Suppose I have something like 50 independent switching regulators all tied a single load and ideally sharing current. I could provide isolation for each regulator using a diode.


simulate this circuit – Schematic created using CircuitLab

Or I could provide GND isolation using a transformer.


simulate this circuit

For such a large number of power supplies in parallel, is there a risk of having massive GND noise on the shared GND plane from all the switching components? Would GND isolation help address this issue? If GND isolation is not required, what design considerations should be taken into account when sharing the GND?

  • \$\begingroup\$ Do you have reason to believe that they are all at different ground potentials? \$\endgroup\$ – Ignacio Vazquez-Abrams Oct 4 '15 at 0:32
  • 1
    \$\begingroup\$ Why do you need 50 independent PSUs in the first place? \$\endgroup\$ – Golaž Oct 4 '15 at 12:18
  • \$\begingroup\$ I'm trying to ascertain if having this many switching components would cause massive GND issues. 50 PSUs is overkill, but it does present an interesting scenario with regards to GND noise and isolation requirements. \$\endgroup\$ – TRISAbits Oct 4 '15 at 17:29
  • \$\begingroup\$ I reworded the question in function of your feedback. I admit that the previous wording was lacking. \$\endgroup\$ – TRISAbits Oct 4 '15 at 17:36

In general you do not need or want to isolate the grounds.

However you can't usually just parallel power supplies. They will not share the load equally.

There are power supply module that are designed for being placed in parallel. Often they are intended to not only increase output but allow "hot-swap" repair. They are often referred to as being in an "N+1" load sharing configuration where there is one more power supply than is actually needed so that when one fails the remaining good ones can power the load while the failed one can be unplugged and replaced with a good one. These are used large rack-based server systems and for telecom applications.

The power output and grounds of all the supplies are connected directly together and often there is a load sharing signal that also connects to all the supplies.

Why do you want to put many power supplies in parallel?

| improve this answer | |
  • \$\begingroup\$ I realize that the load needs to be shared, hence the ideal current sources in the diagram. What I'm trying to ascertain is whether hooking up this many supplies in parallel would cause trouble on the GND plane, and thus, if GND isolation is needed. \$\endgroup\$ – TRISAbits Oct 4 '15 at 17:21
  • \$\begingroup\$ I don't necessarily want to place this many supplies in parallel: it's just an example to help explain my question. \$\endgroup\$ – TRISAbits Oct 4 '15 at 17:22
  • \$\begingroup\$ I reworded my question to address the emphasize that current sharing somehow works across 50 PSUs (as unlikely as that can be). I failed to make it clear originally. \$\endgroup\$ – TRISAbits Oct 4 '15 at 17:42
  • \$\begingroup\$ I've seen those N+1 supplies before, and while the outputs are tied together (same as diagram 2 in the question), I wonder how the inside of these supplies are made. Are there two transformers inside the N+1 supplies (one for the mains, and one for the output), or just 1 (mains). \$\endgroup\$ – TRISAbits Oct 4 '15 at 19:23
  • \$\begingroup\$ I know there's a burning question of why I want to parallel 50 supplies, but the truth is that I don't. I picked a large number to highlight the possible issues of switching PSUs sharing the same GND. The number does not matter. Could the same issues arise with 10? If so, I can update the question. \$\endgroup\$ – TRISAbits Oct 4 '15 at 19:27

You don't need to Isolate ground. Also don't use many power supplies in parallel just use bigger PSU that is capable to handle the load.

| improve this answer | |
  • \$\begingroup\$ I know that using this many power supplies in parallel is not good, but it does provide a good context for my question. \$\endgroup\$ – TRISAbits Oct 4 '15 at 17:26
  • \$\begingroup\$ The thing I'm wondering is that, at 50 PSUs in parallel, wouldn't you get massive GND noise from all the switching? \$\endgroup\$ – TRISAbits Oct 4 '15 at 17:27
  • \$\begingroup\$ In general no. A large power supply would tend to generate more noise than a small one. The ground system needs to be designed to cope with the total power. \$\endgroup\$ – Kevin White Oct 4 '15 at 17:55
  • \$\begingroup\$ I see. If I understand correctly, the noise is proportional to the quality of the GND. The noise could be amortized with a large GND plane across the PSUs, but would begin to increase if it was made smaller or replaced with wires between each PSU. \$\endgroup\$ – TRISAbits Oct 4 '15 at 19:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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