SMPS without any audible noise. Have you any experience or guides?

[Miyuki]

Hi folks.
I need power supplies without any audible noise. Preferring extended hearing range up to 28khz
I wonder are some of the shelf power supplies designed completely noiseless over all load range? 0-100% small and medium sized ones 10-100W
I did not find those readily available.

Do you have any experience with some designs without noise?
Price is not a concern. But 50/60Hz hum is an issue. And I did not find the mains frequency transformer without one.
So I might end up building them.

[Boscoe]

I think this can be solved with some good filtering on the output of the SMPS. Keep the noise in the sub 1kHz band and the amplifier’s (assuming it’s a good one) PSRR will take care of the rejection.

[T3sl4co1l]

Okay cool you want a quiet supply.  How quiet, then?  Can you put a number on it?

Taking your request at face value means < 0 dBa, the very definition of "audible": right on the edge of perception.  And given the frequency extension, that might be flat, not A-weighted.

That's really quiet, like, that rules out most capacitors as well as inductors, because they buzz just perceptibly at mains harmonics.  It might need acoustic shielding to meet that (e.g. potting the supply, then suspending that inside a larger metal box with foam inbetween -- some means of heat transfer is still needed of course).

Easiest way to eliminate light load whine, put a dummy load on the output.  Some synchronous rectifier architectures can run continuous wave as well, but you'd have to hack it to do that, or make your own.

Tim

[jonpaul]

Modern desings use very high freq and soft switching.

F 50k...20 MHz.

Any remaining noise is due to poor loop control and sub cycle oscillation.

I have not had audible SMPS noise in decades.

Jon

[Miyuki]

I know most issues with really loud supplies are from the poor loop and of course standby modes

Forced conduction output rectification looks like a simple solution to keep the steady duty
And dummy load is an option
Digital controller should be easy to use here

With the possibility of potting I count

//Edit: Hearing threshold, might be even lower than this. But it is a good guide. I want to be safe from a minimal value in this one.

[TimNJ]

Some of our medical customers request something like 25dBA @ 1 meter, or higher at a closer distance.

Modern controllers with low-power/burst-mode, operating between ~5-20% load might be the biggest culprits. Often, their burst mode period is directly in the audible range. About 1KHz tends to be the frequency most humans are most sensitive to, so often you can try to push the frequency higher or lower to avoid that range.

I think the hard switched topologies are usually the loudest. Resonant topology is likely quietest. High peak flux densities probably the loudest.

We’ve had some success by varying the varnishing procedure for the magnetics. Also had some success with a slightly softer epoxy for the center leg (in the gap). You can also try to apply some dots of epoxy on the external mating line of the two core halves to join them together more securely.

Finally, be mindful of resonances of other components on the board. Heatsinks tend to be lightweight and if not properly secured, they can resonate when rigidly coupled to a transformer/inductor through the PCB.

[jonpaul]

Hello: Burst mode controllers are easily avoided in noise sensitive applications.

Noise sources can be ferrite cores, steel laminated mains freq transformers, inductors and film capacitors.

Impregnation or encapsulation will reduce but not eliminate a noiuse due to poor design or poor component choice.

Noise supplies are in general due to bad design.

Jon

[Doctorandus_P]

I've bought a power supply in "laptop brick" format which does 24V 8A and I found it quite noisy.
I hung it up under my desk, so there is no mechanical coupling, which helped a bit, but not enough. Then I took a piece of 100mm PVC pipe, glued a piece of carpet on the inside and put that over the power supply. The PVC pipe is open both both top and bottom so the power supply still gets it's normal cooling. I have not heard it since.

At some point you have to consider both using parts that generate less noise, and methods that reduce propagation of the noise that is being made.

[David Hess]

I cannot say that audible power supply noise is something I have studied in detail except when it was audible.  Causes include:

1. Low switching frequency - obviously keep this above 28 kHz in your case but this is hardly difficult these days.  In the past this was more difficult.
2. Avoid subharmonic oscillation - current mode control will require slope compensation in continuous conduction mode.
3. Constant off-time controllers can produce a noisy hiss, but I have forgotten the cause and how to prevent it.
4. I am not sure that there is a difference between continuous conduction and discontinuous conduction, except that the former requires slope compensation to prevent subharmonic oscillation if current-mode control is used.

If you are relying on someone else's product, then I wish you good luck.