Filtering the high frequency noise in switching PSU


It is well known that it is difficult to filter the noise generated by switching power supplies, and this difficulty increases as the PSU is supposed to deliver large currents.

For example, in a usual LC filter, the self may reach the saturation if a large current is passing through.

I would like to know what techniques are usually used in large current capable switching PSU to filter the (high frequency) output ripple.

To restrict the question:

  • I'm not interested in the 50Hz low frequency noise,

  • not interested in the EMI power line filtering,

  • not interested in proper use of ground planes, separation, minimizing area in current loops, not breaking current return paths, identifying high current flow paths and keeping them short and away from noise sensitive parts of the circuit and like.

  • I'm interested only in the output filter.

You still need to focus your question though.

This basic circuit is more or less what I had in mind: switching

A part of the original question has already been answered. So, to focus on the remaining points, my question is now:

What can be done to reduce the common mode noise?

Why should this post be closed?


Are you talking about common mode noise that simultaneously affects both pos and neg output on the DC side (due to capacitive coupling in the transformer) or just differential noise i.e. conventional ripple artifacts? Both are treated differently and both these noise sources depend on different things in a design. Are we talking about DC-to-DC isolating converters or just plain ordinary buck/boost regulators? ‭Andy aka‭ 18 days ago

Can you adjust your question to more fully reflect your aim? ‭Andy aka‭ 18 days ago

I think you misunderstand it from the view of someone answering. Someone answering can make an answer aimed at a specific design and provide good information. Then, if I were answering, the way my brain seems to work is that I could sit back and think how this might be applied to similar (but not identical) designs. If you expect someone to brainstorm the full gambit of possible scenarios prior to making an answer then your expectations are too high. ‭Andy aka‭ 17 days ago

You still need to focus your question though. Asking for a schematic is also problematic because it just may not be the topology that is most appropriate to you. If I answered and you then said "yes but what about this or that converter" I'd have been wasting my time on the original answer. You have to be reasonable. ‭Andy aka‭ 17 days ago

@Andy. See my edit. ‭coquelicot‭ 17 days ago

Show 8 more comments

1 answer


A switching power supply already is inherently a filter. Current is coming thru an inductor, followed by a capacitor to ground.

If you are asking what additional is usually done to filter out changes in the output voltage (not common mode, that's a totally different issue), then the typical answer is "nothing".

It is expected that switching power supplies will have noise on the output at the switching frequency. Any decent power supply will have a spec that says what the worst case noise is. If you can't live with that level of noise, you get a different power supply.

Two ways that the switching noise can be reduced is by increasing the switching frequency and increasing the output capacitance. However, neither of these is an after-the-fact filter like you seem to be asking about. Both these are integral parts of the overall design.

When quietness of the output really matters, but you still want most of the efficiency of a switcher, then a linear post-regulator can be a solution. The switcher output is controlled to a little more than the final desired output, then a linear regulator is used to make final less noisy output. Note that this decreases overall efficiency due to the linear regulator, and it adds complexity. This is therefore only done when you really can't tolerate even a few 10s of mV of ripple. Most of the time you can, which is why this method is far less common than "bare" switchers.

So, your answer is nothing, or at most a linear post-regulator. I thought there is more than that

Maybe in some cases, but usually the switching frequency, inductance, output capacitance, and control scheme are chosen together to get the output noise down to what it needs to be. For example, all else equal, a larger output capacitance results in lower switching ripple. However, this also has effect on stability, and the controller needs to be tweaked for optimum performance.

Put another way, all the filtering you want has already been designed in. If you want more filtering, then choose a different design.

What can be done to reduce the common mode noise?

Argh! This is why it's important to include relevant information up front. As I stated above, common mode noise is a totally different issue than what I answered for. I suppose some of this is my own fault for answering a half-baked question instead of just closing it.

Common mode noise is mostly about capacitive coupling between the input and output, at least for an isolated supply. But, I've already spent enough effort on this question. I don't feel like getting into this whole different subject now, with the distinct possibility of getting jerked around again.


So, your answer is nothing, or at most a linear post-regulator. I thought there is more than that but OK. ‭coquelicot‭ 18 days ago

@Olin. You are completely wrong, your answer is a legitimate answer to my question, and I'm sad you feel like this. I've edited my question to focus on that particular point because Andy asked me to do that, and because it is not dealt in your answer (that's the ironic point). I was vaguely aware that there is something like common and differential noise, but I thought noise could be processed as a whole. I was in the darkness, that's all. ‭coquelicot‭ 17 days ago

@Olin. Please, consider deleting your last edit, or delete the whole post if you wish: This degrades unfairly your answer that was perfectly valid in my point of view. ‭coquelicot‭ 17 days ago

I said this: Can you adjust your question to more fully reflect your aim? and I didn't imply it should be common mode or differential but, what you had in mind. I just want to restate this. ‭Andy aka‭ 17 days ago

@Andy, as I said, I was in the darkness. I felt that at some point, when I said "let's leave that". I've edited the 3 last lines of my question to be fair with the answer of Olin. Hope this is more acceptable. ‭coquelicot‭ 17 days ago

Sign up to answer this question »