Re: [Discuss-gnuradio] FFT bin ordering

[Dirk Gorissen]

Hi Marcus,

Sorry for the delay, thanks for the elaborate reply. My use case is
quite specific actually, in that Im looking for a peak in a specific
part of the spectrum. I had another look and failed to reproduce the
problem I was having so maybe it is a problem between keyboard and
chair issue but I thought it safe to clarify.

So consider this case closed for now :)

Thanks again!

Cheers
Dirk

On 12 March 2018 at 12:05, Müller, Marcus (CEL) <address@hidden> wrote:
> Hi Dirk,
>
> On Sat, 2018-03-10 at 22:22 +0000, Dirk Gorissen wrote:
> …snippety snip…
>> My assumption was that the center frequency is at the middle bin
>> (Shift = True) and that if I want +/- 500Hz (for example) I can just
>> take the bins (middle bin index - 500) -> (middle bin index + 500)
>
> Think about how wide your bins are! An FFT transforms "input vector
> length" =: N samples from time domain into frequency domain. Thus,
> every bin represents Nyquis bandwidth / N of bandwidth. Since Nyquist
> bandwidth is sampling rate, you get f_sample/N*1000 of bandwidth.
>
>> The python block should be trivial in that it just snips out a subset
>> of the FFT bins (around the center frequency).
>
> Now, to act a bit like Clippy:
>
>> It looks like you're building a decimating low pass filter. Do you
> want help with that?
>
> Generally, cutting out spectrum from the DFT of your signal *might*
> lead to a distorted view at your signal; see [1].
> It might be OK (correct, even), however, for specific purposes. What's
> the bigger thing you're planning to do?
>
> Best regards,
> Marcus
>
> [1] https://dsp.stackexchange.com/questions/6220/why-is-it-a-bad-idea-t
> o-filter-by-zeroing-out-fft-bins



-- 
_________________________________________
Dr. Dirk Gorissen
Mob: +44-7763-806-809
Web: dirkgorissen.com
Skype: dirk.gorissen
Twitter  : twitter.com/dirkgor
LinkedIn: linkedin.com/in/dirkgorissen