Print

Print


Hi Chris,

Matlab's resample function that we use filters the data before
downsampling to prevent aliasing so taking that into account my answer
is correct.

Best,

Vladimir

On Mon, Jul 27, 2009 at 9:57 AM, Christophe
Phillips<[log in to unmask]> wrote:
>
> Hello Vladimir and Hamid,
>
> I would rather disagree with Vladimir's answer and say that, the order does
> matter.
> In theory because of the aliasing effect of downsampling, you could well
> ends up with "spurious" signal in the lower part of your downsampled signal
> spectrum.
>
> Example: your original signal is sampled at 800Hz, and you want to
> downsample it at 400Hz, and low pass filter at 150Hz
> Take a signal at 300Hz in your original signal, it's below the 400Hz Nyquist
> frequency, so no problem.
> Keep one sample out of 2, i.e. downsample at 400Hz, then 300Hz>200Hz, the
> new Nyquist frequency. So the orginally 300Hz signal will appear as a 100Hz
> (*) signal. Low pass filtering at 150Hz will not remove this spurious
> signal...
>
> You should thus low pass filter, at least at the new Nyquist frequency (half
> the new sampling rate), your signal before actually downsampling it.
>
> Best,
> Chris
>
> (*) 200Hz - (300Hz-200Hz)
>
> Vladimir Litvak a écrit :
>
> Dear Amir,
>
> In theory there should be no difference. Since after downsampling
> filtering will work faster I'd suggest you downsample first, just
> because of processing speed.
>
> Best,
>
> Vladimir
>
> On Sun, Jul 26, 2009 at 5:18 PM, Amir H Javadi<[log in to unmask]> wrote:
>
>
> Hello everyone
>
> I want to down-sample and filter my data. Which sequence does result a
> better data, down-sampling and then filtering, or filtering and then
> down-sampling?
>
> Have a good time
> Amir
>
>
>
>
>
>