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

[momaka]

Originally Posted by tom66

I guess it doesn't use a magamp for the 3.3V, but runs it directly off the transformer.

It's possible, but this is very rare. I've actually seen that only in a cheap CWT-built Cyberlink with a tiny EI-28 transformer.

If you don't see a second choke on the secondary side, *most* of the time you can assume the 3.3V rail is regulated in with a MOSFET and a 431 shunt regulator.

[rogfanther]

Originally Posted by Wester547

I'm interested in convection cooling!

So thick heatsinks are ultimately the best for heat transfer in PSUs, like those found in older Hipros, Newton Powers, Liteons, and Deltas? I would think that thinner heatsinks would get hotter because while they do transfer heat they do not do so as effectively by way of more limited surface area, and as stated before me, too many fins isn't necessarily good. If anything else, though, I thought thicker heatsinks hugely help mediocre capacitors last much longer than usual, though well built voltage filter circuits and overbuilt rectifiers can do just that, along with fans that blow copious amounts of air. I also think vents on the side and large vents on the back and bottom help too....

Thick heatsinks, along with a lot of area exposed to moving air, would be the best. But then you include engineering constraints, space, and mostly cost, and that has to be reduced.

[Wester547]

If 3.3v is generated from a MOSFET in PSUs that don't have a secondary coil (that is to say, there is only one toroidal coil on the secondary), does that mean...

In a Hipro 300W that has only, say (mine):

2x MBR3045PT in parallel (2x30A/OnSemi/TO-247)

1x P40NF03L (STMicroelectronics/TO-220AB N-Channel StripFET Power MOSFET/40A/30V reverse voltage/0.020 ohm)

1x MBR20100CT (1x20A/OnSemi/TO-220AB)

Does that mean the P40NF03L is not the rectifier for the 3.3V rail but what it's generated from because of the lack of a secondary choke/coil and that it's only 1x30 for the 5V rail (rather than 2x30), 1x30 for the 3.3V rail, and 1x20 for the 12V rail (there's also a 7805 voltage regulator from STMicroelectronics for the -12V rail on another heatsink)? There are also four ferrite coils for the PI filters.

[Behemot]

Isn't it DC-DC from, say, 12 V than?

[momaka]

Originally Posted by Wester547

IDoes that mean the P40NF03L is not the rectifier for the 3.3V rail but what it's generated from because of the lack of a secondary choke/coil and that it's only 1x30 for the 5V rail (rather than 2x30), 1x30 for the 3.3V rail, and 1x20 for the 12V rail

I'm not familiar with that 300W HiPro, but I have the 250W HP-P2507F3P and that is indeed the configuration I have (except that mine has a 40A rectifier on the 5V rail... IIRC).
I don't think the 7805 is for the -12V rail, though. Most likely it's a 7912 (a -12V regulator) or if it really is a 7805, then it has to be for something else.

There are 2 common ways to get the 3.3V rail with the MOSFET:
1) Put a very strong rectifier for the 5V rail and use the rectified 5V rail to generate the 3.3V rail with the MOSFET.
2) Have a separate rectifier for the 3.3V rail (the DC output is usually higher than 3.3V... more like 4V or maybe even higher), then a simple filter for it, and finally use a MOSFET to regulate it to 3.3V.

My HiPro above is like 2). I also have a cheap Raidmax 380W PSU built by Sun Pro that uses 1). Both can be equally good methods if designed properly.

Originally Posted by Behemot

Isn't it DC-DC from, say, 12 V than?

Nope. At least not for the PSUs that use methods 1 and 2 above (those methods, however, do use the 12V rail to drive the MOSFET).

I think I've seen some high-end PSUs that actually do DC-DC from the 12V rail for both the 5V and 3.3V rails.

[Wester547]

Originally Posted by momaka

I'm not familiar with that 300W HiPro, but I have the 250W HP-P2507F3P and that is indeed the configuration I have (except that mine has a 40A rectifier on the 5V rail... IIRC).
I don't think the 7805 is for the -12V rail, though. Most likely it's a 7912 (a -12V regulator) or if it really is a 7805, then it has to be for something else.

There are 2 common ways to get the 3.3V rail with the MOSFET:
1) Put a very strong rectifier for the 5V rail and use the rectified 5V rail to generate the 3.3V rail with the MOSFET.
2) Have a separate rectifier for the 3.3V rail (the DC output is usually higher than 3.3V... more like 4V or maybe even higher), then a simple filter for it, and finally use a MOSFET to regulate it to 3.3V.

My HiPro above is like 2). I also have a cheap Raidmax 380W PSU built by Sun Pro that uses 1). Both can be equally good methods if designed properly.

It's the HP-P3527F3 I posted a while ago. I want to know what it means for the 3.3V rail and 5V rail power wise at that point - there are two 30A TO-247 parts in parallel on one end and the 40A Power MOSFET/voltage regulator on another, and on the back of the heatsink a 20A part for the 12V rail. This PSU doesn't have a second choke on the secondary so 3.3V is generated from that 40A voltage regulator. Does this mean 30x1 for 3.3V and 5V? Or is it more complicated than that?

[Behemot]

You don't really watch the market, do you guys? Basically all todays higher-end PSUs use DC-DC from single very powerfull 12V rail. How else would you achieve close to 90% efficiency, right (also note that blade servers and this stuff uses this for a long time - single higher-voltage rail with huge efficiency). Thats why it came to my mind in the first place, it would be only reasonable to find it in business class-PSUs (like most Hipro OEMs are)

[Wester547]

Well, note that my Hipro has a manufacture date of March of 2006 (and it's an old design since it has a -5V rail, which I think 7805/7905 is for judging by its position), and I'm sure momaka's is older than that...

I guess my question more simply is, do those MOSFETs/voltage regulator (in my case, a 40A/30V reverse voltage one) that generate 3.3V on PSUs that don't have a second choke on the secondary contribute at all to the amount of power the 3.3V rail can provide (assuming there are two other 30A 45V reverse voltage rectifiers very close to each other and one other 20A 100V reverse voltage one as in this case), or are they just there as a voltage regulator and even as a protection circuit of sorts?

[momaka]

Originally Posted by Wester547

It's the HP-P3527F3 I posted a while ago.

Without underside pictures, I really can't say. But it's very likely either method 1 or 2 I mentioned above. If it's method 1, then the 2 rectifiers are in parallel and 3.3V rail is produced by the MOSFET regulating down the 5V rail to 3.3V. If method 2 (like my HiPro) then one rectifier is for 5V and other is for 3.3V.

Originally Posted by Behemot

You don't really watch the market, do you guys?

Not a whole lot, no.
I did look at a JhonnyGuru review a few years ago and I remember how he pointed out that the PSU transformer didn't really have outputs for 5V and 3.3V rails but rather they were produced from the 12V rail via DC-DC buck regulator. And since it was VRMs, there were poly caps in there.

I wouldn't be surprised if in the future they come out with a PSU that produces a single powerful 24V rail that is then regulated down via DC-DC buck regulators for 12V, 5V, and 3.3V. Of course the -12V will still have to come out from it's own transformer winding.... unless they either remove the -12V from the ATX spec or the PSU manufacturers produce -12V rail with a capacitor-diode charge-pump circuit (like they do on the t-con of LCD monitors).

[Wester547]

3.3V can't be generated or regulated from/by any sort of PI filter coil (since it is an inductor and/or choke of the sort), can it, even if there's four of them (and if they're profuse in individual coils however small) and one toroidal coil (I know PI filter coils are noise suppressors)? And a power MOSFET transistor (or voltage regulator as some call them) can't be used as a rectifier, correct? ^^;

[tom66]

Heh, I'm still in favour of moving to 36/40V computers.

[Wester547]

This might sound dumb, but just to be sure, do PSUs that use the DC-DC from 5V method to generate 3.3V (via another MOSFET with no dedicated rectifier for the 3.3V rail) have separate filtering capacitors for the 3.3V and 5V rail or do they share the same set of capacitors? ^^;

[mariushm]

It depends on the power supply.

Modern designs use DC-DC converters to generate 5v and 3.3v from 12v, this way the power supply can deliver almost all its rated power on 12v if needed and if the 3.3v and 5v are not used.
They're separate in the large majority of designs.

There were a few power supplies based on old designs that had a lot of current on 5v, sometimes even close to 35-40A on 5v - the companies adapted the design to use a dc-dc converter to generate 20-25A from that 5v. This gave them more room for 12v output.
It doesn't make these more efficient or better imho.

You still need capacitors for ripple filtering on all rails, no matter if you use dc-dc converters or not.

[momaka]

Originally Posted by Wester547

3.3V can't be generated or regulated from/by any sort of PI filter coil (since it is an inductor and/or choke of the sort), can it

It can. On group-regulated power supplies with the 2 chokes on the secondary, you'll see a small 3rd choke *usually* between the secondary heat sink and the main transformer. This is what regulates the unrectified 5V rail down to about 3.3V.
A choke/inductor is simply a low-pass filter in a circuits that allows DC to pass and attenuates or completely removes AC (depending on inductor turns, core, etc). On the rectified (DC) side of a rail, it only removes noise and ripple. On the AC side, like that 3rd choke I mentioned, it can attenuate the AC to lower values. In the case of a power supply with 2 chokes, the 3rd choke attenuates the AC coming from the 5V pins on the main transformer and brings it down closer to 3.3V. This AC is then rectified into DC just like the other rails and you get 3.3V.

Originally Posted by Wester547

And a power MOSFET transistor (or voltage regulator as some call them) can't be used as a rectifier, correct? ^^;

A MOSFET is a MOSFET. It can be used to regulate voltage but calling it a "voltage regulator" is not quite correct (it really depends on the circuit).
And no, a MOSFET can actually be used as a rectifier. PSU's that use MOSFETs instead of rectifiers are called "synchronous rectifier" PSUs. They are highly efficient. Many OEM laptop power adapters use this topology due to its efficiency. The PS3 power supplies also use that topology. That's why they are so small and don't require much cooling. A PS3 PSU is not even half the size on an ATX PSU, yet it can do 32A on the 12V rail (just shy of 400W) without much cooling or big heat sinks.

[Wester547]

^ Very informative, and that's certainly more power than the 170W (14.2A) on my Xbox 360's 12V rail, a lead free, push-pull topology, 175W (total) PSU from Hipro. :P (it only otherwise has 1A on the 5VSB rail). That has a small fan internally, though. I still worry about capacitors in power bricks and AC adapters, though, since I know OEMs don't have the best choice of capacitors and I know they'd benefit highly from more airflow, IE a fan over them.

But... I don't understand how that 40A N-channel MOSFET transistor plays no role in providing power for the 3.3V rail in my Hipro, then... I understand it uses DC-DC conversion from the two 30A parts on the 5V rail but I don't understand why they'd place a 40A part as something to generate 3.3V from (rather I don't get why they just wouldn't use something smaller if it plays no part in giving 3.3V power in that method of generating/regulating it). As stated, there's one toroid coil and four PI filters on the secondary. I guess what I'm saying is, though 3.3V does not have its own rectifier but a regulator and DC-DC conversion from 5V, I don't get why it can't get any power from the power MOSFET regulating it even if it uses the 5V circuit totally to achieve the combined power. ^^;

[pdavid]

I've looked into some brand new Kolink psus that came with oem cases. Amazing...

One has no model number or anything the other is a Kolink CHS-450

-no graetz bridge just four 2amp diodes
-fake primary caps, real capacitance is 220uF
-no brand 13007 switchers
-2n60 fet for standy
-fake, tiny el-33 trafo
-label claims 5V-25A; 3.3-14A; 12-26A. Diodes are SB2045; SB1640; F1020
-no pi filter, some secondary caps.
-awg 20-22 wires

Strangely one has some emi filtering and connectors instead just plain soldered wires.
The pc cases are also horribly made, you can cut your fingers practically anywhere. The overall builld is paper thin mess.
Avoid.

[Shocker]

It's EI-33 not EL-33.

One of the members here is a common victim of the reverse mistake.

[goodpsusearch]

[Behemot]

At least it does not have diode treatment for +12 V. That's progress, no?

[pdavid]

Originally Posted by Behemot

At least it does not have diode treatment for +12 V. That's progress, no?

Of course, that's the 26 amps right there! And to think one of my coworker has got one of these to save some cash...