TDA 450W PC power supply not regulating voltage with current drawn

[Lars Kuur]

Hello,

I have a Chinese made PC power supply that I would have liked to use for an electronic (3d printer) project. I just need to use the 12V.

The unit presents it self with the following problem:

Unloaded voltage: 11.16V
Load of ~0.7A (big fan): 10.73 V
Load of ~3A (heating cartridge): 9V
Load of maybe some 8A (heatbed): PSU shuts down

So I open up the thing and start probing.

I identify the following:

A LM339N chip compared the output voltage with the reference and sends a signal to what I believe is an TL594 chip (number scratched out, but matches the pins and the numbers that is readable). This one is in push-pull mode and it sends a PWM signal to two transistors who uses a small transformer as insulation and then it makes the charge/discharge on the big transformer.

The system also as an Optocopler (next to the yellow transformer). I do not know what this one does as there are no PWM pulse on it and it is just sending a fixed voltage independent of load.

I have looked at the following:

Unloaded voltage with small transients of some 1V.

With 0.7A load.

Also my DMM does not see AC current (I read that 0.5VAC on output read by DMM is an indicator of bad filter caps)

I then check the big transformer to see the wave form:

Unloaded.

Loaded.

Clearly it is changing the PWM signal accordingly to load. This is also what I see when measuring on the IC's and they pump the transistors and all.

And finally the 12V rectifying diode (output):

Unloaded

Loaded.


Also there is no obviously bad caps and no burned components. But the board really looks like a disasters soldered underneath and also on top it does not look the best.

The fuse has been all-right all along.

I am powered by 110V and 60Hz.

The unit does not get hot anywhere.

This is my 2nd "real" try at repairing "advanced" electronics and electronics is not my profession (else I played some with Arduino etc). My first repair (a battery charger for the Dremel) I changed the bad capacitors and saw green diode light and was thinking all is good - Soon after it burned so many parts that repair is now hopeless and it also fried the Dremel battery - What I want to say with that, is that I am new in this and "you probably already tried so I wont mention that to him" does not apply for me. If you think of something even basic don't hesitate to let me know.

Thank you,

 

[BlackQueen]

Wow, quite empressive

 

[Ramussons]

The primary feedback for voltage control is from the 5 Volt rail. As long as this voltage is steady, there will not be much correction on the 12 volt bus.

There are a lot of schematics on the net for PSU's using this IC combination. You will have to disable the 5 volt feedback sense and adjust the feedback resistors to regulate the 12 volt.

For a reference, see the connection on pin 1 of TL 494 http://danyk.cz/s_atx01d.png

 

[Lars Kuur]

Good morning,

It was my impression that the LM339 is using 3 of its 4 comparators to look at -5, 5 and 12V. I will double check this. According to the schematics I can find it only looks at the 5V. If it only looks at the 5V then a really ugly solution could be to cut that and put a voltage divider on the 12V and feed it as 5V.

I do think it is odd though. How the 12V can completely pull down the PSU.

Before writing the original post I did try to play around with an artificial signal on the error input on the TL chip but that made no difference what so ever.

 

[Lars Kuur]

Actually I will also look at the optocopler again. According to this schematic:

https://www.eejournal.org/wp-content/uploads/2010/10/SMPS-IC-2003.gif

It should not be just flat and unaffected by load. It should be pulsing together with the insulation transformer.. (I think... I wish I was better at this...)

 

[mvs sarma]

lars,
place a dummy load on 5V at say 2 amps. Automatically you would get the improvement .

 

[Tony Stewart]

As previously stated and suggested, you must load the 5V output up to 10% of its rated output for optimal performance. This is by design as the PSU expects this to be the primary load and the feedback and tight transformer coupling of all multi-output feed forward converters depend on some primary load. The actual % of 5v current rating depends on quality of the design up to 10%. This has to do with change in loop dynamics with feedback gain changing with load for stability. If any secondary under voltage protection (UVP) detects a fault , the PSU shuts down. Some may also have OCP, OVP and OTP for over current, voltage and temperature. This protects the Logic and peripherals from operation outside of specified voltage tolerances.

The degree of sensitivity from one secondary load affecting another output voltage is called "cross-load regulation"

 

[Lars Kuur]

Dear all,

Thank you for all the feedback.

This weekend I tried to investigate the solutions proposed. Putting the dummy load increased the voltage but only when the 12V was more or less unloaded. When I put full load (~10A) the voltage still drops hard. But it stays alive which was an improvement.

But I did not follow Tony's suggestion of up-to 10% on the 5V. My dummy load is some modest 0.4A.

I do think that the 12V is also connected to the feedback loop (as seen here on pin 14 on the IC https://www.eejournal.org/wp-content/uploads/2010/10/SMPS-IC-2003.gif ) - I do not have that circuit but mine does the same as far as I can trace. Also I have another of these PSU's and it is powering my first printer just fine without any dummy load on the 5V. And from my scope images in my first post it is clear that the load on the 12V impacts the throughput on the system.

Is the system shown on that diagram for pin 14 what you, Tony, refer to as "cross load regulation" or is that something different?

Also I found that the 5V is actually too high (~5.6V ish) which explains a lot in my mind. The power supply sees the 12V is low but the 5V stays high so it wont regulate sufficiently.

I did try the following:

1) Took out many of the resistors going to the LM339N - Especially in the R3x area (see photo in first post) and tried to add and remove resistance to see if that would do anything. It did not. My thinking was that I would pull the 5V lower to improve sensitivity.

2) Removed and measured the 1 uf capacitor in the upper left corner (I think it is the feedback loop - It couples the feedback from the comparator to the 2nd comparator gate) - It was fine. When it was out the voltage actually went up a bit. I tried higher value capacitors but that made no difference. Again this is to me an indication that the high 5V is making the PSU drop the throughput down to reduce that voltage.

3) Made changes to the feedback coming up from all the output wires down in left lower corner. I believe this is the point where the 5V and 12V regulation voltages come from. As I understand it they have a resistor to each level and then one joint resistor downstream of that. I tried to uncouple the 5V entirely and just use the 12V. That did not work, PSU powers on but quick shutdown. I tried to change the resistance of the entire signal. That did work to some extend. By reducing the last resistor from 100K to 70K the power supply can now handle the complete load while the voltage still drops to 9.xV it remains stable at that - And no thermal runaway or anything. I did reinstall everything else except for this change.


I still put a small dummy load on the 5V of and it still has a marginal impact of maybe 0.3V higher output voltage. Honestly I only keep it because it is a fan that I need anyways.

For the time being it is working at 9V although I am still very open to input. I will buy one more of these power supplies to open it and compare (I am obviously not doing this for the money, it costs like 20 USD). It is not an option to open my other working PSU. Unfortunately I am living in a remote location so it is not something that I can just go and pickup.

At the end of the day I think it is some wrong resistor or bad solder joint somewhere. When I compare to other diagrams found on the internet it looks like mine has a lot more resistors and it could maybe just be a bad combination of the tolerances.

 

[Nigel Goodwin]

Also I found that the 5V is actually too high (~5.6V ish) which explains a lot in my mind. The power supply sees the 12V is low but the 5V stays high so it wont regulate sufficiently.

So why didn't you try what's been recommended all along, a much higher load on the 5V rail? - that's how the PSU is designed to operate, and the main regulation is from the 5V - usually the sole regulation is from the 5V, fairly obviously you can't regulate from more than one output, using two in this way means neither are going to be accurate, or terribly well regulated.

 

[Lars Kuur]

So why didn't you try what's been recommended all along, a much higher load on the 5V rail? - that's how the PSU is designed to operate, and the main regulation is from the 5V - usually the sole regulation is from the 5V, fairly obviously you can't regulate from more than one output, using two in this way means neither are going to be accurate, or terribly well regulated.

Hello Nigel,

First off, thank you for your interest in my problem.

My other PSU works just fine without any load on the 5V. I do not see why this one should not be able to work the same. It is clearly regulating from the 12V. I do not have any interest in making an external circuit just to have the 5V making heat. It is hot enough here in Brazil. If that is the only solution then I will rather buy a new PSU - And it is not the only solution, now it is outputting 9V when fully loaded and nothing on the 5V and that is OK but not perfect - My thinking is that there must be a way of making the "cross-load regulation" so sensitive that it will become even better.

In relation to that;

I do not understand what is happening on pin 14 in this schematic:

For my understanding I would be very intersted if you can comment on that.

 

[Nigel Goodwin]

In relation to that;

I do not understand what is happening on pin 14 in this schematic:

For my understanding I would be very intersted if you can comment on that.

Like I said, the PSU is designed for decent a load on the 5V rail, by ignoring that you're stopping it working properly.

As you seem set against that (even as much as trying it, seeing if it cures the problem), try disconnecting the 5V feedback (as you don't want it - as I said before, you can't regulate from TWO rails, only from one). You can do this by disconnecting R62, and you might need to alter the value of R60 - presumably lowering it?. It might also be a good idea to add an electrolytic before L6 as well, giving a smoother rail - the important rails (5V and 3.3V) already do that.

 

[Lars Kuur]

Like I said, the PSU is designed for decent a load on the 5V rail, by ignoring that you're stopping it working properly.

As you seem set against that (even as much as trying it, seeing if it cures the problem)

I never said that I will not try it to help diagnose. To the contrary - For diagnostics I will do anything you ask. As mentioned previously I actually did try it with 0.4A and it did improve the situation. Seeing that and knowing my other one is working fine I considered that part of diagnostic done. Also seeing that it had an effect it convinced me that it is some combination of voltages set by the resistors which are out of balance. Understanding now that it would be of interest to see what it does with a much bigger load I will try that. I am sorry if that seemed ignorant of me - It was not meant that way.

Also I will try your suggestions for the 5V but I will probably wait until I have purchased a new one that I can compare against first - And so that I have a backup if something goes wrong.


I am still looking for somebody who can explain to me the inputs on pin 14 on previously mentioned schematic. It seems to me that it is taking the input of both the 5V and 12V. Intuitively I think that would give an average of the two voltages and that it is using this average as input to the regulating circuit.

But Nigel has now, firmly, informed me a couple of times that a PSU only regulate on one rail (5V) - So what does that configuration going to pin 14 do? And what does that actually mean? That it only regulates on one rail? Looking at the schematic I cannot identify any place where only the 5V is involved.

 

[Tony Stewart]

Cross load regulation error is when one goes down from being loaded, the unloaded goes up since they are derived by a common current pump. through the transformer which is tightly coupled to all outputs but there is some leakage and thus you are certainly ignoring the correct advice to fix the problem.

Your PSU is not designed for general purpose use with any load on any output.

Historically all SPMS had to be pre-loaded 10% to run at all. That has improved by design evolution.

 

[Nigel Goodwin]

But Nigel has now, firmly, informed me a couple of times that a PSU only regulate on one rail (5V) - So what does that configuration going to pin 14 do? And what does that actually mean? That it only regulates on one rail? Looking at the schematic I cannot identify any place where only the 5V is involved.

As I've also said, the bizarre feedback of two rails means that it regulates poorly - you can't regulate from more than one output, which is obviously if you think about it.

 

[Lars Kuur]

@ Tony Stewart - Thank you for the reply. I am happy to read your explanation, because that was also how I have come to understand that it works. Hopefully you can appreciate my point of view that if I can only make the PSU pay less attention to the 5V going up, then it should improve functionality. And for clarification - I do not need 12V fixed. I just need a reasonable voltage and I need it to not go up and down several volts as I turn the biggest heater on and off.

As I've also said, the bizarre feedback of two rails means that it regulates poorly - you can't regulate from more than one output, which is obviously if you think about it.

Ok, I am sincerely trying to understand but I am still confused;

On pin 14 on previous schematic. There is a 12V and a 5V going in to pin 14. Is that what you call a bizarre feedback? Or is it the mere concept which is bizarre?

My question boils down to this:

Is it correct or wrong that the schematic shown is regulating on a sort of average of the two voltages? Or how is it, that shown schematic is only regulating based on the 5V while taking both 12V and 5V as input?

It would really help if while answering that question we could be specific to the schematic as that will help my understanding greatly.

 

[Nigel Goodwin]

It's quite simple to understand - the 5V is low, so the regulation turns it up - this means the 12V rail is now too high, so that turns it back down, repeat ad-infinitum.

And of course the exact opposite happens as well.

At no time is either rail accurately regulated,

 

[Ramussons]

As I mentioned in my earlier post, The ATX PSU regulates primarily the 5 Volt bus. There are over and under voltage cutouts for the voltages, but the 5 volt bus is the primary one. It can also be seen from the resistor network on pin 14, that the "regulation contribution" of 12 volts is less than half of that by 5 volts.

Try removing R62 and replacing R60 by 47 K to 56 K

 

[Tony Stewart]

Since there is only one primary source of power, the coupling factor in the transformer must be high to achieve low crossload regulation error. Since all computers have some ratiometric loading that varies, so look at the PSU label for max power on each output (VxI) to get a sense of the ratios for which it was designed. 5V is generally the largest load which in turn feeds DC-DC regulators on board for CPU & RAM. The biggest 12V load would be a high end GPU card.

Thus the combined feedback of 12V & 5V on pin 14 is an attempt to improve cross-load regulation between 5 & 12. But still assumes a minimum loading on each for loop stability. and some unknown ratio of power load that you will calculate from above for load and crossload regulation error of 5% total for loads between 10% & 100%.

Pin 14 is just the inverting input of an Op Amp with internal non-inverting input scaled from the 2.50V internal Vref. My calculations show pin 14 should be around 2.37Vdc when in regulation.

The 950k R is a factor jumper to remove to raise all voltages by ~1% (=8k/960k) using equivalent values
The 100k 0.1uF filter is a LPF with a 10mS time constant, so AC line variations that modulate the bridge rectified voltage are tracked only fast enough to to track brownout conditions but suppress faster transients.
The 47k feedback 10nF filter has a breakpoint at 470 us so this should attenuate SMPS ripple by at least 20dB over a 20~50uS period.

More importantly 47k fixes the AC step load response loop gain to 47k/8k= 6x for transient loads where 8k is the equivalent node resistance in pin 14.

The DC loop gain of the integrator will be high and the output will integrate the avearge error with mixed ratios feedback from 12 & 5V Notice the 12V feedback ratio is lower which is only intended to make up for lack of coupling factor in the transformer for cross load regulation.

To make this a variable 12V regulated supply.
Remove 22.1k(5v) and the 121k(12v) feedback R's and choose new ones just for 12V feedback so that the middle of the range is 2.38V on pin 14.

2.38V/12V=20% = R60/(R60+12k) where 13.1k// others=12k
Thus 0.8*R60=2.4k or R60=3k I would choose R60= 1k trimpot + fixed series 2.4k

To improve DC gain stability, add 100k between pin 13 & 14. This prevents the integrator from hunting.

 

[Lars Kuur]

Wow, thanks. I really appreciate the help from all of you.

I have not made a lot of progress at implementing your suggestions. I am attending some further education with exam coming up in two days. Post that I will be back to play around with this. Hopefully next weekend. I really feel that I know where we are heading with this now.

I did find time to try the 10% load on the 5V. I put ~5A load which is much above the 10%. Then when also loading the 12V (with ~3.5A) it puts 12V voltage to 10.5V. For reference my working and same type PSU delivers 11.2V when fully loaded on the 12V and nothing on the 5V. So something is off in this feedback but with Tonys clarifications I really know what to look and try now.

This is not me asking for more help. I am fairly sure that I know what to do now and I really just wanted to say thanks and that all your replies have been read. I will return with, hopefully, a conclusion soon.