Putting computer server power supplies in series.

[Chris Wilson]

I want a regulated circa 100V DC at about 12 Amps and having two identical 50V high amperage (80A from memory, each) switch mode computer server supplies from HP with seemingly isolated outputs I want to try them in series. I have heard people talk of putting diodes in the output in case one starts before the other, how would I wire them, and what would be suitable devices please? Thanks

[rx8pilot]

What computer supply outputs 50v at 80A? That sounds insane!

Sent from my horrible mobile....

[Chris Wilson]

Sorry they are 2000 Watts so about 40A each, dozens on Ebay, cheap mas chips, new, refurbished and used.

[Jeff_Birt]

Many power supplies are not made to be connected in series (or parallel for that matter). Most PC type power supplies have the negative line bonded to earth, so at a minimum this must be lifted on the high side supply. Other types of switching power supplies use a bypass cap from the negative to earth, some may have an earth reference but have the negative rail raised above earth to lift it above the noise.

In general without knowing how the power supply is designed and constructed you don't know how big of a job it is to try and wire them in series. You also don't know how each supplies regulation scheme will react either.

[jeroen79]

seemingly isolated outputs

I'd first make sure that they are really isolated instead of seemingly isolated.

I'm not sure if diodes are needed.
You could wire one between the + and - of each supply so current can flow from the - to the + but blocks current from the + to -.
That way current can flow when one is off and the other on.
(with a diode voltage drop of course)
the diode would need to be rated for the maximum current you expect to put through it.

[elimenohpee]

Usually a diode is used when you want use them redundantly, but usually the primary power supply rail is slightly higher than the secondary so as to reverse bias the paralleled diode.  I'm not sure how effective that technique is at actually creating a redundant system....seems a little hokie to me. 

Be careful like others have said if you put them in series.  If they are earth-ground referenced, you have the chance of literally blowing them up depending on the built-in protection.

[CraigHB]

People commonly use a couple of the 12V server supplies in series for hobby chargers and those work fine after DC ground on the high side supply is isolated from the mains.  It's a pretty simple mod, just a couple connections.  I would suspect that's all that's required for a higher voltage supply, but of course I can't say that for sure.

[joseph nicholas]

One thing to be aware of is that most of these use a micro controller to regulate switching and current overload protection.

[BJ]

Correct me if I'm wrong, but can't pretty much all switching mode power supplies be run on straight (or even bush-mechanic filthy¹) DC, given that rectification is the first (non-filtering) thing they do inside the sliver box?

So, two big bridge rectifiers on the supply side, one for each PSU, and simple electrical isolation of their cases, ideally sealed inside a closed non-conductive box, to prevent accidental shorts between floating grounds. Perhaps even move just the bare boards into a custom enclosure. That would make a considerable amount of sense if you were working with regular PC ATX style power supplies and wanted to tame the tangle of wires those things sport. Not to mention improve the isolation of the unused output rails.

If i'm correct, any number of PSUs could be arbitrarily stacked, provided one and ONLY ONE voltage rail is used on each supply. With that as a given, I don't see why the output side can't be treated like a regular transformer with multiple rectified taps. (ie, A certain amount of respect, and just perhaps not like our friend in the Old Dart, Photonicinduction.)

^1.Damn, I really am feeling my age, I even briefly considered saving four diodes by feeding each supply only a half wave each, but high current diodes don't cost $20 bucks a peice any more, and the price of big fat smoothing capacitors has NOT scaled proportionately.

[CraigHB]

...and the price of big fat smoothing capacitors has NOT scaled proportionately.

I've been going to arrays of chip capacitors over aluminum electrolytics lately.  Can't always go that route, but a lot of times I can.  The ESL and ESR is hugely lower and you don't need as much total capacitance to do the job.  Though it's a lot more expensive than using electrolytics.  It saves a lot of space and provides a more ideal capacitance so I just pay my nickel.  Would be great to see cheap higher value, higher voltage chip capacitors.  They've actually gone down in price a lot over the last few years, but it's still a lot more expensive to use a chip capacitor array compared to an aluminum electrolytic capacitor.

[edavid]

I want a regulated circa 100V DC at about 12 Amps and having two identical 50V high amperage (80A from memory, each) switch mode computer server supplies from HP with seemingly isolated outputs I want to try them in series. I have heard people talk of putting diodes in the output in case one starts before the other, how would I wire them, and what would be suitable devices please? Thanks

Diodes would only ever be used if you were trying to make a center tapped +/-50V supply.  If you just put them in series for 100V, it's not an inssue.

[Cerebus]

What computer supply outputs 50v at 80A? That sounds insane!

Sent from my horrible mobile....

Would it sound less insane if the magic number was 48V? Lots of stuff runs off an internal 48V DC bus and then hangs DC to DC convertors off it at the point of use. It gives you the option of running from a 48V telecom supply instead of the mains if that's the world you're working in. It's quite common in the telecom and data centre world to see 48V supplies at 100s of amps. This leads to really thick cabling (often busbars) and impressively huge fuses - DC fuses tend to be much physically bigger than their AC counterparts

[rx8pilot]

Yes, that does make sense. My first thought was a more consumer level. 48v as a primary distribution voltage in a rack does indeed make sense.

:-)

Sent from my horrible mobile....

[David Hess]

Diodes would only ever be used if you were trying to make a center tapped +/-50V supply.  If you just put them in series for 100V, it's not an inssue.

It is still an issue if overload forces one of the power supplies into current limiting before the other; the power supply which current limits first may not like having a high current negative voltage applied to its output.

My dual tracking power supplies which are effectively two power supplies connected in series have reverse protection diodes to prevent this very problem.