Reducing input noise from pulsed current from Buck Reg

Thread Starter


Joined Mar 8, 2016

I have a circuit which is essentially is a 12V to 8V Buck regulator that is providing power to an LED which is turned on and off
fast using a voltage controlled current source. See picture attached. The current pulses are 5A with a duty cycle of 20%.

I find in operation at full current (5A), the ceramic caps start to sing due to Piezo effect. Most probably because the frequency of this pulsed load is hitting the acoustic frequency band. I can mitigate this on my board by the usual methods e.g. changing the caps to metal standoff ones or to Tants. This is fine to deal with the audible noise on my board but ceramic caps further back in the chain (in the benchtop supply) are still singing. I know that I could put a two stage LC filter on the input to the regulator to try and reduce the noise going back to the supply (I already have a common mode filter on the input supply to my board) but my question is this:

1) Is there anyway I could reduce the noise on the input to the regulator to stop the ceramics singing? If I put an LC filter after the input decoupling caps I would starve my regulator. If I put it before the decoupling then I likely think it would stop noise following back of my board but not stop the noise on the decoupling caps.

2) There isn't a lot of noise on the output of the buck regulator as is typical. Would it be worth trying to filter this more?

Your advice would be gratefully received




Joined Feb 8, 2018
If you are getting audible noise from a 1 MHz converter, either it is unstable or the noise is due to the pulsed loading on the output. If the former, the solution is to fix the converter. If the latter, the easiest solution is probably bulk capacitance at the output of the switcher. Very high switching frequency is typically used with the objective of miniaturization because small capacitors can be used. Small capacitors do nothing to mitigate the problems due a highly dynamic load. A high frequency switcher can respond rapidly to a dynamic load, but that means the effects are passed to the input supply.

Things like common mode filters won't help because the current is normal (differential) mode. CM filters are for the RF spectrum. The only way to keep the ceramics from singing is to provide additional/alternative capacitance to take up some of the current demand. This is difficult because the very low equivalent series resistance (ESR) of the ceramics means that they will tend to be the path for most of the high frequency energy. Providing effective alternative energy storage in the audio spectrum is somewhat less of a challenge. If the voltages are within the range that makes them suitable, polymer tantalum appear to be the best choice, though I think they are quite expensive. Polymer aluminum types are probably the next best overall choice where fairly large capacitance with reasonably low ESR is required.

You cannot remove the ceramic output capacitors of a switcher and replace them with something with higher ESR. This would almost certainly put an "ESR zero" (capacitor looks like a resistor above a certain frequency) in the output filter that is not compensated in the control loop. Adding capacitance in parallel rarely causes problems.

Thread Starter


Joined Mar 8, 2016
Thanks for your reply ebp. I very much appreciate your advice

I am pretty sure my device is not unstable. I use the TI web bench for checking all the parameters out.
So I am sure it is the pulsed load.

Re doing the design with an SP-CAP (Polymer Alu Ely) with ESR 40mohms means the design is still very stable with
about 200uF on input and output.

Is it worth putting an acoustically damped ceramic on as well to bring the ESR down further? Probably more worth putting more SP-caps on to bring the ESR down.

If there any worth in adding a differential mode filter on the input to the regulator as well (before the decoupling caps)?

Any advice is very welcome



Joined Jun 19, 2012
I have also experienced this problem.

it's the Delta-V that makes the caps expand and contract mechanically.
I suspect it's the input caps doing the singing, unless the regulator is unstable. (as previously mentioned)

Check the delta-V on both input and output with a scope, if it's high on the input side, you may have luck with a big fat electrolytic pre-filter cap.