SMPS in Parallel

[tom_pay]

Hi

A while ago in another post I was told not to put two switch mode power supplies in parallel because they would fail and could explode.

Could anybody please tell me why. I can't find out from Google.

Thanks

Tom

 

[cowana]

SMPS work by pulsing the output on and off very quickly. If you connect two together, they output from one could be pulsing when the output from the other is off, forcing voltage into the off one (damaging it).

However, the main issues are:
1) Feedback. SMPS watch the voltage on the output to tell it if it needs to have more or less voltage (ie make a pulse or not). If there is another power supply connected, it will not respond to feedback as it was designed to - if the voltage is too high, turning off the 'pulses' will have no effect. This means one power supply is very likley to shutdown (or both).
2) Current sharing. Whichever supply has a slightly higher voltage will take all the load (and if the load is too high, it will damage the SMPS). The power supplies will not equally share the current demand of the circuit - instead one will go into runaway mode, taking it all. If that one then shuts down from current overload, all the current will suddenly be dumped on the second one, which will also then have to pump current into the other, now shutdown power supply.

Not good - the system won't run happily.

 

[carmusic]

but you can get switching power supply that can be in parallel, they have a function to synchronise together and share current, they are more expensive though

 

[Arkham00]

I see that most servers have the option to add a second power supply unit. If the server has two power supply unit, the load is distributed on the two. If one fails or fails its line plug or is removed, the whole load is on the remaining power supply and the still server runs.

How do they get this ?

 

[tom_pay]

Ok, I see.

But wouldn't the SMPSs have output capacitors to regulate these on/off cycles? This would remove the on/off pulses making it pure DC. If the outputs were of equal voltage (hypothetical situation) wouldn't they be able to be connected in parallel?

Tom

 

[carmusic]

for server they often have a diode (or a perfect diode made with mosfets) at the output, they are just there if one supply fail the other one take the load. one power supply must be able to take 100% of load.

 

[indulis]

If you connect two together, they output from one could be pulsing when the output from the other is off, forcing voltage into the off one (damaging it).

False.

1) Feedback. SMPS watch the voltage on the output to tell it if it needs to have more or less voltage (ie make a pulse or not). If there is another power supply connected, it will not respond to feedback as it was designed to - if the voltage is too high, turning off the 'pulses' will have no effect. This means one power supply is very likley to shutdown (or both).

Maybe...

The feedback will function just like it should. If it's current mode control, the unit with the higher output voltage will supply the current until it starts to power limit. As the voltage starts to drop, the second unit will start to supply current. True, they will not share equally, and it's not very practical, but it will work. If it's voltage mode control, then strange things could start to happen.

2) Current sharing. Whichever supply has a slightly higher voltage will take all the load (and if the load is too high, it will damage the SMPS). The power supplies will not equally share the current demand of the circuit - instead one will go into runaway mode, taking it all. If that one then shuts down from current overload, all the current will suddenly be dumped on the second one, which will also then have to pump current into the other, now shutdown power supply

Any good design will will have current/power limit. Again, current mode will be OK and voltage mode the results may be wierd.

 

[carmusic]

but current limited smps are rare in cheap models, those that have current control have parrallel function most of the time

 

[tcmtech]

I think you guys are assuming theoretically perfect power supply's that don't exhibit real life output pull down effects.

Any number identical SMPS with DC outputs can be paralleled without problems. The one supplying the highest voltage, likely only millivolts more than the lowest, will get pulled down first until the the next highest one starts load sharing and so on and so forth.

Nothing explodes unless there is a gross mismatch of power supplies like putting a 5 volt supply in parallel with a 12 volt unit and the 12 volt one over voltages the 5 volts output capacitors and they burst.

 

[unclejed613]

and if really in doubt, you can "current share" the supplies with "load balancing" resistors, typically anywhere from 0.01 to 0.1 ohm depending on the current. it's just enough resistance to allow the supplies to operate independently with their own feedback loop, but still allow plenty of current to be available from either or both supplies to the load (it must be late, there's something weird about the way i worded that...). the same method is used with paralleled transistors (which have a tendency to "current hog") and paralleled amplifier outputs.

the cure for "strange behavior" between two identical SMPS supplies with voltage mode feedback would be to give them a common reference voltage source....