Free software for fft signal analyses. 1

[Russell9990]

Hi

I have just starting out trying to teach myself signal analysis stuff and I am interested to know if people use any free signal analysis software in real life? I have looked at excel fft functions (limited to 4096 data points) but does anybody use

http://www.scipy.org

or dataplot

http://www.itl.nist.gov/div898/software/dataplot/

or gnu plot (only a front end plotting package) or other public domain software to do signal analysis of mechanical systems?

Russell

 

[SomptingGuy]

You could do a lot worse than to download GNU octave. It functions a bit like earlier version of Matlab (v4 mainly) and as such, is well up to the job of signal analysis.

 

[SomptingGuy]

and...

If you're using Linux, octave comes bundled with most (if not all) distributions.

 

[IRstuff]

Most companies, however, have concerns about program validity, customer support and portability of files. This then dictates that commercial software be chosen.

TTFN

 

[SomptingGuy]

Agreed, but if the goal is simply to "teach myself signal analysis stuff", then free is probably ok. For real work, I'd use commercial software.

 

[VisiGoth]

We are seriously looking at using the FFTW3. The issues IRstuff mentioned are exactly our problem. Our alternative is to use the very good routines supplied with the Board Support Packages (BPS). There is a little lag in recoding for new BSP, and the FFTW3 might (or might not) be less work. The FFTW3 is competitive and sometimes better, than some industrial solutions.

 

[GregLocock]

Don't write off the free matlab clones octave and Scilab. Scilab in particular is industrial strength. If you doubt the accuracy of its fft routines then write your own or check their source code. Scilab can read and write .mat files as well.


Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[MikeyP]

The hearts of many freeware and opensource packages like those mentioned above are built on freely available libraries of numerical functions that have been around literally for decades. Even Matlab is based on an ancient set of linear algebra FORTRAN functions called LAPACK.

1) Longevity on this scale implies that it must be OK.
2) These routines are very efficient with both speed and memory because they were designed to be programmed with punched cards and stored on magnetic tape!

M

--
Dr Michael F Platten

 

[tomirvine]

I have posted the source code for an FFT program at:

http://www.vibrationdata.com/fft.cpp

Also refer to:

http://www.vibrationdata.com/tutorials2/spectral.pdf

Whichever FFT program that you use, make sure that you understand the convention, as well as the error sources and limitations of the FFT.

Tom Irvine

http://www.vibrationdata.com

 

[Russell9990]

Hi Tom

Thanks a lot for sharing your code and for the info. I have had a bit of a play with the different Matlab clones and even had a brief look at the r statistical language as that has an fft function. I know that excel has an fft function but that is limited to 4096 samples and I have played around with that a bit. However I was very interested with all the development around the python langue and its use in scientific computing and the attempts to make it more Matlab like. It seems industrial strength to me and there also seems there is quite a bit of support and it can run on many different platforms. I just wondered if anyone was using it “for real”. Here are the links in case anyone wants to have a look.

http://www.scipy.org/

http://matplotlib.sourceforge.net/

http://ipython.scipy.org/

http://www.enthought.com/index.htm

 

[electricpete]

Is the excel limitation 4092 or just a power of 2?

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[Russell9990]

The limitation is 4092 I belive. Something to due with a 4K limit.

 

[SomptingGuy]

Is the excel limitation 4092 (sic) or just a power of 2?

Excel was never meant to be used for general time series analysis (remember, it's a spreadsheet program). However, if you take a stroll through an average engineering anaysis department you'll see many instances of engineers forcing all their analyses through Excel, regardless of its suitability. A classic case of "When all you have is a hammer, everything starts to look like a nail".

I know I mentioned "octave" earlier, but in general, any command-line driven maths program with scripting capability and decent interactive plotting is all you need to learn DSP.

 

[Rob45]

There's another, primarily intended for acoustics, but it works rather nicely, within the CoolEdit freeware package (CoolEdit 96, I believe?).

Might have to look around a bit, but I think you can still find the freeware versions on the internet.