• Please review our updated Terms and Rules here

AT PSU 12v low

J

justanotherhacker

Experienced Member
Joined
Mar 22, 2013
Messages
72
Location
London, UK
I have an old AT PSU I am hoping to use with my MTM "PC-Retro" kit build (replica IBM 5150) as it seems to be the thing I have on hand that needs the least modification for that job. Taiwan-built Western Systems one from the early or mid 90s. For many years it has been my bench-top "I just need 5v to try something" PSU.

Before committing it to a new use I decided to test it out.
- I'm more used to newer electronics (high cap failure rate) so I was worried about the capacitors. Popped the lid and they all look good as new. I tried an ESR meter and it indicates nice low ESR over all the main caps. I know this isn't absolutely foolproof but it has to be a good sign.
- 5v line is almost 5.2 volts unloaded. It doesn't drop much under light load but I don't have a good way of really loading it up. After all these years I think it's still good. Minus 5v line is similarly generous.
- Unfortunately the 12v line is 10.5v under no or light load - way out of spec! I would expect this to be low enough to cause problems.
- Negative 12v line has the same problem (-10.5v).
- Oscilloscope suggests that the outputs are well regulated with very little ripple, though, again, I haven't really loaded it up and I wouldn't expect to see the worst of a problem until a heavy demand was made.

I can't see anything obvious to adjust an output voltage - I've seen some older supplies (and some new non-PC ones) with little trimmers for this, but nothing here. It's all fixed-value components.

So I'd like to seek this group's advice:
- do I need to buy a proper load so I can test regulation better, or can I skip that step?
- do people think am I right in guessing that this might be some resistors degrading in the very small section that does the voltage-regulation feedback for the 12v line, and that it might be worth removing them, testing, and replacing with known-good ones? (I think I know just about enough about switchmode power supplies to stand a chance of identifying these. They may even be in places in the circuit that I can test in-situ...)
- do people think I'm right in guessing that it is not worth speculatively replacing capacitors if they have no signs of damage and measure low ESR?

Thank you!

(apologies if this is covered in an old thread - I did try search but didn't find something that matched this, maybe I did it wrong)
 
Thanks Stone, HD is a good idea. I have a box of old disks that aren't doing anything. It has few ports so I can't really load it up but it'll do better than my one watt resistor :)

I don't actually have a motherboard I care little enough about to risk connecting to an out-of-spec power supply... I didn't hoard enough stuff :-(
 
A large MFM or RLL drive would be ideal. So would an ESDI or old SCSI. The bigger, the better, as larger motors draw more current, i.e., 12V.
 
Some PSU's including AT and ATX require a load to start and other's require no external load to start but the output voltages will usually not be optimal, When testing PSU's i always load them up and leave them running checking the output voltages over a period of time, I typically use an old power hungry MFM HD + a couple of ide drives and old motherboard. Capacitors can look brand spanking new and still be dead / dying / shorted inside, Load the PSU up as much as you can and check the output voltages again. If the 12v line is still low which i suspect it will be i'd be looking to do a re-cap either in part or full.
 
Thanks for the advice Malc. This one is old and doesn't seem to care if loaded to start or not - I used it as a "don't care" bench supply for years - I know not to expect it to regulate accurately at zero load, but I don't know how much load it needs to start to regulate accurately. However it does seem like I've not been loading it up enough while testing. Based on your and Stone's replies, if I can get it going with an IDE drive hanging off every connector, I might understand its condition better.

The caps are old but not shorted and not gone high-ESR either. Nevertheless if you think recapping stands a good chance of sorting it out, I may as well try - it's an easy step.
 
Put a full load on the thing and then read the outputs. On PSUs, the most common problem is capacitors that have dried out (i.e. very high ESR), not shorted. Also check the state of individual rectifiers. While one gone open doesn't happen often, it can affect the output.
 
As already mentioned: You should load the 5V line somewhat (I remember the specification required 1A or something).

The reason behind that: Those power supplies actually only regulate the 5V line and all the other voltages are determined by the transformer ratio. If the 5V line isn't loaded at all, the other voltages sag a little.

Greetings,
Christian
 
Chuck: Thanks, I didn't think of the rectifier. Seems like it would be easy to test. Shall do that.

Cerker: Thanks also, there's a missing piece of the puzzle for me there. I knew not to expect the PSU to be well regulated at no load, which is why I put a little load on it, not having realised my stack of end-of-life IDE hard drives would make a better load. What I didn't know was that the 12v regulation relied on the 5v load rather than the 12v load. That makes me doubt my measurements.

I'm glad I asked here. I'm a little embarrassed I didn't know/think of some of this stuff, but I'm learning the gap between what I know and what I think I know... very valuable!
 
Further to this:

Rectifier checks out, but I'm glad I checked. So thanks for that.

THANK YOU Stone for suggesting loading it up with hard disks and THANK YOU cerker for correcting my misconceptions about load and regulation. With an HDD on all three Molex there was enough 5v load for the regulation to kick in even under the 12v load of the disk motors. Now 5.05 and 11.89... close enough is good enough. I am now inclined to use this without further messing with it.

(when I load the 12 even a tiny bit with nothing on the 5, as you can imagine, it goes way out. That was the mistake I made before.)

Really glad of everyone's help.
 
11.89V = -0.11V or less than 1% out. Normally, you probably shouldn't be concerned unless it drops below 11.5V or so--and even then, you're probably still okay.

Sometimes I think we're the victims of digital readouts. Had this been a plain old d'Arsonval analog "needle" movement meter, you'd probably say that the 12V line was bang on.
 
Yes, it's well within tolerance. I read the 5150 technical manual and it says the real psu's 12v had a +5/-4% tolerance so I'm not going to lose any sleep over 1%. And my digital meter is not the greatest, I probably shouldn't be quoting that second decimal like I trust it.

I agree digital meters' claims of precision can be a bit double edged. I'd love to have a really good analogue meter someday. Not so much for the sensible rounding-off as for when a value is gradually changing - then I think it's much more intuitive to watch the needle than to see a digital readout keep changing as it tries to keep up. A really good digital meter might display dead-accurate values with rapid update but my brain can't make sense of 'em any faster! A gently swinging needle I can just about cope with.
 
You must log in or register to reply here.
Back
Top