help with ATX PSU conversion - please

[mab2]

hello,

I've done a couple of atx psu conversions before but my current one has thrown up a problem that I've never seen before:-

I started with a psu that was 'new' but had a tendency to go into oscillation (squawking) under load so maybe that was asking for trouble.

I was looking at the output from the transformer on the 'scope when it was under a small load and was running stable (nearly silent) and noticed something odd: the amplitude of the output modified square wave is about 20% higher for one half of the cycle than the other.

My first guess was a dodgy rectifier but they all seem to be OK.
Then I thought: maybe a short on one rectifier outout - but no.
Then I checked the voltages on the primary side capacitors - thinking one might be at a lower voltage than the other for some reason: 165v and 166v - not enough I think to explain a 20% difference in amplitude on the secondary side though.

so now I'm stuck - I can't see what would cause the difference in amplitude given that the transformer is driven via a capacitor anyway, other than an asymmetric load/rectifier problem which I'm thinking I've ruled out.

Anyone got any ideas of possible causes? Or seen this sort of problem?

Apologies for the lack of schematic & scope traces but I do them on paper and getting pics from my ancient camera into the PC is very cumbersome. If needed I can do it but it's likely to take me a day or so .

the mod is basically to get 28v out so I'm full wave rectifying the '12v' outputs of the transformer to give + & - 14 v; the 5v outputs have been removed, and the torroidal inductor has been rewound with two matched & opposed windings, if that helps.

thx

 

[rumpfy]

In my limited experience, these type of supplies either conduct on the forward current part of the cycle or the flyback part of the cycle. Dynamically, the supply source is different for each part of the cycle. You may be trying to get output from the forward part of the cycle for part of the time and the flyback part for the remainder of the cycle.
Hope this helps.

 

[mab2]

hi Rumpfy, thanks for the reply.

ATX power supplies use a push-pull system to drive the transformer symmetrically - hence my confusion as to why this one isn't symmetrical.

I'm going to try to cobble up an isolation transformer to try and see what's going on on the primary side.

I'm hoping there's someone reading who's done a conversion and/or has more knowledge about the primary side circuitry than me - otherwise I'll have to keep digging myself without electrocuting myself.

m

 

[MrAl]

Hi,

Do you have a schematic or can you find one and point out exactly where you measure this problem?
That might help figure out what is causing it.

 

[mab2]

Hi,

Do you have a schematic or can you find one and point out exactly where you measure this problem?
That might help figure out what is causing it.

OK, I've found this one online which is similar:-

http://www.pavouk.org/hw/en_atxps.html

If you look at the top-right of the schematic you will see the main power transformer, the rectifiers and the 'common inductor' (indicated by the dotted line).

If I put the 'scope on one of the 12v outputs of the transformer, I see a 'modified square wave' - i.e. stepping -ve, 0v, +ve, 0v, -ve, etc where the +ve and -ve parts are 20% different in amplitude. If I use 2 probes, one on each 'end' of the transformer, both ends are high step or low step together (if you follow me).

If I put the scope probe after the rectifiers and before the common inductor, I see a squarewave with alternating high and low steps (again with about 20% difference in amplitude). It's the same for both the +12v and -12v outputs.

thanks

m

 

[MrAl]

Hello again,

When you look at the output of the diodes, you see an actual square wave with different amplitude peaks, or is it that one half cycle is longer or shorter than the other as well as different amplitudes?

What do you see at the actual +12v output?

Also, what are you using for your loads, on each output?

Also, what is the power rating of this power supply? If you know the current ratings for each output too that would be good.

 

[mab2]

Hello again,

When you look at the output of the diodes, you see an actual square wave with different amplitude peaks, or is it that one half cycle is longer or shorter than the other as well as different amplitudes?

good question - I think the lower amplitude peak may be wider than the higher one - I say maybe because my scope is playing up at the moment and it's producing a 'double image' which makes it difficult to be precise, sorry.

What do you see at the actual +12v output?

do you mean after the inductors & smoothing caps? it's flat d.c.

Also, what are you using for your loads, on each output?

at the moment just a resistive load between the +12 and -12v (plan is to run it at 28v to charge a 24v battery) - it's only stable with a current in the region of 0.5 - 1.0 amps so that is where I'm testing it (can't get a useful 'scope trace when its 'squawking).

Also, what is the power rating of this power supply? If you know the current ratings for each output too that would be good.

on the plate it's rated 250w, +5v @ 25A, +12v @ 8A, +3.3v @ 14A and -5 & 12v @ 0.5A each.

I've removed the 5 and 3.3v components & and modified the feedback and protection ccts for the intended 28v output.

 

[rumpfy]

mab,
just a couple of things.
1. test set up; I have a couple of back to back 240/24 volt transformers about 50 VA. With this setup, I can probe around these mains coupled power supplies but not with much load. It is very useful for getting started and getting the thing sorted on the power input side. The transformers have a reasonably high series impedance so its possible on the isolated side, to short circuit the 240 volt and nothing drastic happens.
2. the 494 adjusts the output by varying the conduction time.
3. You are seeing peak voltages different on each side so that suggests a problem with the conduction of the main switching transistors. To check this you'll need the scope measuring the collector emitter voltage directly. So back to step 1.
4. You said you modified the original transformer. Is it OK?
5. You might want to put a 1 ohm or so resistor in the emitter circuit of each c2424 to check the collector currents. I have a tek564 with differential input but for differential measurements of low voltages like the volts across a 1 ohm resistor, the common mode interference is too much and the back to back transformers are the go.
6. you comment on your cro. needless to say, these circuits can be difficult sometimes to get a nice stable trace. The tek has terrific triggering, but sometimes has to work really well to get a good trace.
Good luck.

 

[4pyros]

at the moment just a resistive load between the +12 and -12v (plan is to run it at 28v to charge a 24v battery) - it's only stable with a current in the region of 0.5 - 1.0 amps so that is where I'm testing it (can't get a useful 'scope trace when its 'squawking).

on the plate it's rated 250w, +5v @ 25A, +12v @ 8A, +3.3v @ 14A and -5 & 12v @ 0.5A each.

Maybe you should not be doing this whare as the -12 volt side is only good for one amp.
The feed back ciruits may need a load on each to ground not just across both 12 volt outputs.

 

[panic mode]

if the rating says only 0.5A, than that's all you can get without problems...

 

[mab2]

Maybe you should not be doing this whare as the -12 volt side is only good for one amp.
The feed back ciruits may need a load on each to ground not just across both 12 volt outputs.

I have modified the output side by making the +12v and -12v outputs symmetrical - so I can draw a good 8 amps from both. I've also modded the feedback/overvoltage circuits as I no longer have 3.3 and 5v outputs.

Rumpfy,

Thanks again -

points 1. and 3.:- I have cobbled together an isolation tx - I don't have 2 matching ones like yours but have daisychained 3 transformer secondaries to get 115v (the atx psu can be switched for 115/230v so this is probably OK). Following on from MrAls shrewd question re the width of the pulses, I did look at the base-emitter drive waveforms, but so far the only difference I can see with my old 'scope is that the high side driver has a lot more noise and spikes c.f. the much cleaner pulses on the low side transistor - but was wondering if that was simply because it is the high side and by 'earthing' the emitter through the scope (I don't have a differential probe) I'm affecting the waveform.

point 2: - I think the 494 is working normally.
point 4: - I didn't modify the transformer but the 'common inductor' (downstream of the rectifiers and indicated by a dotted line in the schematic) - mine now has just two matched and opposite windings for my two opposite outputs.

point 6: The problem with my scope isn't to do with the noisy signals but an actual fault with my circa-1970 'scope (it's got a power supply problem that's causing a 50Hz wiggle on both the x & y deflection drives - so for the current application I get a double image of the actual trace). It ought to be easy to find but the cause must be in one of the high voltage supplies and I haven't traced it yet. The autotrigger is a bit flaky too which doesn't help.

point 5 - err, OK, that's still on the to do list...

cheers

m

 

[Mosaic]

Possible defective xformer (crack in the core) , or leak in the windings.

 

[4pyros]

Do they work fine by them self's?

 

[mab2]

Do they work fine by them self's?

yes, if you adapt the feedback and overvoltage circuits accordingly. The PSU IS actually working, in that I can get 10amps at 28.8v out of it - but it makes a horrible squawking sound and the output is 'noisy' as it's in oscillation - which I was trying to adjust the feedback and damping when I noticed the asymmetrc pulse issue - and I'm thinking it's reluctance to run stable is connected.

 

[mab2]

Possible defective xformer (crack in the core) , or leak in the windings.

umm.. I hadn't though of that - I guess substitution would be the way to check. Would that make the tx work asymmetrically?

 

[MrAl]

Hello again,

29 volts out of a 12v designed power supply? I hope you checked the output caps to make sure they had a rating that high or actually higher. They may only be rated for 25v or even less. If they are, they could be drawing too much current and it could be asymmetrical if it's more of a problem for some outputs and not others.

 

[rumpfy]

mab,
I think I garbled the bit about the 494.
What I wanted to say was that I 'thought' the 494 provides 2 equal width pulses so on each half cycle you get two equal time durations. As the 494 adjusts the drive period it does so in a slow manner. You shouldn't get one half cycle which is 20% different to the other.
The comments about squawking are interesting. In some application notes I've read, the control IC will do a cycle skip if the SMPS is running on a light load. I'm not sure if the 494 does this or not. So when the switch goes into cycle skipping, the noise from the transformer core changes.
It is possible the ferrite core is broken. My experience with ferrite cores is that sometimes the traffo is hard to dismantle without breaking a leg. Still find it hard to believe this causes a difference in timing for each half of the wave form.

 

[mab2]

Hello again,

29 volts out of a 12v designed power supply? I hope you checked the output caps to make sure they had a rating that high or actually higher. They may only be rated for 25v or even less. If they are, they could be drawing too much current and it could be asymmetrical if it's more of a problem for some outputs and not others.

well it's 29v out of a 24v supply really (+12 to -12) so not that big a difference in voltage, and yes I've replaced the -12v diodes, and the caps (low ESR) so they can handle the voltage/ current. But the problems are there with the output set to 24v so it isn't because I'm asking it to go to 29v.

Rumpfy - yes I think the squawking could be caused by the cycle skipping you mention - though if it's doing it at 10A, 29v it's certainly not because it's lightly loaded. In practice it's squawking under <0.5A (minimum load two 120ohm load resistors to gnd) and >1.0A. I'd like to solve the asymmetric pulse issue as I can't seem to stop the squawking, and I'm thinking it may be connected with the imbalance.

I'd be surprised if the transformer were damaged as there are no signs of damage on the PSU and I didn't mod the tx. Though you never know.

I still haven't put the 1 ohm resistor in to test the collector currents (work keep getting in the way), so I don't really have anything new to add yet.

Thanks for the help.

m

 

[MrAl]

Hello again,

Oh ok so you are taking your output across the +12 and -12 volt outputs, that sounds reasonable up to the point where you mention that you want to take 8 amps out of that. You did also mention that you modified the outputs to be "symmetrical" which sounds very interesting, but it would be good to know exactly what you changed to get to that point. Maybe a circuit diagram would help here so we can see what you did. It could be something simple.
In the schematic you showed they show an inductor in series with the -12v line on the output. If that is the same with your circuit, then that inductor would only be rated to work at 0.5 amps not 8 amps which is much much greater. That inductor might even burn up. So it would be good to know exactly what you did for the mod.

Also, did you try loading the +12 for 8 amps alone to see if that worked ok?
And if it worked ok, how about loading the +12 alone with an 8 amp load to ground and a -12v load 8 amps to ground (two resistive loads to ground) instead of going across the +12 and -12v outputs?

Also, can you find out where the circuit gets it's voltage feedback from? It may be the +12v only or the +5v only or something like that.

A simple question, did the power supply work with the right loads on the outputs before the modification you did? How hard would it to be to get it back to that state, for testing again?

 

[mab2]

Well, 8A if it can sustain that - If I can get it working I can see how it warms up and set the current limit accordingly - I expect it should cope with 7-8A based on a previous conversion of a similarly rated PSU.

I've cobbled up a rough schematic of the output stage as it is; the four L's on the left represent the transformer 2ndary, and the two L's (L5 &L6) in the ±12v lines are actually the common inductor. I did add secondary inductors between the caps but they were non Identical, so I removed them to see if they were causing the problem.

the voltage feedback currently only monitors the +12v line - as I'm planning on loading symmetrically I was going to see how well it worked that way.

I did try loading just the +12v to Gnd alone, but it's no better than loading across ±12v.

thx

m