Old AT type computer SMPS issue

[Chris Wilson]

I have a piece of industrial equipment (engine diagnostics) that runs the OS9 operating system on a bespoke Bosch motherboard. It has the old style AT switchuing power supply, and it's died. I am told I cannot use an adapted ATX supply as the switch on of another Windows based PC in the machine is reliant on the power supply in the bespoke OS9 PC being an AT type....

So, I need to either repair the AT supply, of find another one. I removed all electrolytics, one by one, and they tested OK. I then removed all semiconductors except diodes that would test in circuit. Again, they all checked out OK. I see 320V DC on across the series connected input filter caps. No noises, zero output on 12 and 5 V rails. I do not believe it's oscillating and turning on.

I am trying to get my head around how the damned thing initially fires up. It must have some means of getting the oscillator going until such time as it's generating output voltage, surely? Until then how can the low voltage circuits function? I have found a fairly similar looking circuit, although I do not have a schematic for this particular one of mine, which is an Enermax E201-PV. Hopefully the schematic is attached below. I am pretty new to reading something like a SMPS schematic, in the past if one failed I binned it! But I think start up must be something to do with the 47uF cap at C16, but I am clutching at straws here I have reasonably comprehensive test gear, can anyone suggest how to see if the start up circuit is functioning? I am aware of the potential to damage myself with 320V DC stored energy, and I believe I understand about ground leads on my scope and the mains input side being a no no. But please treat me as village idiot when replying

[Psi]

Looking at the circuit diagram i cant find the bootstrap resistor either

You should be able to use a regular ATX psu.
Can you explain a bit more about why you cant do this as i cant think of any reason.

An ATX psu can be wired to auto turn-on when mains is applied by shorting PS-ON pin to GND
If you need the AT style 230V power switch it would be a simple matter to wire that back in on the 230V side.
Simply cut a hole in the ATX psu on the internal side, insert a cable clamp grommet thing and put some 230v wire though.
Inside the psu wire it between the IEC connector and the 230v feed to the psu board.
Then put a 230V power switch on the other end. 
Since you have a dead AT psu you already have all that you need.

The only possible issue i see is getting a ATX psu that has all the negative voltages. I believe some are optional in the latest ATX spec so some newer PSUs are missing either -5 or -12V which will be an issue if your computer needs that.

[mij59]

Hi,

D14 provides power to the control system when the smps is on.
I think that  R5, R7 and tr1 do the initial startup

[fluxcapacitor]

You can run a regular ATX psu with an adapter :

http://www.ebay.co.uk/itm/ATX-to-AT-PSU-adaptor-cable-20pin-to-P8-and-P9-convertor-AT-converter-switch-/201103915988?pt=UK_Computing_Other_Computing_Networking&hash=item2ed2ba3bd4

[amyk]

Attached is a simplified schematic of the primary side of this type of SMPS. It starts up as a Royer oscillator - one transistor turns on slightly more than the other (R5/R7 provide the startup ), then the windings 2a/2b provide positive feedback to turn either T1/T2 on and the other one off, and increases the current through the driving winding 1 until it saturates, then the field collapses and the cycle repeats again going the other way until there's enough voltage on the secondary for the TL494 to start up and begin to drive the oscillator itself through Tr1.

I would begin by disconnecting the control circuit and getting it to self-oscillate first. It should make a low-pitched buzzing in this mode. As this doesn't seem to be happening with your unit, I would check the passives in the primary side oscillator.

Much more info on this design here:
http://www.electronicspoint.com/threads/pc-smps-startup-by-self-oscillation.18639/
http://www.badcaps.net/forum/showthread.php?t=22129

[SeanB]

Change T1, T2, R5 and R7 and it should start up. Change all the caps anyway, especially C7 and C8, and check the values of C9 and C10. These are the parts that fail, if not then you have a shorted turn in the transformer.

[Chris Wilson]

Thanks to all who have posted, I now have it running. A more detailed component level look at the board shows it is a bit different to the schematic I posted, but in the start up area similar enough for me to pinpoint what to pull and change, especially with SeanB posting his comments. I found a 330K resistor open, and I also found an open 820K across the input to the mains bridge rectifier, whilst poking about .

I am left with one issue, the 5V+ rail is slightly high at 5.2V, the 5V- rail is only -4.54V.     The 12V+ rail a bit low at 11.35V, and the 12V- rail is down at -10.80V This is measured with a single IDE hard drive as a load on the supply.

The output circuit of my supply differs entirely from the schematic and uses  LM339 and KA7500B IC's

I had previously checked the output caps on all the 12V and 5V outputs with my Peak ESR70 meter, out of circuit, and they showed fine.  The info on the KA7500B is at:

www.spelektroniikka.fi/kuvat/ka7500b.pdf

There' a trim pot on the board and it's currently set mid point. I can raise / lower all output voltages with it. Should i raise them a bit?

Any guidance as to why the 12V and 5V  are down much appreciated, thanks for getting me this far I have to say i have not measured a good PC SMPS to see just how accurate the outputs are...

[homebrew]

I am left with one issue, the 5V+ rail is slightly high at 5.2V, the 5V- rail is only -4.54V.     The 12V+ rail a bit low at 11.35V, and the 12V- rail is down at -10.80V This is measured with a single IDE hard drive as a load on the supply.

Looks perfectly fine to me ...
According to your schematic, only the 5V line is directly regulated. The negative voltages levels are normally not very important to PC systems (if not horribly ancient)

[tautech]

Chris, I think the output can be trimmed by R17 that is paralleled across part of the divider for pin 1.

BTW did you fix that welder?

[SeanB]

Good to see the transistors are still good, you often have one that leaks slightly at high voltage and then it will not start. Power the secondary side controller and it will start up and run, so if it never is powered off it will run with leaky transistors till the first power failure or it gets to the point where the leakage is too high. Low rails other than 5V are common, just add an extra 2A to the 5V rail and it will drop and the others will increase. The original IBM had a 2R2 25W resistor inside the case to draw the extra power, unless you had the optional hard drive installed which drew enough off the 12V and 5V rails to keep the power supply happy.

[Chris Wilson]

Chris, I think the output can be trimmed by R17 that is paralleled across part of the divider for pin 1.

BTW did you fix that welder?

I modded my earlier post  to make it clearer the schematic I have does not match the output stages of this supply, there's a trimmer pot, please see post above again.

I believe it may be fine and i was unrealistically expecting dead on 5V and 12V.

No, the welder control box is still faulty, I have borrowed a working box of the Migatronic agent and he's now going to look at my box, as he has a set up that allows it to be fully tested outside of the machine. I had to admit defeat before I made it worse still I am hoping he'll share what he finds, and if so I'll let you know. Sometimes my tenaciousness in trying to fix things myself results in me losing money, especially when it's gear I make a living with. Reluctantly I had to call in the professionals, but thanks again for your help.

[Chris Wilson]

Good to see the transistors are still good, you often have one that leaks slightly at high voltage and then it will not start. Power the secondary side controller and it will start up and run, so if it never is powered off it will run with leaky transistors till the first power failure or it gets to the point where the leakage is too high. Low rails other than 5V are common, just add an extra 2A to the 5V rail and it will drop and the others will increase. The original IBM had a 2R2 25W resistor inside the case to draw the extra power, unless you had the optional hard drive installed which drew enough off the 12V and 5V rails to keep the power supply happy.

OK, that's great, I was unrealistically expecting voltages near as damn it spot on. I'll button it up and try it in the PC than Thanks Sean!!

[David Hess]

The power supply looks a lot like an old Tektronix "high efficiency" resonate design with a self starting primary although they had a separate startup circuit which "tickled" the power stage and bootstrap power was provided through the current transformer instead of the secondary outputs.

Regulation may not have been all that good anyway with the auxiliary outputs which are only regulated through coupling.  Only the +5 volt output really matters.

I would have replaced any aluminum electrolytic capacitor that I removed and probably all of them anyway just because of age.  There is little point in putting an old capacitor back in.  I would also have replaced any paper capacitors and taken a close look at C9.

[Chris Wilson]

I take your point on changing all the electrolytics. I was really doing this as a stop gap. I do not have any immediate access to another AT supply and wanted to know if the OS9 PC it's for was OK. I can now state that it boots and powers up fine, and is on soak test. Likely as not I will find an un-meddled with AT supply and keep this as a spare. I have learnt a lot considering a few days ago a SMPS was a total mystery and for PC's a throw away item. That in itself has been very worthwhile. Thanks to everyone, and thanks for your huge contribution to the Tektronix forum, David, where I lurk agog, usually with my mouth open trying to follow some of the more challenging repair sagas.

[David Hess]

I just hate removing parts, especially inexpensive and likely worn out parts, to test and then reinstall them.  It doubles the wear and tear on the printed circuit boards.

Aren't AT power supplies still available?  I would still have a couple in service if the embedded lithium cells on my Asus P55T2P4 motherboards had not died.