Paralleling power supplies

[kc1980]

Is it fairly common (or safe) to parallel several power supply outputs?  For example, I have an HP dual output power supply and BK Precision single output power supply.  I want to maximize current output to a load, so I set all three outputs to 5V and hook them all up in parallel.  All outputs are isolated from each other.  Does this work?

What other considerations should I keep in mind when doing this?  E.g. source impedance of each output, setting same current limit for each output, etc.

[EEVblog]

Yes, paralleling PSU outputs is quite common and generally it's quite safe to do, as there is usually sufficient output impedance to handle the load sharing. Sometimes though a supply might not like being powered through it's output terminals when it is off or at a substantially lower voltage, so ensure they are the same before connecting or flicking the load switches.

Dave.

[Mechatrommer]

so is that means, both of the following method are safe and can be used?

1) two identical ACDC adapters paralleled together to get double maximum Ampere.
2) two identical voltage regulators paralleled together to get double maximum Ampere.



if paralleling "n" number of devices with "A" ampere maximum for each device, will we get maximum current of (n X A) ampere?

[kc1980]

I'm almost certain that scenario number 2 will not work.  I don't think the two regulators will split current evenly.  One will end up hogging all the current.

[Simon]

I once paralleled the two 5V outputs of two computer power supplies and this did not go as planned

[mkissin]

It can be a good idea to add series diodes to the outputs of each of the individual supplies, just to stop power flowing backwards into one of them, if they aren't designed to be paralleled. Basically you're looking at paralleling the modules via the "droop" method, and should be aware that it's quite possible for one module to end up attempting to supply all of the load power, at least until the load becomes large.

Increasing the current out of an LM317 as you show there in scenario 2 should be done with an external pass transistor, not by paralleling several modules. The datasheet for the device should give an example circuit.

[eswets]

Years ago I built a power supply that had the regulators in parallel.  I will look for the original schematic (came from a book of tested schematics) but I do believe that it has a diode.  I have used it for the past 7years with no issues.

[saturation]

Yes, I would do the same.  The problem with paralleling two voltage regulators is that they will not track identically although they can be close, causing one to take more load than the other.  Using a pass transistor is more reliable and is a standard technique to increase the power rating of the LM317.



http://www.eleccircuit.com/lm317t-voltage-regulator-with-pass-transistor/


To parallel regulators, you need to add more sensing circuitry:



http://electronicdesign.com/article/power/double-your-output-current-with-parallel-voltage-r-9270.aspx

http://www.reuk.co.uk/High-Current-Voltage-Regulation.htm


Which technique to use depends on how much current you'd like to source.

It can be a good idea to add series diodes to the outputs of each of the individual supplies, just to stop power flowing backwards into one of them, if they aren't designed to be paralleled. Basically you're looking at paralleling the modules via the "droop" method, and should be aware that it's quite possible for one module to end up attempting to supply all of the load power, at least until the load becomes large.

Increasing the current out of an LM317 as you show there in scenario 2 should be done with an external pass transistor, not by paralleling several modules. The datasheet for the device should give an example circuit.

[ModernRonin]

> It can be a good idea to add series diodes to the outputs of each of the individual supplies, just to stop power flowing backwards into one of them, if they aren't designed to be paralleled.

I thought about this approach myself, but won't the 0.6-0.7V voltage drop caused by a silicon diode mean that you don't get the voltage level you want? Yeah, if you're willing to spend the big buxx, you can get ~0.1V drop germanium diodes, but even that may screw up your voltage significantly if you're running at 3.3V.

I figure if you want to parallel linear regs, you need to use adjustable regulators, and set them up so they feed Vcc + (diode drop) into the diode. Then it'll work like you expect. It's tempting to try and sense the voltage/current on the far side of the diode and feed it back to your adjustable reg, but I suspect that's gonna lead to problems with one regulator swamping the other...

Edit: And saturation beat me to it while I was writing this comment. Never mind!

[kc1980]

> It can be a good idea to add series diodes to the outputs of each of the individual supplies, just to stop power flowing backwards into one of them, if they aren't designed to be paralleled.

I thought about this approach myself, but won't the 0.6-0.7V voltage drop caused by a silicon diode mean that you don't get the voltage level you want? Yeah, if you're willing to spend the big buxx, you can get ~0.1V drop germanium diodes, but even that may screw up your voltage significantly if you're running at 3.3V.

How about Shottky diode?  Wouldn't that do the trick?

[ModernRonin]

> How about Shottky diode?  Wouldn't that do the trick?

I wish it was that easy. The great thing about the Shottky is that it's a "fast recovery" diode, i.e. it can switch on and off really fast. So you can use it in RF circuits and other fast switching applications.

But it still has "voltage drop between approximately 0.15–0.45 volts"

Also, "the high reverse leakage current [of the Shottky] presents a problem in this case, as any high-impedance voltage sensing circuit (e.g. monitoring the battery voltage or detecting whether a mains adaptor is present) will see the voltage from the other power source through the diode leakage."

- http://en.wikipedia.org/wiki/Schottky_diode

[ModernRonin]

> i see there is still a limit in current (~4.3A), what if we need more?

You can usually get a bigger transistor:

http://www.mouser.com/ProductDetail/STMicroelectronics/2N3771/?qs=sGAEpiMZZMuDHCWHhuFg%252baPN204G4fBcX0boRRvcGRo%3d

That'll do 150W, which is 10A @ 15V. The max current it can flow is 30A (@ 5V.)


If you need more than 30A (offhand, the only things I can think of would be a *very* fancy audio amp, or a motor driver), you'd be better off using a MOSFET power transistor instead of standard NPN transistors.

[kc1980]

kc1980: "I have an HP dual output power supply "
ALL PC PSU that i've come accross (with multiple output), all output at the same Voltage level are joined together inside in one big soldered copper plate, so paralleling the output will not increase the current, its just the effective conductor surface area is doubled. but i'm not sure in your HP or other PSU case.

Oh, i think there's a misunderstanding.  It is an HP/Agilent lab power supply -- not for desktop computers.  Therefore, the outputs should be isolated.  I understand that ATX-type power supplies are NOT isolated.

[EEVblog]

Simon: "I once paralleled the two 5V outputs of two computer power supplies and this did not go as planned"
maybe this proves that Dave's postulate is not entirely applicable:
Dave: "Yes, paralleling PSU outputs is quite common and generally it's quite safe to do"
anyone? or maybe, the PSU's are the cheaper/non quality one?

I did say "generally". It's not always applicable as you mention. In many cases the current won't be shared equally, but that can be ok depending upon what load you require and the overhead on the individual supplies.

I forgot to mention that in diagram A) it would be standard practice to actually parallel the secondary windings of the transformer for increased current (watch the DOT polarity!).

Dave.