It's look like no power of tl434
AT psu starting at free running mode, with out controlling of TL

biggieshellz wrote on 2025-07-15, 05:29:

... When I unplug the internal fan, so it's under absolutely no load, the voltages look a lot closer, but as soon as it gets any load on it, it falls over.

Be aware that most PSU's require a load (sometimes a fairly significant one) to properly regulate...

Do you see similar voltages with the mainboard connected?

Also, if you have an old/trashed (ie: nothing you can't lose) hard drive, try connecting it.
I have seen PSU's which wouldn't power-up correctly without the load of a hard drive (in addition
to the mainboard - probably related to the 12v rail which doesn't see much load on most mainboards.

DaveDDS wrote on 2025-07-15, 10:54:

Be aware that most PSU's require a load (sometimes a fairly significant one) to properly regulate... […]
Show full quote

biggieshellz wrote on 2025-07-15, 05:29:

... When I unplug the internal fan, so it's under absolutely no load, the voltages look a lot closer, but as soon as it gets any load on it, it falls over.

Be aware that most PSU's require a load (sometimes a fairly significant one) to properly regulate...

Do you see similar voltages with the mainboard connected?

Also, if you have an old/trashed (ie: nothing you can't lose) hard drive, try connecting it.
I have seen PSU's which wouldn't power-up correctly without the load of a hard drive (in addition
to the mainboard - probably related to the 12v rail which doesn't see much load on most mainboards.

Using a 12V heavy component, that is preferably very expendable, to generate a dummy load for testing is a rather common practice, especially on such older PSUs.

AFAICR, these PSUs require a minimum load in order for the voltage regulation circuitry to stabilize the power rail within an allowed voltage range. That might not be fully accurate. If I am not mistaken, googling for group regulated PSU might provide insight.

These old PSUs need more load on the 5V rail than anything else. Generally 1 Amp minimum. The 12V rail regulates around the 5V rail, but should also have around at least 0.1 to 0.2 Amps of load.

My method for testing these is 12V auto incandescent bulbs. For the 5V rail use a 12V 20-30W bulb. At 12V, this bulb will use somewhere a little over 1 Amp. For the 12V rail, 5-10W bulb should do (pulling around 0.5 Amps.) In fact, I use this method for initial testing all of my 5V-heavy PSUs. Once a PSU has been found to work OK, I often put it on my hand-built load tester - basically a bunch of sections of NiCr wire (open-type heating element). Does the same thing as the bulbs, but allows me to pull much higher currents from the rails (up to 20 Amps from the 5V rail and/or 30 Amps from the 12V rail.)

momaka wrote on 2025-07-15, 12:00:

My method for testing these is 12V auto incandescent bulbs. ...

I've never been a huge fan of using bulbs for dummy loads (but I do sometimes which it's the only choice)
The main problem with them is that they are designed to "come on fast" which means the resistance is quite
a bit lower when they are cold, and increases as they get to temperature.. in other words the resistance (and
hence the current drawn) changes more than it normally would)

This can have effects like: The PSU comes on, up to voltage and then goes into an oscillation as the load goes in/out
of it's required regulation range - fortunately this is usually visible as the bulb "flashing".

Very good as "current limiters" however - I use bulbs when testing low- draw equipment which might have a
short - as the current draw goes up so does in inline resistance (and also gives you a nice visual indication the
current in excessive)... FWIW

When I was designing the windmill controllers, there were times when you had to quickly "slow it down" when the
wind got too strong .. as you turned it out of the wind - this was done with "heavy load". Which means I still have a few
HUGE wire-wound heavy load 150 Ohm resistors (think 12x4 inches) - these make really good dummy PSU
loads (but most people won't have them)

DaveDDS wrote on 2025-07-15, 12:22:

I've never been a huge fan of using bulbs for dummy loads (but I do sometimes which it's the only choice) The main problem with […]
Show full quote

I've never been a huge fan of using bulbs for dummy loads (but I do sometimes which it's the only choice)
The main problem with them is that they are designed to "come on fast" which means the resistance is quite
a bit lower when they are cold, and increases as they get to temperature.. in other words the resistance (and
hence the current drawn) changes more than it normally would)

Correct.

Which is why I almost never use bulbs higher than 30-35 Watts - the "cold" resistance gets really low with the higher wattage bulbs, and that sometimes can trigger PSU short-circuit protection. On that note, halogen bulbs are even worse. But in any case, even with those, I can usually use up to 35W bulbs - at least on the 5V rail. The 12V rail being higher voltage, I stick to 20-25W or less.

DaveDDS wrote on 2025-07-15, 12:22:

This can have effects like: The PSU comes on, up to voltage and then goes into an oscillation as the load goes in/out
of it's required regulation range - fortunately this is usually visible as the bulb "flashing".

I don't think I've ever had that happen... or not quickly enough anyways.
PSUs are supposed to "crowbar" and not be able to restart after short-circuit protection kicks in until PSU is power-cycled.
Of course, a lot of the older TL494 / KA7500 -based (voltage-mode PWM controller half-bridge) PSUs don't always do that... so I guess you may indeed see the bulb flash in 1-2 second intervals as the PSU starts up, trips SCP, then restarts and repeats the whole cycle again.

DaveDDS wrote on 2025-07-15, 12:22:

Very good as "current limiters" however - I use bulbs when testing low- draw equipment which might have a
short - as the current draw goes up so does in inline resistance (and also gives you a nice visual indication the
current in excessive)... FWIW

Same here.
I use these on a lot of my low-power breadboard projects.
Really helpful if you mis-wire anything, so instead of magic smoke, you get a bulb lit.
Of course in the case of electrolytic caps connected backwards (Electroboom moments 🤣 ), the bulb still might not save you (0r rather, the cap that was connected backwards.)

DaveDDS wrote on 2025-07-15, 12:22:

When I was designing the windmill controllers, there were times when you had to quickly "slow it down" when the wind got too str […]
Show full quote

When I was designing the windmill controllers, there were times when you had to quickly "slow it down" when the
wind got too strong .. as you turned it out of the wind - this was done with "heavy load". Which means I still have a few
HUGE wire-wound heavy load 150 Ohm resistors (think 12x4 inches) - these make really good dummy PSU
loads (but most people won't have them)

Well, if you want high-power "resistors" that everyone likely has in their home: toasters, sandwich makers, space heaters/radiators, stove/burner elements (if you're in the US/NA and have an older stove with user-replaceable "coiled" burners), hair dryers, heat guns, coffee makers, clothes irons ... well, I could probably go on and on here.
But basically, any of these can be used as a high-power resistor, usually - just have to make sure they don't have any "smart" electronics or MCU-based controls, as those will get in the way.

momaka wrote on 2025-07-15, 12:46:

PSUs are supposed to "crowbar" and not be able to restart after short-circuit protection kicks in until PSU is power-cycled.
Of course, a lot of the older TL494 / KA7500 -based (voltage-mode PWM controller half-bridge) PSUs don't always do that...

The AT half-bridge starts with self-oscillations.
This should be enough to start the TL494.
When the protection is triggered, it goes into this very start mode (self-oscillations) with power limitation.
Fundamentally, it cannot completely switch off in case of a short circuit.
ATX with standup power supply - should, but AT - purely from the point of view of circuitry - cannot.

OK, I plugged in a known-good hard drive, and it's still not working. On the +5V rail:

Nothing plugged in: 5.15V
Internal fan only: 1.35V
Internal fan and hard drive: 0.78V

When you say "no power of TL434", what do you mean? Should I try replacing the chip (I ordered a spare just in case)?

I've attached a photo in case it helps.

Thanks!

momaka wrote on 2025-07-15, 12:46:

I don't think I've ever had that happen... or not quickly enough anyways.
PSUs are supposed to "crowbar" and not be able to restart after short-circuit

I wasn't talking about short circuit detection/protection.

I was referring to it providing enough load to properly regulate when it's cold, then dropping below that threshold
after warming up ... most PSUs will drop to a lower V - bulb cools, process repetrs (hence "flashing")

Not common, but I have seen it a couple of times.

Agreed on the space heaters etc. --- I just don't like keeping such big things on my workbench...
The resistors I have a big - but not that big .. plus they are built with metal stand-off ends, which
easily screw to the side of the bench - close enough to hook to, but not in they way.

DaveDDS wrote on 2025-07-15, 12:22:

... Very good as "current limiters" however ...

Alas, at least here in Canada - good old Incandescent bulbs are getting harder and harder to find as
lower-power LED bulbs have taken much of the market. (My wife thinks I'm strange(r than usual)
when we happen upon someplace clearing out "old useless" light bulbs and I always want to grab
a box (or two) of them.

I built a "test power station" which has an easily accessible outlet, large flat light switch (the kind you just
push on to turn on/off) and three light bulb sockets wired together in parallel, and then (the three) in series
with the outlet.

This lets me go in many ranges from a single 25w bulb to 3x100w which gives me a good selection of
"cold" current capability - and I have an easy "cut all power" switch.

biggieshellz wrote on 2025-07-15, 14:39:

When you say "no power of TL434", what do you mean? S

D14/C16
The most common circuit used is from the +12V winding of transformer like s
- Resistor 1-100 Ohm
- Diode
- Capacitor 1-47 μF * 50V to ground
- to the Vcc (12 pin) of the TL494 chip.
The supply voltage should be from 11 to 30 volts depending on the implementation.

OK, there is a 1N4003 type diode leading to a 47uF cap to ground, which is followed by a 220 ohm resistor to that Vcc pin (and a few other things). The voltage on one side of the resistor is ~7.6V and on the other side is ~7.3V; this doesn't change much whether the internal fan is connected or not. So I'm assuming this is not enough voltage to get the TL434 to turn on and start working? What's the next thing to check?

Thanks!

+12 circuit
2* 302 diodes (may be Schottky , in to220/267 case. Something like 120+ volt 10+ amper)

biggieshellz wrote on 2025-07-15, 14:39:

I've attached a photo in case it helps.

PS. Get rid of it - if you want to live.
Not in the sense of "hardware survival", but in the sense of "yourself".
This power supply has a common heatsink for the hot (net) and cold (consumer) parts. Despite the tricks with a fully insulated diode case and thermalpads, in the event of critical damage to the diodes or transistors, you will receive the entire AC power supply voltage on the PC case.
Even in those wild times, such nonsense was done by only very irresponsible manufacturers.

shevalier wrote on 2025-07-16, 12:43:

https://www.vogons.org/download/file.php?id=223446&mode=view PS. Get rid of it - if you want to live. Not in the sense of "hardw […]
Show full quote

biggieshellz wrote on 2025-07-15, 14:39:

I've attached a photo in case it helps.

PS. Get rid of it - if you want to live.
Not in the sense of "hardware survival", but in the sense of "yourself".
This power supply has a common heatsink for the hot (net) and cold (consumer) parts. Despite the tricks with a fully insulated diode case and thermalpads, in the event of critical damage to the diodes or transistors, you will receive the entire AC power supply voltage on the PC case.
Even in those wild times, such nonsense was done by only very irresponsible manufacturers.

I agree with that. Year 1999. my old 386 is getting restarted every few minutes, and i wanted to see if the fan is blowing out from this PSU, connected the hand with the case, and the next thing i remember i was on the other side of the room, my mom and dad are yelling something i didn't hear.... was veeeery lucky that automatic fuse has disconnect itself.....

butjer1010 wrote on 2025-07-16, 20:41:

I agree with that. Year 1999. my old 386 is getting restarted every few minutes, and i wanted to see if the fan is blowing out from this PSU, connected the hand with the case, and the next thing i remember i was on the other side of the room, my mom and dad are yelling something i didn't hear.... was veeeery lucky that automatic fuse has disconnect itself.....

This can be fixed by cutting off part of the radiator.
10 mm of air gap will be enough.
But this power supply has not been repaired yet.

biggieshellz wrote on 2025-07-15, 14:39:

OK, I plugged in a known-good hard drive, and it's still not working.

A single HDD won't be a good enough load to test, as I mentioned earlier. You really need a bigger load on the 5V rail.
Though given how low the 5V rail goes, I suspect this may not be the root of the issue... but rule it out, just in case. Use a 20-25 Watt 12V incandescent auto bulb connected on the 5V rail and see if that does it. If not, then there is indeed a fault somewhere else.

biggieshellz wrote on 2025-07-15, 14:39:

When you say "no power of TL434", what do you mean? Should I try replacing the chip (I ordered a spare just in case)?

No, don't replace it. These almost never fail on their own. Usually the 5VSB standby circuit will kill them - but that's on ATX PSUs. This is an AT design with no 5VSB, so the 494 PWM controller is pretty safe from that.

biggieshellz wrote on 2025-07-15, 14:39:

I've attached a photo in case it helps.

It does, thank you!

This is an example of a decently-built AT PSU from back in the days. I've seen much better, of course, but I would not qualify this as low quality by any means.

shevalier wrote on 2025-07-16, 12:43:

PS. Get rid of it - if you want to live. Not in the sense of "hardware survival", but in the sense of "yourself". This power sup […]
Show full quote

PS. Get rid of it - if you want to live.
Not in the sense of "hardware survival", but in the sense of "yourself".
This power supply has a common heatsink for the hot (net) and cold (consumer) parts. Despite the tricks with a fully insulated diode case and thermalpads, in the event of critical damage to the diodes or transistors, you will receive the entire AC power supply voltage on the PC case.
Even in those wild times, such nonsense was done by only very irresponsible manufacturers.

Actually, there's nothing wrong with this type of design *if* done right and with good quality parts.
FWIW, almost every industrial open-frame PSU I have seen uses a common (connected) heatsink for both primary and secondary. And I'm talking about good quality industrial PSUs from Mean-Well, 3YPower, TDK, and similar. And not only that - higher-power portable power adapters (e.g. laptop adapters) sometimes also use a common heatsink inside.

So this is not really a hazard... but again, only if it was done right.
All in all, it looks like a decent PSU. The 5V rail appears to have a 40 Amp 45V schottky rectifier in a TO-247 case - so certainly not a cost-cut PSU build. I can't tell what/where the 12V rectifier is, though. Could be that TO-220 part on the smaller heatsink by itself, but not 100% sure.

On that note, I agree with the troubleshooting steps you gave to the O/P to check the 12V rail rectifier, as 7V going to the PWM controller isn't right indeed.

This brings up the idea/question:
O/P, check if the PSU has an AC voltage selector switch and make sure that it is set to the correct voltage of your country - i.e. either 115V or 230V. I've seen this switch switched to the wrong voltage before and various issues can stem from that. Well, to be exact, if you live in a country with 220/230/240V AC mains, DO NOT move this switch to the 115V position, or you WILL blow up primary-side parts (the fuse, at the very least... though not uncommon to blow the primary caps and bridge rectifier either.) However, if you live in a country with 110/115/120V AC mains and if the switch is set to the 230V position, that could well be the reason you see the voltage dip. So do check that.

biggieshellz wrote on 2025-07-15, 05:29:

I completely recapped it, suspecting that was the problem, but it's behaving the same way.

Question about that.
Did you replace absolutely *all* of the caps, or just the bigger ones?
The reason I ask is because I see some good cap brands in there (Panasonic's with "T" vents and also Nichicon PM series), along with some really really terrible brands, like Jackcon.
I don't think your recap harmed anything and very likely did not make the original problem worse. However, if/when you get to fix this PSU, consider getting rid of those Jackcon caps, as they are absolute garbage. Same for some of the smaller no-name caps.

shevalier wrote on 2025-07-15, 13:46:

The AT half-bridge starts with self-oscillations. This should be enough to start the TL494. When the protection is triggered, it […]
Show full quote

The AT half-bridge starts with self-oscillations.
This should be enough to start the TL494.
When the protection is triggered, it goes into this very start mode (self-oscillations) with power limitation.
Fundamentally, it cannot completely switch off in case of a short circuit.

Good point. TL494/DBL494/KA7500 -based half-bridge PSUs indeed cannot fully turn off.

shevalier wrote on 2025-07-15, 13:46:

ATX with standup power supply - should, but AT - purely from the point of view of circuitry - cannot.

Well, there were a few AT PSUs built around UC384x PWM controllers, and those can crowbar after a short-circuit fault (or nearly so - they will restart, but the interval between restarts can be so long that it looks like the PSU has latched off.)

DaveDDS wrote on 2025-07-15, 14:57:

I built a "test power station" which has an easily accessible outlet, large flat light switch (the kind you just push on to turn […]
Show full quote

I built a "test power station" which has an easily accessible outlet, large flat light switch (the kind you just
push on to turn on/off) and three light bulb sockets wired together in parallel, and then (the three) in series
with the outlet.

This lets me go in many ranges from a single 25w bulb to 3x100w which gives me a good selection of
"cold" current capability - and I have an easy "cut all power" switch.

I've something similar too, except instead of light bulb sockets, I have 2 wall outlets (and a switch) wired in series. One outlet is connecting my DUT (device under test), while the other is for connecting my "current limiting" device (e.g. desk lamp with incandescent bulb or floor lamp with a high-power halogen bulb... or when heavier currents are expected/allowed, I use a 2 kW space heater or 700W/1.4 kW mini oven.) Yes, such large "current limiting" appliances can indeed take too much space on the bench (and I don't keep them there for that reason)... but sometimes, there's just no other alternative to such high power "resistors" - namely when testing ATX PSUs with APFC circuits, as those APFC circuit absolutely do NOT like low-power current limiting devices (e.g. 100W or lower light bulbs.)

momaka wrote on 2025-07-18, 04:17:

almost every industrial open-frame PSU I have seen uses a common (connected) heatsink

In industry, grounding is carried out using a steel strip around the perimeter of the building, to which all metal surfaces are attached using screws.
And there is a supervisory department/person that checks this.

momaka wrote on 2025-07-18, 04:17:

Well, there were a few AT PSUs built around UC384x PWM controllers,

This is Flyback

shevalier wrote on 2025-07-18, 06:56:

In industry, grounding is carried out using a steel strip around the perimeter of the building, to which all metal surfaces are attached using screws.
And there is a supervisory department/person that checks this.

But this has nothing to do with the industrial PSUs I'm talking about. Most of them come with 5 screw terminals: AC input (L & N), ground, and output (+) and (-). The ground terminal is highly RECOMMENDED, but OPTIONAL. It is isolated from the PSU output terminals, but you could reference one of these to it, if desired.

shevalier wrote on 2025-07-18, 06:56:

This is Flyback

You mean discontinuous forward design. (Flyback is just a term people commonly use for it, though despite not being technically accurate.)
Well, not always.
You can have both continuous or discontinuous designs with UC384x current-mode PWM controllers. The easiest way to tell which one you're dealing with is if there is a large inductor (usually toroidal, like on standard group-regulated ATX PSUs) on the output after the rectifiers (but before the output caps) or not.
No output inductor = discontinuous design **
Output inductor/toroid present = continuous design.

** except half-bridge LLC designs - these have no output inductor either... but then again, these won't use a UC3842-style current-mode PWM controller either, so not really worth mentioning this here. 😁