AnsweredAssumed Answered

vrf Increasing spectral resolution of fft.

Question asked by bill.ossmann on Apr 6, 2006
Nick,

As always, it depends on what you are trying to do. Fundamentally, you
don't have a long enough data sample to justify 10Hz resolution. That said,
there are circumstances in which you can make reasonable accommodations.
For example, if you know your signal of interest is ~0 beyond the time of
the measurement sample, you can simply zero-extend the data to get out the
necessary time interval. This can be a good way to interpolate bandwidth
measurements on a short duration signal, for example. You will need 2200
points which isn't outrageous unless you need to do lots of them in a very
short time.

A 6.5ms portion of a pure sinusoid will have a 3dB bandwidth of about
135Hz, so with FFT processing, that's pretty much what you're stuck with.
The same goes for the noise components. Zero-extension won't actually
increase the resolution on them, but will interpolate smoothly between the
points you had with coarser resolution.

All that said, there are other signal estimation techniques that might
help, but most of what I ever knew I've forgotten due to disuse.

Regards,
--
Bill Ossmann
Philips Ultrasound

"Nick Evans" <nick@genesysibs.com> wrote on 04/07/2006 12:02:32 PM:

> Hi All
>
> I was wondering if anyone could give me some pointers (if pointers
> are to be had) on increasing the frequency resolution of an fft or
> similar.  I will try to explain.
>
> I need to provide a spectral estimation for a digitised signal.
> The signal is digitised at a sample rate of 22kHz.   The total time
> span of the signal is approx. 6.5mS.   I need to get a spectrum for
> the band of frequencies up to 4kHz with an equivalent resolution
> bandwidth of 10Hz.   For a standard fft this is not possible as the
> frequency resolution is directly (inversely) proportional to the
> time span of the signal, meaning that I would need to have a signal
> of 100mS duration in order to get an equivalent RBW of 10Hz.
>
> I have tried things like matlab’s chirp-z function which works ok as
> a kind of zoom fft but it does not increase the frequency
> resolution.   The stimulus signal is (should be) a periodic sine
> wave but with varying degrees of random noise depending on environment.
>
> If anyone has any tricks up their sleeve they would be much
> appreciated.    Alternatively, if I find something elsewhere, I’ll
> post here in case its of interest.
>
> Thanks and regards
>
> Nick
>
---
You are currently subscribed to %%list.name%% as: %%emailaddr%%
To subscribe please send an email to: "vrf-request@lists.it.agilent.com" with the word subscribe in the message body.
To unsubscribe send a blank email to "leave-vrf@it.lists.it.agilent.com".
To send messages to this mailing list,  email "vrf@agilent.com". 
If you need help with the mailing list send a message to
"owner-vrf@it.lists.it.agilent.com".
Search the "unofficial vrf archive" at "www.oswegosw.com/vrf_archive/".

Outcomes