I recently bought several SMPS buck converter supplies for a project. They all had about the same specs, typically 90-270VAC input, 12VDC output. Having an expected worst case continuous load of about 200mA, I selected devices with output ratings of between 300mA and 500mA. Also, since my circuit employs a fairly large capacitor (2700uF), I was careful to select units claiming to be short-circuit tolerant, specifically stating "output returns to normal when short condition stops".

Well the results were a really dismal eye-opener except for one company! Surprisingly, the best device I tested was made by a China based company (SANMIC) and was only rated 300mA. Yet it took on the load, capacitor and all, like a champ. Unfortunately that was NOT the case for at least 4 other boards made by other companies with similar specs. These would basically oscillate slowly (maybe 5hZ) attempting to drive the load, but would never actually start. The diagram and graph I'm posting is admittedly a crappy image made with a paint program, but it fairly accurately depicts what I'm seeing in all but one of these supplies.

So other than the obvious answer of "buy better supplies", I'm hopping to understand (1) Why its so hard for these supplies to "bring up" a capacitive load to rated voltage, and (2) what key features should I look for to get a better (and RELIABLY better!) result. Obviously "short circuit tolerant" has proven to often be false. I can't simply choose an arbitrary higher current rating, since a capacitor at time=0 will always seem to demand infinite current. Obviously some companies know how to do it, but I'd never be able to tell them apart by the specs!

enter image description here

By the way, its irrelevant to the question, but the reason for the capacitor is because an MCU, upon detecting the power "going away", needs to do several things including opening relays in an specific sequence and saving all its program states to NV ram. The 2700 uF gives me about a second to do that.

  • 1
    \$\begingroup\$ An alternative would be to charge the 2700 uF slowly through a resistor. This would limit the inrush current. Some diodes would be needed to connect the output of the regulator and capacitor together. \$\endgroup\$
    – EE_socal
    Aug 1, 2018 at 21:14
  • \$\begingroup\$ It is a possible solution. Of course it would be MORE possible if it weren't one of those cases where the PCB has already been cut, but that's not your fault! ;-) \$\endgroup\$
    – Randy
    Aug 1, 2018 at 23:36

2 Answers 2


You've identified the problem fairly accurately, for capacitative load the instantaneous current requirements are quite large.

In many designs short circuit detection simply disconnects a load when a short is detected reconnecting it when the short is gone . In this case your capacitor does not ever fully charge before the short circuit detection triggers, causing the saw tooth waveform you observe.

SMPS Vendors often include a variety of features, namely

  • soft-start
  • inrush limiting (snubber)
  • short circuit protection.

Sometimes these features may interact with each other.

Soft-start and inrush limiting are intended to mitigate the startup characteristics of a capacitative load.

Short circuit detection protects the supply due to accidental short and typically attempts to restart the supply as soon as the short is removed. But it has negative performance implications during a large inrush on start.

For example, a larger than expected inrush may trigger the short circuit detection, which turns the supply off. As the current load drops the short circuit detect unlatches ... restarting the supply to the same inrush. Rinse and repeat to give you the sawtooth waveform you observe.

General Suggestions

  1. Identify the in-rush current specifications of the supply, these are typically 10x more nominally than the rated supply current. Understand how this relates to your your measured in-rush load.
  2. Identify the nature of the soft-start circuits and inrush limiting on the "working" supplies, the slow supply startup may compensate for larger than expected in-rush.
  3. Use external soft-start or in-rush snubber to avoid slamming the full capacitance load on startup and be agnostic to the features of your upstream SMPS. In a pinch this can simply be a small value series resistor to limit the inrush current, say 100Oohm in series with the capacitor, which is disconnected from the circuit when it is charged.

Note: It is hard to say if the worse performing supply is actually a "worse" product, in principle you are simply exceeding its specification, if the supplies are nominally the same specification and you are exceeding both of them, it is probable that the one that allows the large in-rush will function but fail prematurely. Or it may have a slower soft-start and does not trigger the short circuit detection, meaning it is a better design for your purposes.

  • \$\begingroup\$ Thanks. Very informative and helpful. Two observations, maybe relevant... 1) A great deal of the boards I tried were "no name" boards, the ones that actually worked without fail were the ones who put their co. name on the silkscreen! Hmmm! 2) I see what you mean about calling a supply "worse". But a great many wall-plug adapter supplies also seem to have similar circuits, and though cheap they tend to be very tolerant of hanging a big cap on their output. Maybe they are made anticipating supplying unknown products, which may also have big caps hiding inside. \$\endgroup\$
    – Randy
    Aug 1, 2018 at 22:51
  • \$\begingroup\$ One other thing, based on your suggestions I thought I'd investigate just how much resistance in series with my test circuit would allow some of these "poor performers" to work. The worst case was 10 ohms, the best was about 4 ohms (and that was marginal) and the "best" of course was the one that needed none. Its unfortunate that I'm controlling some 12V relays where 10V is not perfectly reliable for coil pickup, which might be over-ridden to "ON" by a mechanical switch (so the voltage is needed right away"). But at least I can use some of my "duds" now in less critical circuits! \$\endgroup\$
    – Randy
    Aug 1, 2018 at 23:26

The ones with adjustable current sensing limits will be more inherently stable.

After all a voltage source into a cap with near zero ESR relative to Vout/Iout=Zout min.

There are many methods of switched mode power regulation and low impedance voltage feedback only is sure to cause stability issues. They must combine current feedback in linear mode with Voltage feedback to be potentially stable. To simply have OCP is not enough.


Your Answer

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

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