Oh I see.... I can't really tell for sure what is happening in the upper right,
but that might amount to the same thing. I hope you program a little better than
you draw :^)
-jim
>
> After all the wise advices in this thread it looks quite immature ))...
>
> Thanks!
>
> ----------------------------------
> http://www.ufo-scientific.com
> Your favourite audio/midi software
>
> >>It is pretty simple to accomplish what you want to do. All you have to
> do
> >is
> >>downsample the data by 4 and then take your 1024 point FFT.
> >>
> >>-jim
> >
> >Should this downsampling be complemented by any kind of filtering?
> >
> >
>
> well, now I start to realize that it's much the same idea as I sugested in
> my drawing above. But it looks much more elegant.
I didn't see a drawing.
>
> Thanks for this.
>
> But what kind of concerns me a bit is the alias effect that will appear
> after downsampling.
>
It's hard to give anything more than generalized advice since I don't know how
you intend to accomplish any of this. I don't know if you are going to roll your
own code or if you are just using ready made functions (like Matlab, for
instance). If it is the latter then there are probably functions available to
you that will take care of the filtering and downsampling probably all in one
function call.
-jim
> that's why I'm asking if i should pre-filter the data before
> downsampling.
>
> ----------------------------------
> http://www.ufo-scientific.com
> Your favourite audio/midi software
Reply by hq9000●March 18, 20092009-03-18
>Yes that would probably be a good thing. If your data has energy above
>the 0-Fs/8 band that you are interested in then you will need to low
>pass filter as part of the process of reducing the sample rate. If the
>data already is free of energy in the band from Fs/8 to Fs/2 then you
>could just keep every fourth sample and discard the rest.
> Or you might just acquire the data at the lower rate in the first
>place, but that should have a filter in the process also.
>
>
>-jim
Yes, didn't see your post when replying.
Thanks for explainations!
------------------------------
http://www.ufo-scientific.com
Your favourite audio/midi software
Reply by Jerry Avins●March 18, 20092009-03-18
hq9000 wrote:
>>> It is pretty simple to accomplish what you want to do. All you have to
> do
>> is
>>> downsample the data by 4 and then take your 1024 point FFT.
>>>
>>> -jim
>> Should this downsampling be complemented by any kind of filtering?
>>
>>
>
> well, now I start to realize that it's much the same idea as I sugested in
> my drawing above. But it looks much more elegant.
>
> Thanks for this.
>
> But what kind of concerns me a bit is the alias effect that will appear
> after downsampling.
>
> that's why I'm asking if i should pre-filter the data before
> downsampling.
And that's why you've been answered yes. The signal will be corrupted if
it contains frequencies above half the new sampling frequency. Filtering
them out before discarding samples is a necessary part of the
downsampling procedure. Note that the filter needn't calculate samples
that are to be discarded.
Jerry
--
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������
Reply by hq9000●March 18, 20092009-03-18
>>It is pretty simple to accomplish what you want to do. All you have to
do
>is
>>downsample the data by 4 and then take your 1024 point FFT.
>>
>>-jim
>
>Should this downsampling be complemented by any kind of filtering?
>
>
well, now I start to realize that it's much the same idea as I sugested in
my drawing above. But it looks much more elegant.
Thanks for this.
But what kind of concerns me a bit is the alias effect that will appear
after downsampling.
that's why I'm asking if i should pre-filter the data before
downsampling.
----------------------------------
http://www.ufo-scientific.com
Your favourite audio/midi software
Reply by jim●March 17, 20092009-03-17
hq9000 wrote:
>
> >It is pretty simple to accomplish what you want to do. All you have to do
> is
> >downsample the data by 4 and then take your 1024 point FFT.
> >
> >-jim
>
> Should this downsampling be complemented by any kind of filtering?
Yes that would probably be a good thing. If your data has energy above
the 0-Fs/8 band that you are interested in then you will need to low
pass filter as part of the process of reducing the sample rate. If the
data already is free of energy in the band from Fs/8 to Fs/2 then you
could just keep every fourth sample and discard the rest.
Or you might just acquire the data at the lower rate in the first
place, but that should have a filter in the process also.
-jim
Reply by hq9000●March 17, 20092009-03-17
>It is pretty simple to accomplish what you want to do. All you have to do
is
>downsample the data by 4 and then take your 1024 point FFT.
>
>-jim
Should this downsampling be complemented by any kind of filtering?
Reply by jim●March 16, 20092009-03-16
hq9000 wrote:
>
> Hello people,
>
> Standard FFT lets us retrieve amplitudes of evenly distributed
> frequencies.
> What if I need at some frequency range greater/smaller density of
> frequency points...
>
> Let's say I take fft of 1024 sample long chunk. After standard FFT it will
> give amplitudes of
> f[i]=i*samplRate/1024
>
> the step is sampleRate/1024
>
> now let's say I want to have step equal to sampleRate/(1024*4) for
> frequency range [0; sampleRate/8].
>
> The question is if I can take this FFT only partially, since I'm not
> interested in frequencies higher than sampleRate/8.
>
> Any ideas? Tricks?
It is pretty simple to accomplish what you want to do. All you have to do is
downsample the data by 4 and then take your 1024 point FFT.
-jim