help with ATX PSU conversion - please

[mab2]

Oops! forgot to add the pic:

 

[MrAl]

Hi again,

Oh that's nice that you drew that because i think that helps.

From looking at that and assuming that the problem is because of something on the secondary, one thing that is interesting is that if one of the inductors (like L6) went into saturation, that would cause a huge current (or at least greater than the other) to flow into the caps and that could upset the flux in the transformer which would cause all sorts of problems as well as audible noise.

But you are saying that loading the +12v alone is causing a problem by itself, so maybe L5 is saturating. I wonder what happens with only the -12v line loaded?

Where did you obtain L5 and L6? If there is a problem with them then you'll see a similar action.

If you can not determine if they are saturating then you could try swapping them, or you could take them out entirely and replace them with maybe a 2 ohm power resistor (one each side). See if it runs like that.

Alternately, you could disconnect the diodes and connect power resistors across the secondary to test the circuit up to that point. That would tell you if the output circuit is really the culprit or it is really something on the input or the transformer itself.
You would look at the AC signal out of the secondary with it loaded appropriately. A good idea would be to solder the resistors into place for this test so they dont pop off and blow out the switching transistor.

 

[mab2]

L5 & L5 are a single toroidal inductor with two matched windings connected in opposite directions for the two outputs (I don't know how to 'draw' that on LTspice) It shouldn't saturate when operating within the supplies design current (in theory) as it's part of the original supply - I've just modified it as the original -12v winding was of too-thin a wire so I've rewound it with thicker wire, but the same no. of turns as original.

I can't test the -12v output by itself as the feedback is only on the +12v, so it doesn't react to the increased load on the -ve rail alone.

unfortunately I've had to focus on paid work today so I've not had a chance to do anything with it .

Thanks

m

 

[Mosaic]

Just thought I'd ask, but this seems to be a lot of effort for a relatively available part....lots of 36VDC 10A switching supplies on Ebay....for abt $32 incl shipping.

A 36V supply can charge your 24V batts with proper CC/CV regulation.
They have a +/_ 10% adj pot, so u can go down to abt 32V

 

[alec_t]

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

Does the tranny have two primary windings for its push-pull operation? Have you modified the primary side?

 

[mab2]

Does the tranny have two primary windings for its push-pull operation? Have you modified the primary side?

hi,

No I haven't modified the tranny - ATX psu's use the line power to charge two electrolytic caps in series; one side of the tranny primary is connected to the midpoint of the caps via an unpolarised cap, and two transistors aternately connect the other end of the primary to the + amd - ends of the capacitor pair.

Just thought I'd ask, but this seems to be a lot of effort for a relatively available part....lots of 36VDC 10A switching supplies on Ebay....for abt $32 incl shipping.

A 36V supply can charge your 24V batts with proper CC/CV regulation.
They have a +/_ 10% adj pot, so u can go down to abt 32V

Well firstly, there are a lot of projects on this forum that you could say 'why don't you just buy one?' but that kind of misses the point of this forum IMHO.
secondly, whilst there are some PSU's of that price on the UK ebay, they tend to be from china and of unknown quality (until you've bought them).
Thirdly, 32v is still too high so I'd end up modifying an unfamiliar type of PSU.
Forthly, I had an ATX psu doing nothing and usually they don't give me much trouble - though I admit this one's being a pain.

 

[MrAl]

Hi,

My 30v 10amp power supply set me back quite a bit, so i wont say go out and buy one, unless you find that you just cant do it any other way.

But now that we've had a chance to talk about this back and forth a bit, the picture is starting to get more and more clear. I think we are narrowing down what might be wrong, although it could take a little more effort in posting back and forth.

The basic circuit looks sound. Everything looks like a typical power supply, except for one thing. That's the "common" inductor. That's something we dont have much information on other than that it came with the power supply and you've added a new winding and removed the old winding. What this means is it would be very hard to figure out what to put in a spice simulation for that object, although every other object we could pretty much guess at what we could do, but even there we dont really need to because the basic circuit is quite simple.

So in this case my recommendation is that you remove the common inductor and replace it with something else, even if just power resistors for now, just for testing. Modifying something like that without knowing all of the effects that mod has on the entire construction is not something we like to do really. Although it seems like it should work there is always the chance that it did not work as expected.
So an idea for troubleshooting would be to remove that inductor for now, replace it with either two known inductors or with two resistors just for now to see if you can get the supply working better. You dont want common inductors on a general purpose power supply anyway.

Even better as a starting point as i said before, run it with just resistors on the output of the transformer, no diodes or chokes. See if it can function with that kind of load first.

Another idea might be that the diodes are different. If one diode has too much more voltage drop it could cause a DC current to flow in the transformer.

 

[mab2]

OK, I'll sub in a couple of resistors - though I'll eat my hat if it cures the asymmetric pulses - but it'll be interesting to see if it runs stable with resistors.

Sorry I'm being a bit slow at my end but paying work has to take priority, and when that's finished I'm often too tired to risk fooling around with 330v d.c. caps.

thanks

m

 

[MrAl]

Hi,

Yes above all remember these three things:
Be careful, be careful, and be careful.
When these guys bite, they bite very hard.

Those input side capacitors can store some nasty energy too, possible even after the power switch is turned off.

Well that test is to eliminate at least one component from the list of things that might be causing the problem. The full AC only output test would be next, which would eliminate the diodes. That will eliminate everything on the secondary side so if the problem persists we'll know it's on the primary side, but if it goes away then it is most likely on the secondary side. One thing though is that you'll have to use some diodes and a resistor and filter cap for the feedback i guess.
The asymmetry is easily caused by an offset DC current in the transformer, but if that turns out to be true then we'd have to find the cause.

 

[rumpfy]

Mab,
1. I looked at your original post and you said there was a problem with squawking then, and presumably the PS was 'funny' in its original state.
2. Mr Al is suggesting the 'common mode' choke to eliminate it. These common mode chokes are quite common and the primary mode current supply simply gives a nett zero flux. When the ground noise and other parasitic noise gives a different current in each + and - lead, then the inductive reactance acts to reduce the unbalance current amplitude. There's no way this kind of parasitic stuff could be simulated in spice and all that stuff, but I would be astonished if it had anything to do with the squawking; but to bypass it as Mr Al says could not do any harm. If it has any affect, it would be that the currents in the + and - leads are different.
The squawking. I am going to suggest, is due to changes in the magnetostriction in one of the cores due to some operating condition. Normally, with these things, to keep the noise down, manufacturers inject a bit of glue (epoxy) into the gaps.
Mr Al's comment about DC unbalance might be happening but if it is, then you would pick this up with the CRO. A DC unbalance would have to show up as differences in the collector currents at the instant of switching. BUT maybe, the original transformer was a dud. I experienced once, a dud TV Line output transformer where the tapping from the winding were terminated arse about on the connecting lugs. Hard to imagine here on the primary side, but the secondary has lots of separate windings.
I'm not forgetting too, that you said you've some home work to do on the CRO; SO....

 

[MrAl]

Hi again,

My main reason for wanted to eliminate the common mode choke is because he said he modified it, and any modification can come with a mistake somewhere that is hard to figure out at first.

But yes most likely this is a DC primary current problem. We could find this out if the entire secondary circuit was eliminated but yes that is a harder test to perform because we still need proper feedback to get the circuit to try to regulate as it usually does. But if the output diodes are eliminated (and the other stuff because of that) then there's nothing left on the secondary to cause the problem, so it would have to be something on the primary.

One procedure to troubleshoot something like this would be to drive the control circuit with a stable voltage and turn up the main power input (sometimes called the "B+" supply) slowly, a few volts at a time while looking at the output for problems. This doesnt require any feedback as long as the input is kept low enough. Yes, that's not an easy test to do either, but when you tried everything else you have to dig in deeper and use more complicated tests to find out what exactly is wrong. If this was a production line we might even have a test jig set up to test the output transformer, but we dont have that luxury here so we have to design tests to help find out what is wrong and some of those tests may not be easy to do.

Here's the long and the short of the possible primary DC current problem...

If one transistor turns on and it's voltage drop is 1.1v, and the other transistor turns on and it's forward drop is 1.0v, then that puts Vcc-1.1v across the primary during one half cycle and Vcc-1.0v across the primary during the other half cycle, and that means there will be a net DC current flow in the primary.
The actual level depends on the primary resistance, but the primary resistance is usually kept low in power converters to keep efficiency higher. The series capacitor should help here though, if this particular unit actually has that capacitor, so it could very well be unequal secondary loading which could be caused by the diodes having a different voltage drop.

 

[mab2]

hello again,

well i tried two 10 ohm resistors instead of the common inductor - it did seem quieter, but there was still some audible noise, and I couldn't draw a large current through the 10 ohms (I'm a bit limited on choice on pairs of power resistors). At light loading the transformer waveform was more a noisy sine with spikes at the crossovers and under load it looked more square but I couldn't see if it was symmetrical.

I did try an unmodified common inductor from another PSU with no obvious improvement

not yet tried removing diodes altogether and putting a resistor across the tx.

Still not tried resistors on the emitters of the primary side drivers - got to work out where to put them (and find a pair of suitable values).

I did try measuring the dc voltage across the 1uf cap that's in series with the primary of the transformer (PSU lightly loaded) - with a cheap digital meter; showed ~50v; same test with an analog meter showed around 20v d.c. on the cap.

 

[mab2]

Hate to admit this, but I've just tried a matched pair of (separate) 300uH inductors and it seems a lot happier up to 4 amps: but I can't check the waveforms when I'm burning my finders with load resistors - other than thinking they looked OK for the second or two I could see them - I need to set up a suitable variable load to test more fully (big transistor on a big heatsink)....

OK - with the separate inductors it seems to be stable and quiet right up to 10amps .

At low powers there's still an asymmetry in the pulses but it seems to disappear at higher currents (or at least diminish to the point where I can't see it on my dodgy scope).

Question is: should I worry about it if the PSU is otherwise working OK?

 

[ChrisP58]

When using a common output inductor for two equal voltages, the number of turns on the core for the two windings need to be exactly the same. The best way to ensure that they match is to wind the two bifilar.

You can almost think of the two windings as two windings on a transformer (which it really is.) Of course, they also need to be phased properly.

The fact that it seems to work better with two independent inductors makes me think you might have a mismatched turn count in the two windings.

An ATX power supply usually has the same inductor for the 5Volt output as well. The turns count of the 5V and 12V windings also has to match the 5:12 ratio.

 

[MrAl]

Hate to admit this, but I've just tried a matched pair of (separate) 300uH inductors and it seems a lot happier up to 4 amps: but I can't check the waveforms when I'm burning my finders with load resistors - other than thinking they looked OK for the second or two I could see them - I need to set up a suitable variable load to test more fully (big transistor on a big heatsink)....

OK - with the separate inductors it seems to be stable and quiet right up to 10amps .

At low powers there's still an asymmetry in the pulses but it seems to disappear at higher currents (or at least diminish to the point where I can't see it on my dodgy scope).

Question is: should I worry about it if the PSU is otherwise working OK?

Hi,

Sounds like you have made some progress.

It also sounds like if you dont need the common mode choke then maybe you wont mind using the two individual ones. You probably dont need the extra filtering and EMI characteristics offered by a common mode choke anyway, and individual loading characteristics will be better with individual chokes anyway.

If you still want to try to use the common mode choke, then maybe you can try to alter the turns ratio or something, or it could be that the core cant take the extra current through the -12v line. If it were my power supply and cost wasnt a huge issue i would prefer two individual chokes, but that's up to you.

 

[mab2]

Well it's working as is so I'm tempted to leave it there (if it ain't broke...), though I may do a few experiments with re-winding the common choke - as ChrisP said I may not have been careful enough - or then again as this PSU squawked before I started on it the original core might have been bad?

I could try fewer turns in case it's saturating I suppose. The main reason for experimenting is that this PSU is one of a job-lot of 5 and they're all a bit squawky (I checked yesterday), so if I solve the problem on this one then I'm likely solving it for all of them - maybe.

Still a little concerned about the asymmetry at low power levels, but if it works and the drivers don't get too hot then perhaps I shouldn't worry about it.

Thanks for all the help - it's funny how you think 'it can't be that' until someone tells you to check - and it turns out to be that.

 

[MrAl]

Hi again,

Well i am happy to hear you are getting somewhere with this now in any case.

When something is wrong i go to the last thing that was changed and check that first.

I've had a little more time to think about this too now and it could just be that the inductance is not enough for the power supply's natural switching frequency. That is, at the required frequency a certain inductance is needed or else the *peak* current goes way up. That could be the problem with the other inductor, that it just does not have enough inductance.

Unfortunately increasing inductance in a DC choke is not as simple as adding turns, because although that increases the lower current AC inductance it actually hurts the DC characteristics making it saturate more easily. I guess if you had time you could try it, but if it is working now why bother.

Hey if you feel like posting a few pics so we can see what it looks like it real life that would be nice too.

 

[mab2]

well you may have to wait a day or two for pics - as I said it's a bit cumbersome with my old camera - but I'll try.

actually I still need to do another mod - the point of this PSU is to charge a 24v battery - as mosaic correctly guessed I think. It works in Cv/cc modes already, but it will be drawing its power from a stand-alone hydropower system which is using a capacitor-excited induction motor as a generator, which is currently operating at 120v (ish), 25Hz, and the psu needs to monitor the input voltage and reduce power if the input voltage drops below a specified level (110v for now). I expect I need to add some extra capacitance on the input side too due to the low freq.

but now it's working with its current mods, I can proceed - work permitting.

 

[MrAl]

Hello again,

Ok sure, whenever you get a chance to take the pics.

Yeah at 25Hz i would think the input capacitance should be doubled as referenced to 50Hz which is close to 60Hz too. Inrush current may increase a bit though so you might have to look out for that blowing the input diodes, i dont know if they use surge limiting thermistors on your particular supply.

 

[mab2]

Well it took a bit more than a couple of days but here are the pics.

It works in constant volts (25 - 29.5 - using the pot on the left side) or constant power (rather that CC) as I simply modded the overload 'trip' to limit the output (adjustable using the silver trimmer hanging off the edge of the board - currently set to 7A at 29v).



I've not yet added the extra caps or the input voltage sensing (work again).

Yes there is a surge limiting thermistor.