This doesn't even belong in the hall of shame.

[Th3_uN1Qu3]

Okay... i couldn't resist posting another one before i go to sleep. Since it's missing some caps and almost all others are bulged (our old friends at f-word), you could say i'm not fair. Therefore you can see in pic 6 how it would look if it had all caps. Unfortunately the one in pic 6 suffered some water damage at some point in time and some PCB damage during packing or shipping or whatever. But at least you can have it as reference.

Surprisingly, the thing still worked in its current state (!!!) albeit with an ear-splitting whine. So let's go ahead and check 5vsb... Uh... I'll let you guess who made this one. I know, but i'm not telling.

PS. Yes, that which you see is the primary heatsink. And there is an invisible heatsink for the 5vsb switch too.

[kaboom]

Re: This doesn't even belong in the hall of shame.

WOW!

Between the 'wrong' Y caps and the vivas, fooyoos, and 'diodes on a bracket,' I thought that was a Deer.

Load that '5V' standby "till it pops." On a good day, this would've been good for 50 watts, so where the hell does it get the other 180? Do you think it pulled the energy from the ambient air and cooled the room?

-Paul

[Scenic]

Re: This doesn't even belong in the hall of shame.

CWT.. omg

and almost 20V on 5VSB ..

[Th3_uN1Qu3]

Re: This doesn't even belong in the hall of shame.

Now, if i tell you this still powers up and all other rails still work would you believe me? It'll even take a direct short on "12v". It'll do 4 amps at 0.38 volts... Not even enough to light up the bulb i put in series with the primary so things don't blow up in my face.

[b700029]

Re: This doesn't even belong in the hall of shame.

CWT sure has come a long way since then...
http://www.hardwaresecrets.com/image...hp?image=13051

Interesting to see they're still using the same heatsink design and QC sticker on the primary cap!

[kaboom]

Re: This doesn't even belong in the hall of shame.

4A @ .38V, funny...

Does that have a 3842 or some such 'current mode' controller? Has too, one primary switch, and I think voltage mode control would make it blow up with those caps.

If you were to scope the top of the source resistor, you'd probably find some nice current peaks in the primary.

Like I said in that other thread, "Heatsinkless PS- a new feature!"- haha...

-Paul

[Th3_uN1Qu3]

Re: This doesn't even belong in the hall of shame.

Does that have a 3842 or some such 'current mode' controller?

Yup, a 3843. Btw, i decided to actually give this thing a chance (and it'll also double as an excuse to get rid of the fukyoos i have left from an Antec). I'll fix it up, and load it till it blows.

[kaboom]

Re: This doesn't even belong in the hall of shame.

As you are aware, the source resistor will make the 3843 fold back if you try to pull too much. It's probably something like .1-.27 ohms/2W. Don't bypass it since the 3843 may depend on the 'ramp' created by the IL time constant of the primary winding- it may not start. Worse, it may just go bang, short the FET, blow the 3843 and piss you off

Sleep well and have 'fun' at the 'party.'

-Paul

[Th3_uN1Qu3]

Re: This doesn't even belong in the hall of shame.

Thanks, now i really am going to sleep, the fun comes tomorrow.

I changed the two little fukyoos and one capxon in the primary of the two transistor 5vsb with Rubycon YXA.

The good news: The darn whining stopped.
The not so good news: 5vsb reads 15v. Most likely, due to the circuit's nature of cooling the room, one of the resistors that bias the optocoupler drifted permanently.
The dramatic news: Out of the blue, the secondary side IC (probably the supervisor) decided to let the magic smoke out.
The downright ridiculous news: While that IC is pouring out smoke, the rest of the power supply keeps functioning perfectly.

[kaboom]

Re: This doesn't even belong in the hall of shame.

Possibly, the 5 and 12 outputs are summed with resistors and feed something like a TL431, which then drives another opto just for main supply feedback. The supervisor was probably just OVP. And I do mean "was!"

It may have had ~20v, but only in 'pulses' as the supply squealed away. Now, after you recapped the aux, that IC had a nice steady supply of DC, and it finally 'cooked good' after being tormented by the 'pigs in the power supply.'

Be good and stay out of that PS for tonight- it may be plotting against you. After all, it's already violated the laws of thermodynamics!

[Th3_uN1Qu3]

Re: This doesn't even belong in the hall of shame.

I believe you are correct. The supervisor is a TPS5510P, and it is powered from the 5vsb rail (even tho it was supposed to be powered from 12v). Absolute maximum rating is 15v. Now, this chip also does the power good signal, so i'll check that too... i wouldn't be surprised if the PG signal is held high even with the chip up in smoke.

Since i have a good one on the other board, after i fix the 5vsb rail i'm going to transplant it to this one. Let's give it all chances shall we?

And yes it has TL431. Two of them to be precise.

Edit: Yup, the PG signal is held high. So, given that the standby rail wouldn't have killed it already, your computer would still be running while the power supply roasts itself!

[Alex Atkin UK]

Re: This doesn't even belong in the hall of shame.

What? No videos showing us the magic smoke?

[Th3_uN1Qu3]

Re: This doesn't even belong in the hall of shame.

I would've made one, but the thing decided to kill itself before that.

This PSU is hopeless. The supervisor decided to short all rails together. Killed the 5vsb. I removed it, 5vsb was ~3v, then after a few seconds it would jump to over 20v and the PSU would draw some serious current from the mains and the horrible squeal came back.

Found an open 18v zener in the 2 transistor SB circuit. Replaced it only to have its replacement short out. Within about 10 seconds the standby switch blew. And that's the end of it.

Edit: I found some specs. 3.3v @ 14A, 5v @ 22A, 12v @ 8A, -12v and -5v @ 0.5A, 5vsb @ 2A. All ratings are followed by "MAX" which means that it'll do them for maybe a couple minutes. And my bad it has 3x TL431. Stay tuned while my big soldering iron heats up, it's time for an autopsy.

[Th3_uN1Qu3]

Re: This doesn't even belong in the hall of shame.

Topology is single switch forward. Input rectifiers are 2A05 - sure they're all teeny weeny but will do for 230v input.

Primary switch: Toshiba 2SK2545. 6A 600v 0.9ohm. Assuming an efficiency of 80%, we have a max current of 0.96A at 300vdc (those caps are crap so i'm quite sure they won't keep the rectified DC at 320v). At 60C operating temperature, following the RDSon curve in the datasheet we get 1.3 ohms. 0.96A over 1.3 ohms solves to 2W. With fan cooling, that tiny heatsink is perfectly adequate...

Standby switch: Toshiba 2SK3067. 2A 600v 4.2 ohm. Typical standby switcher, i've seen worse. 10W at 300vdc is 0.0333 A, which solves to 0.006W over 5 ohms (specified maximum value for RDSon) So, if the 5vsb line worked properly, the lack of a heatsink wouldn't have been a problem. 5vsb is rectified by a FR302 (3A). Enough for 2A.

Moving on to the secondary side. 12v is taken care of a pair of FR302s. 8A? Maybe, if you ask them nicely. 5v has a STPS1545CT (15A). 22A? Not a chance. 3.3v first goes thru a 10GWJ2CZ (10A) then it is chopped by a STP3020L (40A 30v 0.022 ohm). There is no dedicated 3.3v inductor, this MOSFET uses a winding on the main coupled inductor for that, talk about cheap! And it seems it didn't like it one bit, as there is significant thermal damage to the core. Out of all the rails 3.3v has the biggest chance of actually meeting spec. Well, at least i got myself a nice MOSFET. Btw, except for the 5vsb switcher as mentioned, all big silicon tested well.

There are pi filters but they're filled with fukyoo TMR and a couple Su'scon SD, one of which was bloated. Now, Su'scon aren't bad caps. They are used extensively in CRT monitors and i have yet to see a failed one in that application. And those Hipro supplies have them too and they're all good. However, looking in their catalog, the SD series only has 1A ripple current rating. As for the fukyoos? Maximum ripple current isn't even mentioned. There's only 1 cap per rail (well, with the pi filter cap that makes two, but the first one is the most important), so it's quite possible that this supply ran the caps out of their specs.

Now, from the timing resistor and cap, oscilator frequency should be around ~70kHz. At this frequency a forward converter needs a significant amount of turns... And given that the core is an EI33 and there's even a bit of space left, there are likely to be significant losses in the transformer as well.

Let's make another guesstimate with the component values as found:

• 12v: 6A (72W)
• 5v: 15A (75W)
• 3.3v: 14A (46.2W)
• -12v and -5v: 0.5A (8.5W)
• 5vsb: 2A (10W)

And we get 210W. Not far from the 230W rating. Plausible? Yes, if the parts were good quality. Would it have worked with good caps? With that inductor... i don't think so. Still, the greatest issue remains the 5vsb rail. The lack of a protection zener for the 5vsb is inexcusable. If the zener were there, at least the standby transistor would have blown up before it had a chance to damage the motherboard.

And that, about wraps it up. Next.

[Evil Lurker]

Re: This doesn't even belong in the hall of shame.

Heres one for ya... came in a sunpower 250W rated PSU. Believe it or not it has the infamous "Rulycon" primary caps (and yes I had to open one up just to see if there was a smaller capacitor inside... there wasn't dammit). I would of hooked it up for testing, but was afraid it would damage my mains.

[Th3_uN1Qu3]

Re: This doesn't even belong in the hall of shame.

The transformer would do 250W. But with those secondary caps... it'll be lucky if it does 50W while staying in spec.

[everell]

Re: This doesn't even belong in the hall of shame.

Since the 5vsb circuit seems to be the biggest problem, perhaps you should add a DM311 mod like the one I used in the Bestec ATX-250 12E. It did a great job of taking care of the 5vsb overvoltage problems. Except for a few bulging caps, the rest of the Bestec works quite well. With a decent pwm chip in the 5vsb circuit, maybe your CWT psu will do much better.

[kaboom]

Re: This doesn't even belong in the hall of shame.

But with those secondary caps... it'll be lucky if it does 50W while staying in spec.

What secondary caps?

Lurker, add an EMI filter to that thing! What a QRMer... At least the lil' supply's growin' up- a PI filter for 5, maybe 12

[c_hegge]

Re: This doesn't even belong in the hall of shame.

The trouble is, you can't add an input filter as the PCB isn't even screened for one. sunpower

[Th3_uN1Qu3]

Re: This doesn't even belong in the hall of shame.

With a decent pwm chip in the 5vsb circuit, maybe your CWT psu will do much better.

Or with a plain old 7805. But... nah, not worth fixing, besides, i gutted it of everything already.

I have another patient that due to 5vsb overvoltage has shorted the standby transistor, blown the 2.2 ohm resistor in its emitter (it blew so the fuse didn't have to ), shorted the 494, made the primary transistors die one shorted and one open (probably when the 494 died its outputs went all wonky), and who knows what else will i find in there. Too bad, because the power supply looks respectable otherwise.

Secondary caps were mostly CS and all of them are alright, but what do we get for 5vsb? RULYCON. After replacing the standby transistor, the 2.2 ohm resistor and changing the RULYCONS for the real thing written with a B the standby still doesn't come up so there'll be more digging to do... First i will drop in another 494 and use an external supply to power it. Once the main power supply is alright then i will focus on the standby circuit.