On Feb 2, 1:16 am, Wox <nom...@[EMAIL PROTECTED]
> wrote:
> On Fri, 1 Feb 2008 07:16:26 -0700, David Fanning <n...@[EMAIL PROTECTED]
>
> wrote:
>
>
>
> >Wox writes:
>
> >> Example below filters in time or frequency domain:
>
> >> ; Time domain
> >> freq1=2.
> >> freq2=3.
> >> freq3=4.
> >> dtime=0.1
> >> ntime=1000
>
> >> time=dtime*findgen(ntime)
> >>
signal=sin(2*!pi*freq1*time)+sin(2*!pi*freq2*time)+sin(2*!pi*freq3*time)
>
> >> ; Time domain Filter
> >> f_low = 0
> >> f_high = 2.5
> >> timefilter = DIGITAL_FILTER(f_low*2*dtime, f_high*2*dtime, 50.,40)
> >> signal=convol(signal,timefilter)
>
> >> ; Frequency domain
> >> nfreq=ntime/2+1
>
> >> freq=findgen(nfreq)/(dtime*ntime)
> >> fsignal=fft(signal)
>
> >> ; Frequency domain filter (instead of time domain filter)
> >> if n_elements(timefilter) eq 0 then begin
> >> steep=20.
> >> freqfilter= 1./(1.+(freq/f_high)^steep)
> >> fsignal*=freqfilter
> >> endif
>
> >>
plot,freq,abs(fsignal[0:nfreq-1])^2,xtitle='frequency',ytitle='spectrum'
>
> >Wonderful example, but I'm trying to understand this whole
> >subject. Do you think you could flush this out with a little
> >explanation of what you are doing and why you choose the terms
> >you use, etc.? What kind of frequency filter are you constructing
> >here? I don't necessarily see it doing any filtering of the signal,
> >at least if I pass it the original signal, rather than the signal
> >that had already been filtered in the time domain, as written
> >in your example.
>
> >Cheers,
>
> >Confused
>
> Ok, sorry for the confusion, but I was just illustrating that you can
> do the same filtering in the frequency domain as in the time domain.
> You do one or the other, not both at the same time. Btw, convolution
> in one domain becomes multiplication in the other:
>
> filtered = signal "convol" filter
> fft(filtered) = fft(signal) x fft(filter)
>
> But I guess you already knew all this.
>
> The filter used is the Kaiser-Bessel filter. At least I think
> digital_filter is using this. For the filter I constructed in the
> fourier domain, I'm not quit sure whether it is really identical to
> the KB filter, but if you plot it, it looks like a nice lowpass filter
> to me :-).
>
> I'm just typing this in a hurry... Did I answer your questions?
Thank you again for your detailed explaination.
Now I am clear about it.
Best regards,
Du
|
