discuss-gnuradio
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Discuss-gnuradio] Signal processing question : FFT frequency resolu


From: Tomaž Šolc
Subject: Re: [Discuss-gnuradio] Signal processing question : FFT frequency resolution
Date: Tue, 7 Nov 2017 08:51:02 +0100
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101 Thunderbird/52.4.0

Dear Steve,

in general you cannot have 1 Hz resolution by analyzing only 100 ms worth of signal. A simple way of looking at it is that you are only fitting one tenth of a 1 Hz waveform in your window. So by only looking at the waveform through your narrow window you cannot know whether the waveform repeats, say, every 1 s or every 200 ms.

If you want to have a FFT with 1 Hz resolution every 100 ms what you can do is to have 1 s long sliding window (e.g. first FFT is 0.0 - 1.0 s, second FFT is 0.1 - 1.1 s, etc.) This obviously requires changing the FFT size (and 1 M sample FFT is quite large).

I think the correlation approach you propose is similar to doing FFT on a 100 ms window that has been zero-padded to 1 s. Indeed you get 1 Hz frequency bins, but their content is effectively only an interpolation between 10 Hz bins. You gain no new information.

However, if you can make some assumptions about the signal you are analyzing (for example, you know that there is only a single tone between 1 and 10 Hz), then some kind of correlation or curve-fitting approach might work better than FFT.

Best regards
Tomaž

On 06. 11. 2017 16:42, Steve Gough wrote:
Hi all,

Could you please help answer a quick signal processing question ? I apologize if this is the wrong mailing list.

I have a sampling rate of 1M samples/second, and I perform an FFT every 100ms. This gives me a frequency resolution of 10Hz (i.e. each bin size is 10Hz). Is there any way I can increase my frequency resolution (say from 10Hz to 1Hz) without changing the FFT size ?

If I "manually" correlate different frequency sinusoids (1Hz sinusoid, 2Hz sinusoid etc) with the samples acquired in the 100ms duration, can this give me a higher frequency resolution (1Hz in the above case) ?

Thanks in advance!
Steve




_______________________________________________
Discuss-gnuradio mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio




reply via email to

[Prev in Thread] Current Thread [Next in Thread]