The long "leads" of the stack itself will typically have a few mOhms of resistance, which will be enough. And moreover, most motherboard buck regulator circuits aren't usually sensitive and won't object to the even lower ESR of Tantalums and solid polymer caps. The exception is a few ECS and similar cheap boards that used a TL494 / KA7500 PWM chip for the CPU VRM (e.g. K7S5A, K7SEM, and etc.), because these actually are built to work with a minimum capacitance spec. Also, probably older socket 3/5/7 boards may fall in this category, as some of them used linear regulators throughout, and linear regulators may sometimes oscillate with too low of an ESR. But then, it's also worth noting that linear regulators don't really stress the capacitors that much either (exception being newer boards that have hot-running linear regulators that put high stress on the caps with the extra heat they generate). So recapping these old boards with a quality Japanese brand of electrolytic capacitors should provide another 2-3 decades of service, at the very least.
And all in all, you can't really avoid the use of electrolytic capacitors. PSUs will always have at least a few in them.
auron wrote on 2024-07-09, 19:20:
an ESR tester can give some data on that, with even in-circuit measurements being possible (capacitance usually fails in that case, but ESR can be still readable).
Anyone who's serviced motherboards will tell you this is a bad idea.
Motherboards tend to have multiple capacitors in parallel to get a larger overall capacitance and lower overall impedance / ESR. This is especially the case for the CPU VRM. If one capacitor has failed, you will NEVER see it, because the other ones, even with marginal ESR/impedance will mask the dead cap's poor/bad ESR. Unless you read the particular cap's datasheet and do precise calculations to see if the readings you get on the ESR meter actually match up with the overall parallel capacitance and ESR spec, you won't see a bad cap. And even if you do the calculations, you still won't see it. That's because capacitor ESR/impedance spec tends to vary quuuuite a bit with temperature. Mix in the 20% tolerance (which is typically only the stated tolerance for capacitance... impedance spec can vary as much as -300 to +50% at very high or very low temperatures, respectively) and it's pretty much an exercise in futility to catch a bad cap with in-circuit readings. Perhaps if a motherboard rail has only one or two caps and one is very badly failed, only then you *might* see it. But overall, don't count on it.
The best bet to mobo "bad cap hunting" is just to see what brands and series of capacitors you're dealing with and then go from there. Most non-Japanese brands are not worth a squat and should be replaced at this point, especially on the retro hardware we're dealing with that is 20+ years old now.
That said, even I have a few motherboards myself still with the original shoddy electro caps on there. But these tend to be low-stress, cool-running boards (a few Pentium 3 era boards, a Pentium board with a linear reg for the CPU, and similar.)
