the gutless, bloated, and fried power supply hall of shame

[PeteS in CA]

Originally Posted by ben7

The only good about it is it looks like it has proper Y capacitors ... not regular, unsafe, ceramic capacitors.

Yeah, no input common mode inductor, no X cap, no MOVs to prevent the input lytics from venting (a safety standards violation, IIRC). Those heatsinks, the MJE13007-copy BJTs, the 470uF input lytics, the TL494, the -26 material output inductors all yell, "I'm a mediocre 250W power supply!"

[elitist rant] Dammit, the UC384x family PWMs and power MOSFETs have been around for 3 decades, everyone knows who the good cap vendors are, powdered iron cores have been around longer than I have, and heatsinks have been around and understood as long or longer! There is no excuse for manufacturing and selling a TL494-and-BJT-switch-device-and-26-material-based, unfiltered input, wimpy-heatsunk, crappy-fanned, crappy-&-wimpy-capacitor-populated POS like this. Not without a multilingual warning label such as, "Warning: this POS is a waste of money and is hazardous to your checking account, computer system, and home!" [/elitist rant]

[PeteS in CA]

Those Chinese PSU names are funny. Though still not as funny as PowMax, IMO.

Don't know if they're still around, but in the mid-late 80s I came across a Korean power supply company named Torch. Just inspires so much confidence in their reliability. Then there's the classic American power supply company, Acme Electric (Beep! Beep!). In fairness to the latter, they predate Roadrunner cartoons.

[Pentium4]

Your post made me laugh Actually, the heatsinks are better than they look. That metal is actually pretty hard to bend with your fingers, I was surprised. I'm not saying they're good, just that they aren't the Leadman style ones that look like they were forged from melted beer cans I can load it up to around 200W and see how hot the toroids get. By the way, what's that third coil doing in there?

[cheapie]

Originally Posted by momaka

Yes, very important for your SDRAM to have clean power!
Oh wait, it's not 1999 anymore?

You realize that we still use SDRAM today, right? (DDR, DDR2, DDR3... it's all SDRAM) I'm not sure if it's still powered from +3.3V though.

[Wester547]

That third toroid is required for the magamp circuit on the +5V and +3.3V rails as it derives the +3.3V from the +5V transformer pins (hence the combined rating). You will find it in pretty much every PSU with that design (and possibly another one of those), except for PSUs that use linear regulation (or other kinds of DC-DC conversion) to generate the +3.3V rail.

[Pentium4]

Cool, thanks for the info. I guess the design of this one just threw me off, the secondary design is a little different than I'm used to seeing. It's amazing that they use these old designs and "semi-upgrade" them. I'm also used to seeing two rectifier slots available on the +5V rail, but that Power Star has another rectifier slot on the +12V rail.

Even though this design sucks, I wonder how much it would benefit from another +12V rectifier. Although recently I've noticed that a lot of old half bridge PSU's refuse to use schottky rectifiers on the +12V which is unfortunate because they work fine on the +5V and +3.3V

[momaka]

Originally Posted by cheapie

You realize that we still use SDRAM today, right? (DDR, DDR2, DDR3... it's all SDRAM) I'm not sure if it's still powered from +3.3V though.

What i meant by SDRAM was that good old classic PC100/PC133 168-pin stuff.
On crappy motherboards (or maybe even most motherboards and not just crappy ones), it was directly powered from the 3.3V rail, so any instability on the 3.3V rail could make the PC crash.

The most recent motherboards I've worked on is Pentium 4 S478 stuff and AMD AM2. And even on those, *most of the time*, RAM is powered by the 3.3V rail. However, newer RAM (DDR and newer) is 2.5V or lower, so the 3.3V rail is stepped down - usually either through a MOSFET in a linear fashion or with a synchronous buck regulator. Because of that, ripple/noise on the 3.3V rail rarely affects the RAM (and thus system) stability.

Buy try hooking a really crappy PSU (i.e. one with no PI filters and insufficient capacitance) to an old motherboard and see what happens. If the motherboard doesn't have a lot of caps near the RAM to filter power from the PSU, you'll probably get lots of rebooting and/or BSODs.

[Behemot]

That was the principle of the design so I don't understant your crap talk.

The same, powering by +3,3 V is a must to provide minimum load on +3,3 V rail. Nothing else is drawing power from this rail anymore. I am expecting new ATX revision any day which will most likely increase +12 V even further, exclude -12 V rail and also require the PSU to be able to run with no +3,3 (+5) V load.

In the long-term, I expect +3,3 V and finaly also +5 V to be exluded as well, possibly replaced with something like +24 V (for higher efficiency). But it won't be easy and probably will brake backward-compatibility.

[mariushm]

You'll have to forgive me for the quality of this post - I'm visiting my parents' place for a few days so I'm demoted from a eight core system with 2 1920x1200 screens to a Sempron pc with a 19" lcd screen.
In addition, I've snapped the following pictures with my sister's camera, with batteries nearly dead (literraly turned off and on the camera every 2 pictures due to flash getting the batteries below the threshold).
I'm also writing this through remote desktop connection to my pc so if i make grammatical mistakes it's the lag.

So please excuse the out of focus and poor quality of pictures, the post,everything, i'm not in my environment.

Anyways, on with the story. Father complained a few weeks ago that the system was erratic, not powering on , or powering on only once every few times. Then one day it died.
Knowing it was a crappy psu that came with a 30$ case 3-4 years ago when this system was made, i just told him to go with the computer at a local service and let them install a new cheapo psu in the system. He only uses the computer for eBay (collecting stamps and coins) and news.

So here's the 400w rated psu. Supposed to be 400w ... based on transformers it looks more like 250w max.
Full of CS capacitors, 560uf 200v 85c on primary, 13009 transistors, ie-33 transformer, i think it's the two transistor job for the 5v sb
3.3v, 5v, 12v is 22a , 40a, 18a but i doubt it, the beefiest diode that i can see on the heatsink is a SB3040PT which is a 30A shottky barrier rectifier.

If you check the pictures carefully, you'll see why it died.

Picture 001.jpg
Picture 002.jpg
Picture 003.jpg
Picture 004.jpg
Picture 005.jpg
Picture 007.jpg
Picture 008.jpg
Picture 009.jpg
Picture 011.jpg
Picture 012.jpg

[momaka]

Dead cap on 5VSB output, looks like. Easy fix. Yeah, it may not be a 400W PSU, but it's probably more than enough for that Sempron PC. Only concern is the 5VSB sections - it looks darkened. I don't see a critical cap on there, though. Perhaps it just needs new output caps on 5VSB (although if you replace the output caps on the rest of the PSU, that wouldn't be a bad thing eitehr).

Originally Posted by Behemot

That was the principle of the design so I don't understant your crap talk.

Well, you wondering why cheap PSUs almost always have their 3.3V rail filtering built a little better than the other rails? It's precisely because of that.

[mariushm]

Bingo. That schottky rectifier on the secondary also looks a bit cooked, may be dead.
Honestly, can't be bothered to carry it with me back to my place and fix it, all i have here are screwdrivers.
I would have loved to remove the big caps and measure them, for Uranium's thread, but it's just too much hassle.

[c_hegge]

I blew one of those PX-400s up in the 1012 el cheapo PSU roundup. It did 250W in spec, but it was pretty inefficient.

[RJARRRPCGP]

Still looks better built than the Deer PSUs I had.

[Behemot]

It's been selling like hot cakes under several names here like ten years ago…well, cheap craps is the most sold one, we have to face it

[goodpsusearch]

This thing is not that bad. It needs a full recap and some oil in the fan and then I would check that the caps spins fast enough to keep the psu cool.

[Behemot]

I am not really sure…doing like 60 % of rated power (and I won't even guess how much in spec) is not that bad?

[c_hegge]

^
As I said, it made it up to 250W with everything in spec.

[Heihachi_73]

Originally Posted by Behemot

In the long-term, I expect +3,3 V and finaly also +5 V to be exluded as well, possibly replaced with something like +24 V (for higher efficiency). But it won't be easy and probably will brake backward-compatibility.

To be fair though, a new PSU form factor would be no worse than the switch from AT to ATX in the 1990s. Even though ATX supported 12V, 5V, -12V (early revisions ATX only) and -5V like its predecessor, it was a completely different connector to the infamous 2x 6 pin AT connectors (where putting in the mechanically-identical connectors the wrong way would result in releasing the magic smoke from your board). I won't get started on Dell...

If a better PSU design means that an old Pentium II or Socket A won't boot with that PSU, so be it. If you were to shove 24V down a mechanically-compatible ATX connector, regardless of which pin was used (well, obviously not GND), your board would go boom anyway. Adding extra pins to maintain compatibility with old systems just makes things more bulkier and expensive, and also forces PSUs to become more cramped inside (if the same standard form factor, also from the late AT era, is used), since they have to support half a dozen different voltage rails as well as the mains input.

[cheapie]

I don't think it would be that hard to create an "ATX24V" (or "EPS24V") spec, as long as somebody produced compatible motherboards and adapters were available to convert the connectors to SATA and such.

Maybe something like a Molex Mini-Fit Jr. style connector (similar to the current CPU power connectors) would work. It could have five pins in a row - two 24V, two ground, and one for communication (like some sort of 1-wire protocol). All of those connectors would be mechanically and electrically compatible. The power supply would output 24V all of the time, oblivious to whether the computer is actually on. The motherboard would simply instruct devices to power on or off as desired.

I think that the same old form factor would be okay, as having a single-rail supply means no VRMs or anything like that, so there would be more space inside for other parts.

[c_hegge]

What I think would be good idea is to have a universal voltage for all ICs and components (CPU, RAM, GPU, chipsets, etc). That way, the PSU could go directly down to that voltage, and there would be no VRMs on motherboards.