Old power supply issue -- any ideas ?

You can start by disconnecting the fans and seeing how the voltage holds up . If the loading was causing some kind of overtemperature foldback, I don't see the rectified voltage holding up at 30V. Note the nominal 20VAC peak r ectifies to sqrt(2)x for nominal 30VDC.

Is it a 20-0-20 output rated at 2A per output? (an 80VA transformer)

Okay, so about 4W of fans.

That is normal. ~sqrt(2) * AC voltage minus a diode drop, (AC voltage will be a bit higher with light loading than at full load).

It should be dissipating about 6 watts.

Does the supply have a current limit function? If so, and it's set too low, you'd see that.

What happens if you disconnect one or the other of the fans?

Best regards, Spehro Pefhany

When you say the regulator is becoming very hot, do you actually mean that the heat sink is? If the regulator is becoming much hotter than the heat sink, then the problem may be that the heatsink is no longer in good thermal contact with the regulator.

Sylvia.

if the heatsink isn't also getting very hot it sunds like the regulator is incorrectly mounted or possibly the regulator booster transistor has failed, all that anyone can really be sure of is that youu awr describing a broken PSU.

if you can't tap +12V from somewhere in the server buy a cheap 12V powersupply, either reatil or online,

Brand name and model?

Have you measured the output ripple voltage? Could be a stability issue under the pulsed load presented by the fans. Stability may degrade as tempertures increase/reduce.

You say you're adjusting the previously fixed 15V to 13V. Is this adjustment made when loaded or unloaded?

Some common linear circuits will limit earlier if the regulated output is adjusted low, without other compensation.

If you've lost a phase in the rectifier (+/-20V suggests half wave rectification, ratherthan fullwave), there will be increased input ripple and degraded load regulation. Dried out caps can also cause input regulation effects and output stability issues.

If you were happy with it in the past, it's probably worth fixing, something that might not be as easy with a modern replacement.

RL

I am a bit puzzled by some of your questions. You say:

The transformer has a 20 - 9 - 20 V 2 Amp AC output, and the rectifier output is +/- 30 V DC, taking into account the diode drop sqrt(2) x 20 V. May I know why you say that it is half wave rectification, given that the input AC is a standard 1 phase AC ? I have adjusted the output voltage with and without the fans connected, and the result is the same, fans run at full speed for e.g. 45 minutes, and then slow down.

That is indeed a possibility, but careful visual inspection does not appear to indicate any leaking.

Hmm I'm not sure how you can assume it's the fans? Try a resistor as load maybe.

How about a nice Phihong wall wart switcher.

You can find them cheaper.

George H.

Possible, but they generally didn't use those caps in linear power supplies- they were consumed in motherboards where they got to eat HF ripple at high current.

This will save about 60% of the power being wasted by running a 30V supply into a 12V load.

Yeah? Well, make and model isn't such stretch.

It could have been a single 20V winding with a full wave bridge. In that case, you couldn't just lose a phase; it would be all or nothing.

If just one of the 20-0-20 windings is not being processed, you'd still get 30V from the single phase, but ripple would drop from 120hz to 60hz and various other problems would start to show up.

You mean 4 to 5 minutes? Thats electronic thermal time constant, but the reason for thermal stress and effect on regulation or limiting hasn't been identified.

Do you adjust the output voltage using a supplied on-board adjustment pot, or are you doing it on your own, using another method?

RL

Not to mention the fans will slow if the volts drop to 8.4v

that current being drawn by the fans is increasing, leading to the droop. "

You seem to have answered your own question right there. Hot and low voltag e outout = overload. There is a slight possibility that a zener on the ou tpuit is getting leaky when warmed up, but the odds of that are slim and no ne and slim is nowhere to be found. Usually when that happens and overvolta ge condition happneed to cause it.

The long and short of it is anything with moving parts is suspect first. On e of those fan motors is starting to bind up when warm. Actually most motor s are the opposite. When the bearing get shitty they run better when hot, s o I suspect you have taken their life cycle beyong that, to the point where dust and shit gets in between the bearing surface and the shaft. That is w hen most of them start exhibiting these tendencies. On the other hand it co uld just be so worn out that somehow part of the armature is starting to ru b part of the stator. You didn't mention exactly what kind of motors they a re.

If they are brushless slotless DC motors using drive chips, there may be so me electrolytics going bad. They don't get leaky but they f*ck up the signa l from the Hall effect devices or whatever it uses and makes the motor work against itself. Like a car engine that jumped time in a way. Of course tha t only applies to certain chips and configurations.

if you don't think it's the motor(s) take a car taillight or something that 'll pull about an amp and run it for a while. If the voltage doesn't drop, look at the motors.

load maybe.

Old brush motor fans for sure, i won't make that bet on electronically controlled fans (BLDC).

?-)

According to Maxim, it's more-or-less linear above a threshold.

Although given their (mis) use of the term "deadband" I have some doubts.

Best regards, Spehro Pefhany

All the ones I've encountered do.