Maker Pro
Maker Pro

White smoke - Replacing high voltage rectifier caps in PC power source

My brother asked me to check out his 1000W power source for whining noise at high loads. Oscilloscope showed an out of ordinary high ripple voltage on the input side that follows thru the rest of the unit, making the second stage MOSFETs noisy. I have di-soldered and tested both input electrolytic caps. They showed 15% and 20% loss of capacity. (Panasonic HD 470µF at 450V) After 10years of intensive use that seems pretty good but I decided to replace and asses the situation after. Power source is designed to be uses with both 120AC and 240AC.
So here is where I need help. 470µF x 2 = 940µF seems very little for 1000W power source but there is no room to put anything bigger... My thought was lets lower the voltage as it will be always on 120VAC. So I did a quick calculation 120 * 1.4 = 168 plus a little for good measure and picked up 1200µF 200V caps (LGW2D122MELA50). Polarity was correct but white smoke after 10s :(
What did I do wrong? Is 200V DC not enough to rectify 120V AC? Is ripple voltage that high or is it something else altogether?
Really appreciate your help.
 

Harald Kapp

Moderator
Moderator
They showed 15% and 20% loss of capacity.
That is well within specs.
My thought was lets lower the voltage as it will be always on 120VAC.
Imho not a good idea. How can you ensure that? The least I'd apply is a BIG sticker saying "120 V only".
Is 200V DC not enough to rectify 120V AC?
Theoretically it is. But:
  • Allow for +-10 % tolerance of your mains and the margin dwindles to a mere 14 V.
  • There may be voltage spikes from the switching that feed back to the input capacitors.
I suggest you use higher voltage capacitors.
 
There is some info lacking to know for certain what has happened, but I'll speculate and throw this out there anyway! ;)

The problem was not the cap voltage rating. That would just short out the cap.

What second stage mosfets? If you don't have a schematic, pics might help. I wonder if what you're referring to are secondary aka low side diodes, which may still be in a TO-220, 247 or other 3 pin package.

Regardless, as Harold noted, a certain % of capacity deviation or loss, is not that significant.

"White smoke after 10s", from what component? Details matter.

If I had to just guess based on... statistics or probability or something, that guess would be that either this PSU never saw this load range previously and has an inductor, not transistor, resonating at a particular frequency at the particular load, causing the noise because its coils aren't wound or fixated well enough, OR it is not the primary high side capacitors at issue, rather the secondary low side capacitors have degraded and being an LC, shifts the duty cycle at the particular load to where, again, loose inductor windings are making noise.

The other possibility is you made some slip up in the repair and that caused a short somewhere, or the secondary cap(s) were on the brink of shorting and disturbing the PCB to do the repair put one over the edge... I've had it happen, a system with bad caps, little flaky but still runs, get a bluescreen, move the system slightly and that little bit of motion causes more of a short and PSU shuts off.

Consider that this is a 10 year old PSU. At a minimum to be worth the bother, it probably needs new secondary side capacitors if not all of them, and a new fan, and we haven't even gone into the issue of whether it is a good design with good components otherwise, or overrated generic junk.

It probably isn't too terribly bad, to have lasted a decade we could argue that it's pretty good and served its purpose, but at that age, everything with a known lifespan in single to low double digit thousands of hours lifespan, is subject to scrutiny.

I'm just wondering if this is a gem of a psu or served its useful life and not worth bothering with any longer because even if you get it back in service, there may be multiple capacitors and the fan at least, due to be a soon failure point.

I'm not trying to discourage you from doing what you can, but at the same time, a 10 year old PSU has a certain low value based on expected remaining lifespan, and your time has value, and 1000W of even remotely modern components in a system, have a significant value that is not worth risking on a 10 year old PSU even if you managed to repair it.

Now it could be that the PSU is grossly over-spec'd and the system only uses, say 500W but even that is far above normal consumption, would require several hundred dollars worth of hard drives or a pretty high end video card to come close, so frankly this PSU even if fixed, should be destined for some lower risk use.
 
Thank you for all your help, I appreciate your input.
In no particular order:
- Yes, 120V sticker is a good idea.
- After connecting to power everything was in a good shape. Only after applying the load input electrolytic capacitors heated up and one of them smoked up. I haven't noticed any other damage.
- As power source is still operational I am reluctant taking out too many things at the time. Ideally I would approach it in stages. Power source is silly cramped. Heat-sinks are soldered to the board and there is no access to screws holding 3 pin power components. I may try to 90 degree bend a phillips screwdriver to get to them as I have heard that they tend to come undone...
- Visually everything looks good. No burn marks or blown caps.
- I don't have a circuit diagram, although circuit looks pretty strait forward. Input filters (LC type x2) > rectifier > caps > (35N60C3 x2) > transformer > opto-coupling and control module > (037N08N x 7) > Inductor > (2200µF x5) > 5V inductor based DC-DC + 3.3V inductor based DC-DC. Plus there is a fan control module and what looks like a small standby DC-DC converter with it's own controller ect.
- On idle noise is barely noticeable.
- I wouldn't testify on it but it sound like the noise is coming from 037N08Ns and not from any coils... Seem unusual to me. Noise is getting louder with increased load.
- 1000W power source is not cheap to replace, plus if I recap it it may be good for a while longer.
My brother ordered new set of caps 250V this time. I have told him to get 300V but I guess it got lost in translation.
My plan is to check output caps first as dave9 suggested and then put first stage caps in.
 
Update:
Now I know what happened and where I went wrong. The circuit is more interesting then I originally thought. It does not have (LC type x2) filter but rather it has C - L - C filter (where C are bypass caps). Then it uses next L as a DC-DC boost converter feeding-in to the electrolytic caps. What got me confused is the controller only activates boost at increased load (probably to increase the efficiency at low loads). So when I probed it with my oscilloscope it showed 168VDC. I was not able to draw sufficient load on my bench to get DC-DC boost to kick in. I guess to achieve 1000W output would require a massive transformer. Alternatively you increase 168V to lets say 350V that lowers current on a main transformer... That is why newly replaced capacitors smoked up only after an increased load.:eek:
Schematic is more like this: Input filters > rectifier > DC-DC boost > DC-DC buck (12V) > DC-DC buck (5V) + DC-DC buck (3.3V)
What surprised me is that there is no capacitors before DC-DC boost converter.o_O
 
Top