Sumvision PXPRO800 ATX PSU - partly repaired

[jarsudsco]

I have a quite new Sumvision PXPRO800 ATX PSU. I apparently over-loaded it, even though my calculations said it should be fine, which resulted a bang, a flash and some smoke.

It took a little while to find anything obviously burned-out but hidden under the hot-side heat-sink was a burned out resistor (R7). The middle band was almost completely gone but my best guess was it was 2K7 so I replaced it and it (optimistically tried powering up) - it promptly burned-out again.

Some further inspection revealed that R4 appeared to be cracked and although it was marked as 1-ohm it was actually O/C.

While I had the power transistors off-board I tested those, the Q2 (marked FQPF2N60C) seemed to be ok (MOSFET DMM test only) but Q01 and Q02 (both marked as 3DD13009) had both gone S/C B to C.

Replaced Q01 and Q02 with MJE13009's, replaced R7 (again) and R4.

Now, the +5SB line comes up but that's all, it won't otherwise turn-on. Very similar to another thread I've just looked at - this PSU also uses a 339 chip and a 7005 chip for control.

I removed the output ferrites L1 and L2 to get at the cold-side heatsink, then the 4 pairs of Schottky diodes - they all tested ok.

(Edit to add - PCB is marked ATX-1000 and SH10022 by Xinhuida Electronics Co. Ltd.)

This is my first attempt at an ATX PSU, and I'm a little out of practise in general, so any suggestions as to where to go next would be very much appreciated.

Regards,


- jarsudsco

[Agent24]

Re: Sumvision PXPRO800 ATX PSU - partly repaired

Is it just me or does D31 have a crack in the body?
Are you getting 5v out of the PS_ON pin?

This may help http://www.electricaltechnology.org/...sistor-by.html in the event that a wrong resistor value is preventing power on


Ultimately though, if the PSU blows up on overload rather than shutting down, it's a shit PSU anyway, and you should get something better.

[c_hegge]

Re: Sumvision PXPRO800 ATX PSU - partly repaired

It looks like about a 350 Watter, maybe 400W. I sure do hope it wasn;t labelled as an 800W (like the model number suggests).

Does the fan briefly twitch when you try to power it on?

[jarsudsco]

Re: Sumvision PXPRO800 ATX PSU - partly repaired

@Agent24 - All the Schottky's looked OK when I inspected them and all seemed OK when I checked with a DMM diode tester - there are 2 different device types - D30 & D33 are STPS2045CTC and D31 & D32 are ZA13P59 - junctions were all between about 0.2 and 0.5 volts ISTR (I can't find my notes!)

+5SB seems stable at 5.11V.
PS_ON unconnected reads 4.94V.
All other outputs read 0V (with PS_ON grounded).

Thank you for the link - method 1 was a good idea but unfortunately R7 was a sub-miniature type so the burned-out section is relatively large compared to the body. I did try method 1 but could only get O/C, S/C results.

I'm fairly sure that R7 reads red-purple-red (pic attached) but the third band looks a little more orange than the first, but the other sub-miniature resistors on the board all have very much paler orange bands so I think the third band has just got bit discoloured from the heat.

I agree entirely with your comment, I don't think much of a PSU that dramatically fails rather than shuts-down.

@c_hegge - Yep, sold and labelled as 800W (attached label photo), I reckon I was pulling maybe 400W (mainly on the 12V line) when it blew.

No, the fan does not move when I try to power-on.

Regards,


- jarsudsco

[goodpsusearch]

Re: Sumvision PXPRO800 ATX PSU - partly repaired

How did you manage to pull 400W of it?

[Agent24]

Re: Sumvision PXPRO800 ATX PSU - partly repaired

@Agent24 - All the Schottky's looked OK when I inspected them and all seemed OK when I checked with a DMM diode tester - there are 2 different device types - D30 & D33 are STPS2045CTC and D31 & D32 are ZA13P59 - junctions were all between about 0.2 and 0.5 volts ISTR (I can't find my notes!)

You're probably right. I just see some kind of dark line but it's more likely to be a reflection or dust. No way it would test OK if it had a hole blown in it!

I'm fairly sure that R7 reads red-purple-red (pic attached) but the third band looks a little more orange than the first, but the other sub-miniature resistors on the board all have very much paler orange bands so I think the third band has just got bit discoloured from the heat.

It does look like a 2.7K to me too. Where is R7 located in the circuit?

I found a datasheet for your PWM controller. I guess you could check to make sure it's getting power at least.

The SDC339 is most likely just the same as the LM339 comparator

[jarsudsco]

Re: Sumvision PXPRO800 ATX PSU - partly repaired

@goodpsusearch

How did you manage to pull 400W of it?

That was an estimate, I was building a machine with dual graphics cards. The cards were each specced to require 2 x 75W PCI-e power connectors, but I actually made up some PCI-e extension leads with 0.01-ohm shunt resistors in the 12V lines so I could measure the actual current and they maxed out at around 4.5A per connector so, 9A per card = 18A in total, x 12V = 216W.

The CPU according to Intel was rated at I think it was 70W, and I assume this is predominantly drawn from the 4-pin CPU power connector so another 70/12 = 5.8A on the 12V line.

The motherboard had almost all BIOS functions disabled, there was not even a hard-drive, I was initially booting from a USB flash drive with 2x2GB of system RAM so I would guess less than 100W for everything else and probably most of that was on the 5V or 3.3V lines.

So that's 216W (measured) + 70W (according to spec) + 100W (my over-estimate) = 386W.

And only about 24A worth of that was on the 12V line on which the PSU claimed to be able to deliver 19+20=39A (it only has one 12V rail externally).


@Agent24

I'm pretty confident that the 4 Schottky diode pairs were ok, and the more I look at R7 the more sure I am that it's 2K7. From memory, I've marked it on the PCB copper side and it's between the base and emitter of Q01 - one of the 13009 power transistors.

Thank you for the data sheets - that was good of you. I'll check as soon as I have time.

Regards,


- jarsudsco

[redwire]

Re: Sumvision PXPRO800 ATX PSU - partly repaired

If the burnt resistor is across the E-B junction on one of the MJE13009 power transistors... then I'm afraid it's toast, as in open E-B junction and probably C-B shorted. Most schematics have that resistor as 2.2K or 2.7K ohm. There should be another one on the E-B junction of the second MJE13009, for comparison.

[jarsudsco]

Re: Sumvision PXPRO800 ATX PSU - partly repaired

@redwire - Thanks for the reply. You are quite correct - in fact both the power transistors (originally 3DD13009) had gone S/C B to C - they've both been replaced by MJE13009. I don't recall another resistor in the same location on Q02 but I don't have the PCB to hand - I'll check shortly. Thanks for the confirmation on R7 - it's currently 2K7 and that's were I've got to. +5SB line up, nothing else. I need to do some further checking/measuring.

Any other suggestions would be appreciated.

Regards,


- jarsudsco

[redwire]

Re: Sumvision PXPRO800 ATX PSU - partly repaired

I suggest working upstream of the 13009's. There's could be another open resistor like R4's counterpart. Also a 1uF cap and diode from the driver transformer to the 13009's base. They should be checked as they could get damaged too.

[goodpsusearch]

Re: Sumvision PXPRO800 ATX PSU - partly repaired

I would check any diode/resistor/capacitor near the 13009s, I would also check the rectifiers and finally I would start suspecting the ICs..

[PeteS in CA]

Re: Sumvision PXPRO800 ATX PSU - partly repaired

It may just be the wide shot, but one of the smallish 'lytics between the heatsinks looks swollen and aslant.

Agree with the 350W-400W opinion - the MJE13009s, the main transformer core, the heatsinks, the current ratings of the rectifiers, and the size of the output inductors are all pretty consistent with that. It's the TL494-clone, half-bridge topology, bipolar transistor, 20KHz-30KHz design that is pretty much ubiquitous in low-cost PC power supplies. Done well, it's inexpensive adequacy. But that won't do anything close to 800W continuous (unless "continuous" is measured in minutes, maybe single-digit).

[momaka]

Re: Sumvision PXPRO800 ATX PSU - partly repaired

And only about 24A worth of that was on the 12V line on which the PSU claimed to be able to deliver 19+20=39A (it only has one 12V rail externally).

Never, ever trust the label on a cheap PSU!
(Especially one without any safety agency marks of any kind... speaking of which: most are often faked on cheap PSUs so only look for UL and CSA marks with numbers).

D30 & D33 are STPS2045CTC and D31 & D32 are ZA13P59

I can't find anything on the ZA13P59 diodes. Check your numbers on those again. Most likely, they are lossy fast recovery rectifiers rated for 100V to 200V and 16A (or 20A if you are lucky).

In any case, since this is a half-bridge topology, the two STPS2045CTC(s) are for the 3.3V and 5V rails, while the two "ZA13P59" are likely in parallel for the 12V rail. This doesn't mean you can pull 16+16=32A or 20+20=40A from them, though. With rectifiers in parallel, the most current goes through the one that has the lowest Vf drop. Because of that, the one with the lowest Vf drop will heat up more than the other rectifier. This will in turn cause more current to flow through it, and so on until it burns up if the current is much higher than what it is rated to handle. Now, because both of these rectifiers are couple to the heatsink, the diode with the lower Vf drop can't get that much hotter than the other diodes, so this will usually keep things in balance and you may be able to pull a bit more current from this parallel rectifier duo than from a single diode of the same rating. That said, it's usually not much more. With two 20A parallel rectifiers, I would expect no more than 24 to 25A max. Maybe up to 30 for brief periods, but that would be pushing it quite far IMO.

And that's why the board appears very discolored near D31 - it most likely was getting overloaded.
....
There may also be a few small glass diodes near the primary transistors - check those as well. Since the original main BJTs shorted, the two small electrolytic caps there need to go as well.

[jarsudsco]

Re: Sumvision PXPRO800 ATX PSU - partly repaired

@redwire, @goodpsusearch, @PeteS in CA, @momaka

Thank you all for your comments and advice, I haven't had much time this week to look at the PSU but I'll get back to it shortly.

Regarding the discolouration near D31: you may well be correct, but there is a low value resistor directly behind D31 (I think it was 100-ohm - I don't have the details to hand) and that seemed to be the centre of the scorching although the resistor tested ok.

I'll double check the labelling on the diodes too, I should have taken a picture whilst I had them off the board.

I also wanted to double check the specs of the replacements I got for the 3DD13009's - I bought them as MJE13009's but the labelling on them is just "E13009L" under a Fairchild "F" logo. I can't find this marking (or indeed a MJE13009) on Fairchild's web site and I've even tried telephoning their UK office who gave me another number for someone who they said should be able to help me track down the correct datasheet but the number has not been answered when I've called it.

Regards,

- jarsudsco

[Agent24]

Re: Sumvision PXPRO800 ATX PSU - partly repaired

13009 will be the basic part number. The letters at the start will vary depending on the manufacturer - many manufacturers make the same parts. Letters at the end can denote some variation, like different package types or a lead-free version. The manufacturer code is not always present on the package marking.

See attached datasheet(s) for your Fairchild ones. KSE13009L is TO-3P package while KSE13009 is TO-220. They are the same except the TO-3P has a bigger package and can dissipate more power\heat.

[PeteS in CA]

Re: Sumvision PXPRO800 ATX PSU - partly repaired

I can't find anything on the ZA13P59 diodes. Check your numbers on those again. Most likely, they are lossy fast recovery rectifiers rated for 100V to 200V and 16A (or 20A if you are lucky).

The "ZA13P59" may be the date/lot/fab code. Some semi mfrs use the simple 4-digit date code (WWYY or YYWW). Others include the country, fab, lot and date in their code, and some one wanting to know what it all means needs that vendor's "secret decoder ring". I know STMicro and IR have that sort of complex code system.