# How do I attenuate 30kHz noise?

I'm trying to design an LC low pass filter to attenuate 30-35kHz noise on a 3.3V power rail. It's preferable to attenuate higher frequencies as well. So I designed an LC filter with L=10u, C=1u (which is wrong, but this is just an example) - and I found it boosted frequencies near the cut-off point. Is there any way to avoid this? The boost of 22dB corresponded to an increase in 12.5x the voltage. The rail will be powering a microcontroller, which draws 100mA, so this increase in voltage would likely damage it.

-
The 3.3 V is a signal or a power rail? What is it used for? What are your source and load impedances? Can you draw your circuit? You can enter it in falstad.com/circuit and then export link from the file menu – endolith Feb 23 '11 at 19:02
@endolith, Yes, it's a power rail. As the micro could execute various code and enter sleep modes, power consumption may vary - is it still possible to use an LC filter? – Thomas O Feb 23 '11 at 19:08
Cascaded RC filters are simpler, but will waste energy in the resistors, so it's only good for signals where the energy is not important. For a power rail, you don't want to waste energy (voltage drop), so either use small value Rs, or use LC filter. But LC filters can resonate, which produces the peak you're seeing. – endolith Feb 23 '11 at 20:25
if your having 30khz noise on a power rail you should consider the cause of that noise. It shouldn't be there and you should solve the problem at the source of the noise rather than bandaiding it with a filter. – Mark Feb 23 '11 at 20:34
@Mark, it's mainly to attenuate noise from a noisy battery line which also powers a motor, so lots of current is drawn and this causes glitches in the power line. But I'm not certain I will still need it after doing a scope of the power line. – Thomas O Feb 24 '11 at 9:47

In simple terms, add a resistor in series with the inductor. But then you could rather use a RC filter because you're decreasing the quality of your inductor.

Why do you prefer and LC filter? The filters are harder because at a certain frequency, you will get a resonance point. The voltage you put on the filter will be 'amplified'. The amplitude of how much it will shoot up , is called the Q factor. To dampen the Q , you add a series resistance to the coil to dampen it out.

A lot of math stored on: http://en.wikipedia.org/wiki/RLC_circuit

But because you're only operating at 30kHz, I think a RC filter might do the job better/easier.

Edit: Sorry your post was confusing because first you were talking about a 3.3V >signal< and then about a power line.

The answer still stays similar, add a small resistor, but it will reduce the steepness of the filter. Alternatively you can better add a filter which it's cutoffpoint is way less than your noise. I.e. if you have 30kHz noise, add a filter with a cutoff point at 3kHz or something. You won't get the increase in amplitude at the noisy frequency and a pretty good attenuation.

However, as mentioned, patching the source is always better.

-
1. Put linear regulator before uC - they will react fast enough to cut low-freq noise. Especially good if you have some margin between power supply voltage & uC voltage.

2. Just add large(some 10-1000 uF) caps before uC: some electrolytic and some 1uF and 0.01uF ceramic ones all in parallel to prevent resonance. Ceramic ones should be as close as possible to power pins of uC. If it were more powerful uC, I would consider adding ferrite bead, but it might be not needed in this case.

-

If you are using LC filter for any voltage (power supply). Don't select the cutoff frequency of filter has same as frequency you want to attenuate. Generally , you should select 10 times less than the frequency you want to reject. Make sure that Inductor SRF is away from the frequency. Inductor current rating should be 2 times of the max current comsumption of load and voltage rating for capacitor should be greater than 3 times of the selected voltage.

-
Exactly. The cutoff frequency should be around 1Hz. Suggested values: L=20uH, C=10,000uF. – EJP Jan 29 at 3:06