incomplete time serie

[PRovira]

Hello!!

I try to do the spectrum analysis from a incomplete time serie. I have zeros in the "no-data". The "unevenly sampled spectrum", "power spectrum" etc... doesn't work well becouse the nodata-zeros contamine the spectrum.

Then, i try to do a deconvolution of a mask and the contamined serie... but the result isn't the original serie (a test serie, i know the complete serie and a incomplete serie)

Anyone help me???

Thanks regards.

[muks]

Just let me know If my understanding is correct.You want to delete all zeros since they mean adding unwanted noise?

[PRovira]

Yes, i want to do the fourier transform (auto, power, lomb...) and delete all component (frecuencies) due to the zeroes of the time series. Then, i "clean" the serie and i have the original serie's FT(fourier transform).

d=c·m (d dirty serie, c clean serie, m mask serie ) FT(d)=D, FT(c)=C, FT(m)=M.

by a property of fourier transform, convolution:

d=c·m <-> D=C*M ( * is convolution )  --->   D deconv M=C

if the time serie is complex or not obvious, the labview's deconvolution doesn't work well.

the question is, somebody knows some method or algorithm to solve this problem?


thank's

[Guillem_A]

Hi,

If the LabView routines don't fit into your needs try to build your own or adapt it from pre-existing C/Fortran routins using built-in LV functions and the MathScriptnodes. 

You can find a good compilations of numerical routines in the online numeric reciepes.

http://www.nr.com/oldverswitcher.html 

Hope it helps.

Regards,

Guillem.