Filtering noisy signals? which for when

[Jun_l]

Hello, I'm hoping someone can give me an advice on how to filter noisy signals:

I am measuring voltage and current of a capacitive test sample, voltage will be measured over a capacitive divider and current over a resistance. Picture phaseshunt.png shows the signals over the resistance without my test sample. Looks good, no noise. Picture noise.png are my signals when I add my test sample in the test setup, my voltage signal looks good but not my current, which is causing me possible errors in my calculations for phase.I'm not sure if the signal is noisy because of the low currents (ca. µA) or because of other factors?

I know I have to filter it, but which filter should I use? I want to measure voltage and current over a function of frequency ( 50 Hz - 2 kHz), if I use a lowpass does it mean I have to change the highest and lower frequency everytime I change my input frequency?

I have no experience with filters so i'm going to do some reading too but some tips would be highly appreciated!

Vi is attached. I'm sampling with a rate of 200 kHz and 100k samples, using NI USB 6366.

Jun

[Ben]

The range of your current axis is 20X what it is when you say "no noise".

First off change the scale setting to see if you have the same amount of noise with no widget in your tester.

You can decrease bit noise if you set the range of the acquisition to a smaller range (increase the gain of the DAQ input).

"Noise Reduction Techniques in Electronic Systems" by Ott is an excellent reference that looks at all of the types of noise we can encounter starting with Maxwell's equations, establishes boundry conditions, does the math and develops rules of thumb for each type of noise.

In that book he indicates that removing noise in a signal is a last resort. The first steps should be to;

Eliminate noise source

Eliminate noise propagation

Shield noise entry in the system

and only then try to remove the noise that did get in.

That said...

What ever type of filtering you use, make sure the same type of filtering settings and ranges are applied to all of the signals being compared. All filters (aside from zero phase shift filters) will introduce a phase shift (Depending on the order of the filter).

As to which type?

Play with the filter examples that come with LabVIEW. They each have some advantages and disadvantages. Find the one that best meets your needs. But before you settle on one of the fancy math versions, over-sampling and then doing a box-car averaging is a good approach for reducing random noise.

Have fun,

Ben