3
\$\begingroup\$

Can two ATX SMPS be connected in series to get +10 and +24V DC (from 5V and 12DC output of each SMPS)

Needing lots of current at 24V with atleast 5A so cannot use -5V or -12V inside SMPS as negative voltage have less than 1A rating

I know transformer based PSU can be uses this way, but professor telling that SMPS cannot be connected in series the same way.

He not tell me reason as he says I will understand better next year when SMPS in course offering.

If he correct, then why not SMPS work like transformer based PSU?

\$\endgroup\$

6 Answers 6

7
\$\begingroup\$

No, you can't connect two ATX supplies in series and get double the voltage. There are many issues to consider, and most of them require detailed knowledge about the insides of each power supply.

One issue to consider is what happens when one supply is turned on, while the other is off? This can happen for many reasons, but #1 reason is that both power supplies will never come up at exactly the same time even if you use the same power switch for both. When one is on, and the other is off, the off supply will be subjected to -5 & -12v on it's outputs. This is a case that the supply was never designed to handle. Odds are that you'll blow up the caps on the output, and maybe a diode or MOSFET as well.

There are ways to protect against this sort of fault, but why bother? You will be much better off getting a supply that does what you want-- and the risk of burning down your house is much less.

\$\endgroup\$
2
  • 4
    \$\begingroup\$ But the main reason you can't connect two ATX power supplies in series is that the multiple outputs on each unit share a common ground. This means that there's no way to connect the two 5V outputs in series while simultaneously connecting the two 12V outputs in series. The ground of the "upper" supply can only be connected to one or the other. \$\endgroup\$
    – Dave Tweed
    Commented Jan 7, 2013 at 1:05
  • \$\begingroup\$ @DaveTweed And since the ATX output GND is connected to the AC input GND you would be shorting the output(s) of one supply to GND. \$\endgroup\$
    – user3624
    Commented Jan 7, 2013 at 1:40
3
\$\begingroup\$

Under certain conditions you can stack power supply outputs to get higher voltage, but there are caveats.

If the power supply rails have their output return tied to earth, you obviously cannot connect the return of one supply to the positive of the other supply without creating a short-circuit. The ATX spec v2.2 says:

3.4.6. Output Bypass
The output return may be connected to the power supply chassis. The return will be
connected to the system chassis by the system components.

This is ambiguous - the power supply return may or may not be earthed inside the power supply depending on the whim of the manufacturer.

David's earlier point is valid. Since you cannot guarantee that the output of each power supply stage was intended for paralleling, you most likely would have to add ORing diodes to provide some measure of protection and deal with the power dissipation.

The other issues that were touched on in earlier posts are valid. The outputs will not come up together, so your rail will go from 0-12V, then 12-24V in discrete steps. Also, overcurrent protection will not happen at the same threshold for both supplies.

You are definitely better off just purchasing (or designing) a 24V supply.

\$\endgroup\$
2
  • 1
    \$\begingroup\$ I don't think anyone is arguing for paralleling. TehNthDegree was the first to bring up the issue of load balancing. \$\endgroup\$ Commented Aug 18, 2011 at 21:20
  • \$\begingroup\$ ATX PSUs are always grounded. Never seen one made that's Class 2. \$\endgroup\$ Commented Jun 23, 2016 at 15:13
0
\$\begingroup\$

You CAN connect 2 SMPS in series, better if they are the same brand & model... because they are designed to behave as theoric batteries. This way of connection has some problems anyways, a price to pay for : Poor load regulation, partial or total lost of overcurrent protection and similar stuff.

Starting-conditions get worse but they are not bad at all because the baddest starting situation behaves as only-one supply starting by itself (the cheapest one).

Paralel operation is not suggested as current sharing is 100% unpredictable.

\$\endgroup\$
2
  • 1
    \$\begingroup\$ This might be true for two single-output SMPS units, but doesn't apply to the multiple-output ATX power supplies the OP is asking about. \$\endgroup\$
    – Dave Tweed
    Commented Jan 7, 2013 at 1:02
  • 2
    \$\begingroup\$ Many AC to DC SMPS are not completely isolated-- GND on the output side is connected to Earth GND on the input side. You can't connect those in series. Connecting them in series would effectively short the output of one supply to GND. The ATX supply that the OP mentions is one of these. \$\endgroup\$
    – user3624
    Commented Jan 7, 2013 at 1:38
0
\$\begingroup\$

Needing a very specialized supply that wasn't available in the voltage or wattage I wanted, I was considering to setup 2 of those PSU in series but found it very complex to manage. You have all the points that were already made with regards to grounding and timing of power-ons but there's also the issue that one of the PSUs will have double the voltage that it's 12v rail was designed for running through it. To make sure that it's not going to blow up, you'll have to go through every component of the rail, lookup it's model number on google and check what's the maximum voltage it's been designed to operate on.

\$\endgroup\$
0
\$\begingroup\$

well just for the &hit& and giggles I tried to do this but even though it worked connecting them continuously the Voltage didn't increase to 24 it was still 12

\$\endgroup\$
-1
\$\begingroup\$

The outputs of Switching Power Supplies can NOT be connected in either series nor parallel unless they are specifically designed to allow for multi-unit syncronization. Very few SMPS devices are designed for this.

The reason is that each SMPS is switching 20,000 - 80,000+ times per second. If you connect their outputs together without them being perfectly syncronized, they will not share the load properly and there can be very large voltage transients created which will destroy the SMPS components and possibly even the load you have connected to them.

\$\endgroup\$
2
  • 4
    \$\begingroup\$ Your statement about synchronization is false. I've never in 12 years of power engineering seen a power supply sharing scheme that involves frequency synchronization! Switching frequency is irrelevant when it comes to load balancing, which is generally done with DC current sensing and DC setpoint adjustment. \$\endgroup\$ Commented Aug 18, 2011 at 19:00
  • \$\begingroup\$ @Madmanguruman - Synchronizing switchers is tremendously important in ultra-low-noise analog systems (assuming you're forced to use switchers in such a system in the first place). However, it is indeed not needed just for load sharing. \$\endgroup\$ Commented Jan 7, 2013 at 2:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.