Go Down

### Topic: Low pass filter -- how hard can it be? (Read 8128 times)previous topic - next topic

#### errno

##### Aug 24, 2012, 10:14 am
Hello all,

I'm trying to remove noise from a SMPS 5V supply. You can see a 'scope view of the noise (two 'groups' of noise present) in the two attached images. The built in low-pass filter on the 'scope can dramatically smoothen the noise when set with 25Mhz as the cutoff frequency, and I'm trying to emulate this filtering with simple components on a board (see attached breadboard image which features DC power input, 100uF decoupling capacitor, and RF lowpass filter from 220ohm resistor and 68pF capacitor -> cutoff of 10.6Mhz).

So far, I have not been able to significantly attenuate the noise at all. I've tried low and high cutoff frequencies with simple 1-pole RC (many different resistor/cap combos) and 1-pole LC filters (I only have a 100uH coil, but many different caps). I've also tried many many different power rail decoupling capacitors, ranging from a few pFs to bulkier electrolytes.

Trivial circuit diagrams:
- RC: http://en.wikipedia.org/wiki/File:RC_Series_Filter_(with_V%26I_Labels).svg
- LC: http://www.ampbooks.com/home/amplifier-calculators/LC-ripple-filter/LC-lowpass.gif

Nothing is shaking the noise.

Any ideas what I'm doing wrong, or ideas of other things to try?

Thanks.

#### Tom Carpenter

#1
##### Aug 24, 2012, 11:41 am
If you want to remove noise from a power supply rail, you would be bettern off using a series inductor followed by a capacitor to ground.
If you use a series resistor in a power supply rail, as soon as you start to draw current from it, the resistor will start dissipating large amounts of the voltage. take your 220R resistor for example - at 10mA, it would drop nearly half the +5v supply (2.2v). An inductor by comparison has negligible resistance (generally) and will do a better job at filtering.
I have used a cheap 220uH inductor and a 10uF capacitor to filter the power supply for an audio DAC with great success. This combination would give you a cutoff of:

1/(2*pi*sqrt(L*C)) = 3393Hz approximately.

There are two things you may have also failed to account for. First, you cant just use any old capacitor to make an RC filter at 10MHz. Cheap capacitors tend to have a tollerance of +-20%, sometimes as high as 80%, which means you can't gaurantee the capacitance value and thus your cutoff point.
The second thing is 'cutoff' is a bit deceptive. What it means is the frequency at which the signal drops to half the magnitude (aka the 3dB point). It is impossible to get a brick wall filter as it is known (unless your signal can time travel), and so you have to allow for this in your filter. For simple filters like the one you have made has not been carefully designed to have as sharp a cutoff as possible.

For DC power supplies you aren't just trying to cutoff the high freqency component, as the noise is not just at the 15MHz your scope estimate, but across large parts of the specutrum including low frequencies. For a power supply, you want to remove as much noise as possible and only keep the DC component. This means you want your cutoff as close to 0 as you can, without attenuating the DC voltage.

~Tom~

#### errno

#2
##### Aug 24, 2012, 01:09 pm
Appreciate the response Tom!

Resistor vs Inductor -- you're absolutely right. In my finished project, I plan to use inductor+capacitor for the reasons you describe, but have been playing with RC filters since I already have many different values of 'R' and 'C' (and currently only 1 value of 'L' -- 100uH). So if I can it working with RC filters at all, I'll proceed to LC.

Cheap capacitors:
- I am certainly using cheapo ebay capacitors. Is there a specific type/material of capacitor most suited to this purpose?

Cutoff:
- I've experimented with many different cutoff values (from 10kHz - 20Mhz) and haven't seen any dent in the noise shown in the images. Are there any drawbacks to shooting low (e.g. <=10kHz), and can I safely assume a lowpass filter with a 10kHz cutoff point will also effectively attenuate a high frequency (e.g. 25Mhz) signal? (e.g. since 10kHz <<< 25MHz, one could expect 25Mhz noise in the signal to be attenuated to ~nothingness?).

Thanks again.

#### pito

#3
##### Aug 24, 2012, 04:33 pm
To reduce those noise you have to use a ferrite bead in series and ceramic capacitor (ie 10uF || 10nF) to the ground. The 100-220uH inductor will not help you much.
For further reading see ie. www.linear.com/docs/11877

#### MarkT

#4
##### Aug 25, 2012, 12:54 am
Yes definitely ceramic capacitors - you need capacitors with a low impedance at the frequencies of interest, no electrolytic can cut it, ceramic caps at 10uF and 0.1uF are a good start - small series indurctor (it needs a high impedance at the frequencies of interest).

The problem is that capacitors have stray inductance and that inductors have stray capacitance.  A key spec to look at is the self-resonant frequency -- your decoupling caps and inductors should have self-resonant freq ABOVE the frequencies of interest - otherwise they behave completely the wrong way round (and act as high-pass filter!!).  Since wiring has inductance its vitally important to use good layout to the capacitor(s), otherwise the noise can simply bypass it.

Also be aware that scope-probes can pick up noise that's not actually in your circuit - its picking up capacitively and/or inductive coupled noise directly from the power supply, so make sure to measure the noise at the load, away from the supply.
[ I DO NOT respond to personal messages, I WILL delete them unread, use the forum please ]

#### Docedison

#5
##### Aug 25, 2012, 02:37 am
There is also 'common mode' noise which is just the ground differential voltage.... ground is Not necessarily Ground when it comes to pulses with that rise time... Especially if you have a computer connected to the USB port. Real noise will be diminished by the filter common mode noise will not, It's not flowing through the filter...

Doc
--> WA7EMS <--
"The solution of every problem is another problem." -Johann Wolfgang von Goethe
I do answer technical questions PM'd to me with whatever is in my clipboard

Go Up