Date: Thu Nov 07 2013 - 13:12:55 MST
Not sure how you are judging a "good result"
Attached are 2 png files.
They low and high pass response functions for differing weights
with 'fca=1.0/365.0'
You coulalso experiment with an fft ([1] detrend. [2] taper pior
to using the FFT [ezfftf ... then set all coefficients .le. 1/365
to 0.0 ... then use ezfftb
This has problems [eg: Gobbs phenonema and sidelobes]
See: http://www.ncl.ucar.edu/Applications/filter.shtml
On 11/5/13 3:28 AM, Erika Folova wrote:
> I am experimenting to remove the variability greater than 1 year in my
> daily dataset. For that, I did a high-pass filter as follow:
>
> * nwt = 251*
> fca = 1./365.
> ihp = 1
> nsigma = 1.
> wgt = filwgts_lanczos (nwt, ihp, fca, -999., nsigma)
>
> however, I still do not get a good result from the filter, any suggestion
> of how many number of weights that's normally used to remove variability
> greater than 1 year in daily dataset? if you have experiences in signal
> process please kindly share it.
>
> Thanks
> Erika
>
>
>
> _______________________________________________
> ncl-talk mailing list
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>
_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
