I have a 48V 50A SMPS driving a fairly erratic load. The load varies wildly between 1 and 10A and can suddenly drop from full load to low load. I don't have an oscilloscope at hand, so these are estimates.

The load is very tolerant and won't care much about voltage variations in either direction. The problem is that the over-voltage protection on the power supply is fairly aggressively tuned at around 54…56V. It trips pretty often. I believe this is because the output regulation can't quite keep up with the sudden drops in the load.

What can I do to remedy this issue? I have considered the following:

  • Add a resistive dummy load in parallel to help the regulation in case it struggles at low loads. How should I size it? 10A? 480W is a heck of a lot of power to dissipate in a resistor.
  • Add a capacitor across the power supply output. I have no idea what size it should be to have an effect, but my guess is huge.
  • Add a TVS diode across the power supply output. This was suggested here and indeed it seems ideal for filtering inductive spikes. But I'm not sure how well it plays in this case. Conveniently, there is a 51V part available though.

Is my diagnosis correct? What would you suggest to solve the problem?


On a transient load dump all of the energy already stored in the SMPS's inductor will have to transfer to the output capacitor. There's nothing the control loop can do to prevent that. So the output capacitor has to be sized to absorb that energy before the OVP trips.

Aside from that, a poorly designed control loop can cause excessive overshoot. If that's the case re-tuning the supply or switching to a better supply can help.

I don't know if you can modify the SMPS at all, but your options are as follows:

  1. Ensure the SMPS control loop is optimized. If so, add more output capacitance or reduce the SMPS's inductor value if possible.

  2. Build a comparator circuit that turns on and clamps the output voltage to a given value before it hits OVP.

  3. Use a TVS to do the same this as in (2) (both have to be sized for the energy involved) Be careful with the breakdown voltage variation on the TVS, it can be large.

You can estimate the energy you have to absorb by looking at the peak overshoot voltage you expect to get. Then you will have to reduce the 1/2CV^2 energy in the caps by the amount of voltage you need to get below the OVP threshold. By the time the output voltage peaks, most of the 1/2LI^2 energy in the inductor is already in the caps.

That should give you a starting point for the amount of energy you have to handle, then you can add some margin from there.

The easiest thing to do might be to just add more output capacitance if you have space.

  • \$\begingroup\$ Sadly the power supply is pretty much an off-the-shelf unit so I don't know much about its internals and I can't modify it. The output capacitors are located in a cramped space so I can't read the markings on them. The voltage peaks probably reach a maximum of 10 volts over nominal voltage. Is there any way to estimate the required capacitance to an order of magnitude or so? I was also concerned about the ratings on the TVS. The breakdown voltage is between 48.5 and 53.6 volts on the 1.5KE51A. Should I just buy a few and hope one of them doesn't blow up? \$\endgroup\$ – ntoskrnl Aug 26 '14 at 8:24
  • \$\begingroup\$ As an experiment you could add some capacitance until you observe a reduction in the overshoot. Then the capacitor you added will have absorbed 1/2C(deltaV)^2 joules, and you can either keep or add the appropriate amount of capacitance to further reduce the overshoot or size the TVS to absorb the energy. \$\endgroup\$ – John D Aug 26 '14 at 15:03
  • \$\begingroup\$ I don't have any suitable capacitors at hand though, so I'll have to buy anything I want to test. What do you reckon will happen if I slap, say, a 10mF cap on it to avoid multiple trips to the local electronics shop? \$\endgroup\$ – ntoskrnl Aug 26 '14 at 16:56
  • \$\begingroup\$ 10 millifarad or do you mean 10 microfarad? \$\endgroup\$ – John D Aug 27 '14 at 3:06
  • \$\begingroup\$ Millifarad. That would be a huge cap, and it should be enough without requiring further fiddling I guess. Or can it be too large? \$\endgroup\$ – ntoskrnl Aug 27 '14 at 7:06

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.