Data filtering or smoothing?

[Willsong]

Hi all,

Does anybody know about data smoothing or filtering?
I'm working on data processing. The experimental data has a lot of noise on it so I need to remove the data.
I'm studying about filtering with FFT and trying to program a subroutine in Fotran. It is extremely tough to understand and program it.
So if anybody has an experience with programing a filtering, please help me out.

Thank you very much in advance,

Will

 

[IRstuff]

see:

http://www.nrbook.com/a/bookf90pdf.php

TTFN

FAQ731-376

 

[Willsong]

Thank you IRstuff.
If you don't mind, could you talk to me specifically?
Actually, I don't know what and how to use.
I have spent almost a week on this. I don't have any progress. Please help me out.

Thanks again.

 

[IRstuff]

Is this for school?

TTFN

FAQ731-376

 

[Willsong]

No, it is not at all.
I'm building my own program to process experiment data.
It is pressure data having a lot of noise. So I want to remove the noise.
I'm studying FFT digital filtering over and over, but it is difficult to understand. Also, I don't know whether the digital filtering is the best way to remove noise or not.
If you can help me out, I really appreciate it.
Please help me.

Thank you.

 

[IRstuff]

> Perhaps you should start with a better pressure sensor
> How do you know it's really noise?
> What is the actual behavior of the data, i.e., constant, varying, etc?
> FFT filter is not that simple because most data is not periodic, so windowing effects can alter your data
> The type of filtering depends on what your data looks like

TTFN

FAQ731-376

 

[Willsong]

Thank you for your comments.
The pressure data was measured in a combustion engine cylinder.
It is change with time and crank angle. The data shows the pressure trend, but it's noisy. So I want to remove the noise. Otherwise, I get unreasonable calculation result based on the measured pressure data.
If you have better idea, could you recommend me what to do?
I thank you very much for your time.

 

[HuVeit]

You should build in the butterworth filter (24dB) not FFT-filter.

 

[IRstuff]

Ostensibly, a Butterworth filter has a Fourier transform.

> How many samples do you have relative to the features of interest, or equivalently, how many samples do you have along this "trend?" Fewer samples is problematic; more samples allows more flexibility in determining the optimum course of action.

> How much distortion are you willing to tolerate?

TTFN

FAQ731-376