Page 1 of 1

Help needed for using frequency analysis

Posted: Wed Jan 04, 2012 4:36 pm
by screamer
Hey guys,

I'm need some help to determine the fundamental frequency of the attached graph (i have attached the graph and also the .txt file). The sound being analysed is baby crying. From research topics I found out that the fundamental frequency should be some where around 750-1200 Hz. But I don't know how to read the graph. I would greatly appreciate if someone can point me out to the fundamental frequency of the sound in question.

Thanks

Re: Help needed for using frequency analysis

Posted: Wed Jan 04, 2012 5:28 pm
by steve
It's probably that big spike on the left.
Change the "Axis" option from "Linear" to "Logarithmic" and you'll be able to see it more clearly.
Then position your mouse over the graph and move the mouse pointer so that it is directly under that big spike - at the bottom of the window there will be a read-out for the actual cursor position, and for the nearest peak. It's the peak that you are interested in.

Re: Help needed for using frequency analysis

Posted: Thu Jan 05, 2012 5:55 am
by screamer
One more question is why is the dbl negative. I know it's a silly question but I don't know. Will it ever be positive?

Thanks

Re: Help needed for using frequency analysis

Posted: Thu Jan 05, 2012 6:35 am
by screamer
How can I manually scale the x-axis of the analysis graph?

Re: Help needed for using frequency analysis

Posted: Thu Jan 05, 2012 5:31 pm
by steve
The x axis is scaled automatically according to the sample rate.

I think that it's been requested previously for manual scaling, but the ability to do so would probably be less useful than it first appears, unless there was a major overhaul of how the tool works. The analysis breaks down the audio into lots of frequency bands. The "Size" parameter determines the number of frequency band and there are half the number of frequency bands as the value set in "Size", so if the FFT size is set to 128 there will be 64 frequency bands. The highest "resolution" is when Size is set to 16384, which produces 8192 frequency bands spaced at something like every 3 Hz for a sample rate of 44.1 kHz. If you stretch the spectrum window to full screen (or larger) you can already see a readout accurate to around 3 Hz, so being able to "zoom in" (re-scale the x-axis) you will not actually gain very much.

For people that require greater accuracy the spectrum plot can be exported as a text file, which produces a table of frequency values (Hz) and their value (dB) which could be extrapolated in a suitable spreadsheet application, but that's only really relevant for people conducting scientific analysis, in which case there are probably better tools than Audacity (such as Praat or Sonic Visualiser)