Sources for high amperage 2:1/1:2 transformers?

A

AreEss

2[H]4U
Joined
Jul 5, 2005
Messages
2,180
Okay, I'm stumped. I know these are out there. I just can't find them. Here's the problem:

BIG FRIGGING INDUSTRIAL PSU (22A @ 24V) -> ??????????? -> Lots Of Tasty Hard Drives!

Obviously, the question marks are where the problem resides. In order to properly handle these drives (which do NOT have a DLY jumper) I want to play it safe with about 26A on the 12V and 22A on the 5V. So, as we all can see, I need to build a switched step-down.

More specifically, a 2:1 fused stepdown that'll handle 30A inrush and a configurable 5V output stepdown that'll handle 25A inrush.
(I will now deny any and all charges of "overbuilding." I'm just paranoid.)

I don't suppose anyone's seen something like this out in their catalogs?
 
12V @ 30A is 360W, which isn't hard to find. You can probably find a 24->12V full brick from Datel, Vicor ($$$), Cherokee, etc... 5V @ 25A can definitely be found from one of these companies, probably even in a half brick.

Only thing is, you've just bought 4-5 known brand computer power supplies if you buy these bricks. I'd stick with computer power supplies...
 
I covered in another thread why a standard pc power supply is not going to cut it. Trust me, they aren't. I've been doing this sort of stuff for nigh upon ages. The problem is simply that I am not going to spend $900 for the N+1 modular system I really should be using.

The drives are not Western Digital crap; these are FC-AL 2Gbit 10,000RPM beasts - 8 to 10 of them. (Also, for the record, those numbers from WDC are useless.) Startup draw is incredibly high because of the motor; normal operation is a lot more reasonable. I could use a 280W industrial +5/+12 if I could do a 3s delay per drive, but that's not possible with these. Once they're up and running, they're very low draw idle.

I'll check Datel, Vicor, etc. Maxim comes so close, but they lack the amperage.

(EDIT)
Remember how I said PC's don't cut it? Here's some tasty evidence thereof;
http://www.pcpowercooling.com/products/viewproduct.php?show=S36ATX&view=techspecs
The Silencer 360W cannot do it. Sure, it's got the +5V load, but it's short on +12V. All the amps are over on the 3.3V line, which I don't even use. Even if I did need the 3.3V, I'd need at most 1.25A. At 3.3V, that's pretty much nothing.
 
I've checked a few catalogs which I have for DC/DC bricks, and it looks like buying two 24->12, 24->5V bricks with the ratings you want is going to push the cost up close to the $900 you're trying to avoid. Same goes for off-the-shelf AC/DC power supplies.

How about a PCP&C Turbo-Cool 510?

http://www.pcpowercooling.com/products/assets/T51ASL/specs.pdf

34A on +12V, 40A on +5V.
 
The 470 cuts it too close, and the 510 won't fit with the temperatures. (Limited cooling, which is why industrial. Stuff that doesn't start dropping till past 70C.) :(

Thinking about just doing it from scratch with an AC/DC inverter at this point.. :(
 
You should check eBay, as big-iron NAS/SAN arrays sell there every day, along with redundant power supplies and things like that. Lots of stuff sells for much less than they should, so check it out.
 
And if I had the room for off-the-shelf, I'd just use my Hitachi 9K parts that are literally collecting dust. Unfortunately, I don't have that luxury. I have very limited room, so I have to cram things in and make them work that way.
 
AreEss said:
The 470 cuts it too close, and the 510 won't fit with the temperatures. (Limited cooling, which is why industrial. Stuff that doesn't start dropping till past 70C.) :(

Thinking about just doing it from scratch with an AC/DC inverter at this point.. :(
Click to expand...
The PCP&C is good for 50C... can your hard drives, computer motherboard, etc do that? Almost every computer motherboard I've ever seen is only good up to 50, and even most Intel thermal solutions (Intel HSF's) are only good for 40 ambient...

What's the application anyway?
 
Right, 50C internal. The closed housing of the 510 makes it run hotter than an open frame PSU, and NFW I'm gonna void warranty on a PCP&C 510.
There is no motherboard in there, but for the record, most of my stuff is rated to 60C if it's not -40 to +85C. I don't do chintzy consumer crap anymore.

As far as the application, it's an external drive cage. With very heavy duty drives, rated to 55C operating. I know full well that a closed frame PSU is going to get too hot, because the space contstraints prevent me from mounting it directly to the rear plate.

Which means my best bet is probably to hack it into an open frame configuration. I'm seriously considering two 1U PSUs though. Figuring to rewire the whole thing, run them to a PCB to combine the 12V, 5V, grounds, and adjustable resisitve loads for unused lines, and just run the connectors to the drives from there. My concern is that they won't actually balance, and one PSU will get more load than the other. :(
 
Well, two problems with that. I looked at it before.
One, not enough amperage on the F3A and F2B. I'd have to get the F1C, then step-up the +5 to +12 to get roughly +5V/40A +12V/32A, before factoring step-up/step-down losses.
Two, you don't even wanna know what it costs. Even with the discounts I get... no.

Combining two 1U ATXes is looking like a better option.. especially since I found a pair for <$20.
Just gotta figure out how to combine them properly.
 
If the power supplies aren't designed for it, don't just wire them together... use one power supply for each half of the drives.

"Proper" redundant PSU's are designed with current sharing capabilities, remote voltage trim and sense, and a separate board which diode or mosfet OR's the outputs together. Making a typical TL494/KA7500-based, silverbox ATX supply capable of this is next to impossible.
 
Goddamnit, I just lost a long huge reply with lots of detail. Now I gotta start again. FUCK.

Anyways.

Yeah, I'm well aware of what goes into a proper redundant system. 90% of the desktops I build have them. And somewhere around here I have a proper ATX-GES combiner, which does it's own volt-sense and distribution.

The way I see it, what I need is not as complex as everyone seems to think. This isn't redundancy; it's simple load sharing. There's no way it could be redundant; one PSU alone cannot handle random seek load on the +12 if I got all the drives going. Forget spin-up; I'd create a frag grenade out of the PSU.

So, since I only need to combine, I don't really see a need for anything more than a cut-out rigged to the PW-ON if I lose voltage on one PSU.

What I was figuring I might be able to get away with is something like this, since certain pins must be combined anyways:

PSU0 +3.3V(1&2) -> 2.5A Resistive Load (to meet min)
PSU1 +3.3V(1&2) -> 2.5A Resistive Load (to meet min)

PSU0 PW-ON -> Common PW-ON <- PSU1 PW-ON
PSU0 GND(15) -> Common PW-GND <- PSU1 GND(15)

PSU0 +5V -> Ferrite! -> Maybe a big +5V capacitor cycle-matched? -> Common +5V
PSU1 +5V -> Ferrite! -> Maybe a big +5V capacitor cycle-matched? -> Common +5V
PSU0 +12V -> Ferrite! -> Maybe a big +12V capacitor cycle-matched? -> Common +12V
PSU0 +12V -> Ferrite! -> Maybe a big +12V capacitor cycle-matched? -> Common +12V
PSU0 GNDs -> COMMON GROUND <- PSU1 GNDs

Obviously, I have to keep up min. load on both PSUs, but I really only need to split. If I lose one or the other, I'm dead no matter what. Bear in mind, I'm no electrical guru, but the way I see it, there's no need or reason to go for a full out redundant. The PSUs I'm looking at do provide Vsense (though I'd have to tap it) but it's not necessary. I just need to evenly split the amp load.
 
I'd still say forget the combining... put half the drives on one power supply, half on the other. Connect the grounds and PW-ON's, nothing else.

Doing the wired-OR of power supplies really isn't a good idea if the power supplies don't support current sharing - You might end up with one power supply sitting overloaded all the time while the other puts out next to no power. Your only adjustment being the pots within the supplies if they exist, which you can tweak for decent balance and then cross your fingers it never drifts with temperature/time/load/etc.

And this completely ignores the fact that if you connect two dissimilar power supplies with different control loops, you can end up with two supplies that "see-saw" current back and forth into each other and explode... Additionally, computer SMPSes go unstable with too much capacitance, refer to the ATX spec for the limits. The output capacitance of PSU1 could make PSU0 unstable, and vice versa...

If you had 3 power supplies in your hands with current sharing capabilities driving a common bus, so that if you lost a supply you'd still keep running, doing a system like this would make perfect sense. But since you don't have any redundancy at all (one PSU down = failed system) really there's no point in the combining to begin with.
 
gee said:
I'd still say forget the combining... put half the drives on one power supply, half on the other. Connect the grounds and PW-ON's, nothing else.

Doing the wired-OR of power supplies really isn't a good idea if the power supplies don't support current sharing - You might end up with one power supply sitting overloaded all the time while the other puts out next to no power. Your only adjustment being the pots within the supplies if they exist, which you can tweak for decent balance and then cross your fingers it never drifts with temperature/time/load/etc.
Click to expand...

Well, file me under tard for forgetting to mention parts again. No, they don't actively support current sharing. But no, the internal pots aren't the only control I'm planning on. I'm figuring on pots on the actual board where they connect as well, I just haven't got a clue where exactly I need to put them. I thought about a variable resistance pot on the 3.3V, but then I've got uneven 3.3V load.

And this completely ignores the fact that if you connect two dissimilar power supplies with different control loops, you can end up with two supplies that "see-saw" current back and forth into each other and explode... Additionally, computer SMPSes go unstable with too much capacitance, refer to the ATX spec for the limits. The output capacitance of PSU1 could make PSU0 unstable, and vice versa...
Click to expand...

Actually, no it doesn't; the two PSUs are exactly identical across the board, right down to the capacitors. That's also why I want to have some sort of isolation between the combined feed and the lines feeding it; I know all about backfeed and rebound, and figure the easiest way to do that is to put something in the way to keep it from backfeeding.

If you had 3 power supplies in your hands with current sharing capabilities driving a common bus, so that if you lost a supply you'd still keep running, doing a system like this would make perfect sense. But since you don't have any redundancy at all (one PSU down = failed system) really there's no point in the combining to begin with.
Click to expand...

Sure there is; $20 for two PSUs versus $130+ for one that does the job. I can easily change it to four, but then my temperature management is screwed to hell. Even three would screw it up, so either way I'm probably going to have to redo the cooling. The drives run hot, the PSUs run hot, and there's not a lot of room for fans.
And theoretically, at idle, it would be possible to run all drives off one. Plus the 'hey look what I did' factor is always cool, especially since the whole assembly's already going to be custom built. ;)
 
You must log in or register to reply here.
Back
Top