Re-capping an old ATX power supply

[HwAoRrDk]

I'm in the process of re-capping an old ATX power supply. I decided this one is worth saving because it is early ATX standards compliant, and features the full complement of voltage rails that includes -5V. Bizarrely, though, the PCB silkscreen proclaims it is "ATX Ver 2.3", but there's no way it can be, because that standard revision wasn't even created until 5 years after this PSU was made. Perhaps a typo, and it is meant to say "1.3".

Anyway, one of the main points of evidence that this PSU was getting tired was bulging electrolytic capacitors. What intrigues me though, is why one particular cap was so much more bulged than the others. I managed to track down a datasheet for it (G-LUXON LZ series) and it seems to be one that is more highly specified than any of the others in the PSU: 3000h @ 105C, high-frequency low-ESR (ESR is specified at 100kHz rather than the usual 120Hz). So why has it degraded the worst?

I traced the circuitry, and it is directly on the output of the +5VSB rail. I guess this means it is likely to have been powered up for far longer periods than the other caps, but is this the sole reason? I don't think it would see anything in the way of transient loads. Would the ripple current rating come into play? It is rated for 1A.

I'm thinking of replacing it with something with a longer lifespan rating - maybe 5000h. Worth it?

[wraper]

99% of LOW ESR caps are specified at 100kHz. Lifetime spec without ripple current rating means nothing.

So why has it degraded the worst?

Stressed the most, heat source located nearby or simply quality of capacitor was lower than others. 5VSB rail is working all of the time, other rails only when computer is ON.

[HwAoRrDk]

Hmm, maybe it is just power-on hours. There's nothing hot near it (e.g. heatsink) when the PSU is on - and it's even right in the path of the fan airflow.

Speaking of the fan, the PSU's label proudly states it has "thermostatic fan control". But looking at the circuitry, I can't decide whether that's a lie or not. The positive terminal of the fan connection is straight to the +12V rail, and the negative has a large (1W? 3W?) power resistor in series to ground. Markings on the resistor seem to indicate it's 31 ohms, but measures as a dead short. Could it be an NTC thermistor? I've never seen one in such a large axial resistor package.

Edit: I made a mistake and traced the circuitry wrong on the PCB. The large power resistor is nothing to do with the fan. It seems to be providing a minimum load on the +3.3V rail (and actually measures about 4 ohms). The fan negative is in fact connected through a jumper wire straight to ground. So the label is a lie! There are unpopulated components - a zener diode (of which the wire jumper is in place of), a transistor, a couple of resistors, a capacitor and a component right next to one of the heatsinks labelled "SNTC" - that appear to provide capability for that function, though.

[wraper]

Hot does not mean heatsink. It may be a diode, transformer or resistor.

[jh15]

From the days of the Bad capacitor plague?

[HwAoRrDk]

Just before. I think it was made in 2002.

The capacitor plague was around 2005-06 wasn't it? At least, that's when I remember having to deal with a bunch of Dell stuff that was faulty because of it.

[jh15]

I bought some Epox motherboards around Sept 2001, and they all suffered. Also, I have recapped boards for industrial use, from around that time a few years later, so as to keep a running system in place without  starting over.
     

[wraper]

You should probably get rid of it and buy new, 20 years is a lot. Due to modern PSUs being significantly more efficient (unless you buy trash), likely over the years you will save more on electricity than money spent on new PSU. At early 2000's there were plenty of trash capacitors. But even decent ones will eventually fail, you cannot blame them after PSU lasted almost 20 years.

[HwAoRrDk]

Well, of course a new PSU will be more efficient, what with the advent of mandated active PFC, 80+ certification, etc. But you can't buy PSUs like this any more. As I stated originally, this one is still useful because it supplies -5V rail (modern ones don't), so it can run systems with an ISA bus. I plan to use it for such things.

Actually, I tell a lie, I do recall seeing a couple of years ago some US company that was making new PSUs to old standards - e.g. AT, ATX v1.x, etc. - but they cost a fortune (IIRC, like $200+). I'd rather spend £5 on some new capacitors and keep this one.

[janoc]

Well, of course a new PSU will be more efficient, what with the advent of mandated active PFC, 80+ certification, etc. But you can't buy PSUs like this any more. As I stated originally, this one is still useful because it supplies -5V rail (modern ones don't), so it can run systems with an ISA bus. I plan to use it for such things.

I believe the only thing that ever used -5V on ISA was an old SoundBlaster cards and some obscure networking cards. If you really really need that, it is easy to generate that voltage rail with a small inverting regulator.  Otherwise an ISA system will run just fine without it.

See e.g. here for description: https://x86.fr/atx2at-smart-converter-negative-rails/

That's really not a very good reason why to keep such an ancient supply.