I have tried connect two 12V switch mode power supplies, plus to minus, to form a virtual ground and hence a dual rail supply. However, I am aware this is not ideal. The SMPS seem to interact or work against each other. One overheats and periodically cuts out. It eventually failed.
I have done this successfully before with transformer regulated supplies.
What is different about SMPS's that causes this and can it be overcome, perhaps by adding diodes in series with the output?
It's best not done at all. I never do it unless I know exactly how the supplies work, meaning when I have designed them from scratch by myself and synchronized the PWM controllers.
You can get all kind of nasties. Loop instabilities, hissing, plumes of smoke, molten solder splattering about, sirens wailing...
Switch mode supplies sample the output voltage and adjust accordingly. Now you have two samples simultaneously. Not good! The supplies will likely try to "fight" each other, while trying to stablize their outputs.
Not in parallel, series ("dual rail supply"). Which is even worse, as paralleled, the highest set supply will merely hog all the current. But with them in series, one will be short-circuited through the ground lead. How stupid!
At the very least, you have to cut the ground wire. But the ground is an integral part of the circuit, so you can't do that.
You might just maybe want to go back to linear supplies, or else get a
*proper* +/-12V supply.
Tim
-- Deep Fryer: a very philosophical monk. Website:
by ground you mean Protective Earth (ground is a TERRIBLE term). Its quite likely that each smps output is connected to PE, which would explain whats happening.
The OP can test this in 10 seconds flat with a meter.
But the PE connection is for safety reasons, and need not be to the 0V output. After all, 0V is relative. So If you took a +12V SMPS, cut the connection from PE to 0V (DONT remove PE entirely) then hooked it up to
+12V, one can happily re-define the former +12V as 0V, and the former 0V becomes -12V. Voila.
Of course if you remove the 0V-PE connection from one supply, then series them, this is already achieved. But modifying the individual SMPS would be (marginally) safer.
The problem could also be with cross-coupled loads (+12V -> -12V), but I dont think that would give rise to the OPs symptoms.
Your first suspect should be sequencing; if one PS comes up an instant before the second, that second unit sees (through the load) current from -12V source on its not-yet-charged +12V filter capacitor. This will kill power filter components.
A couple of clamp diodes to keep each PS from seeing reverse voltage would be useful.
Are these PC power supplies (or similar) where the 0V output terminal is also connected to the mains earth and the casing? If so then you may have shorted one of the supplies through the earth wiring of the two supplies.
If the above situation is not the case, then perhaps you are running the supplies too close to their current rating. If they have fold-back current limiting, then the second supply to start up might have a difficult job, depending on your load.
In any case it might be worthwhile to connect a suitably rated diode across each supply, so that in the event of a +12V to -12V short circuit, each supply will only see its output voltage drop to about zero, rather than the possibility of one of the supplies having the reverse polarity voltage forced across its output. With many switched mode supplies roughly the same protection is already provided internally because of the way the rectifiers are connected, but it is hard to know for certain without a schematic.
I have read this thread, and after much thought (approx
15 mSec), have decided to reply.
My employer makes both custom and "jelly-bean" SMPS.
So please believe me when I say that unless the units were designed to be paralleled, are to operate in series, Do NOT attempt it.
We had a customer that was too cheap to fork over the extra $2.85 for the parallel option, then had the audacity to complain that the units would not properly share current. What a poor excuse for an engineering manager.
Another customer, for a medical application, decided to series some units to make a 124Vdc output. And because the output caps were referenced to chassis (which is typically the protective earth for a Class 1 power supply, killed the 60V-rated caps. He also complained. What a maroon.
If the end-use application designer ignores the (published) conditions of acceptability for a unit, I have a rather anti-social trait of forwarding the (former) customer's company name to the U.S. CPSC. And if they are a client of an NRTL or NB, I also let the specific agency know.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.