How does Adobe Audition calculate the dB scale in the Frequency Analysis window

Forum|Forum|4 years ago
August 18, 2021
1 reply
1561 views

Hello, I am a research student at Florida Atlantic University and I have been trying to mimic the waveform obtained in the Frequency Analysis plot:

I am able to get the x-axis from the python code that I wrote, but I do not know how to get the dB y-axis. This is what I actually plot from my python code:

I get a positive magnitude when I do a FFT analysis and plot the waveforms. How is the y-axis value actually calculated? Is there some kind of normalization that takes place?

Any help will be greatly appreciated to help me get a resolution as to the y-axis dB scale. Thanks!

How to

This topic has been closed for replies.

SteveG_AudioMasters_

Community Expert

We can't give you a definitive answer, because there's no way that Adobe will ever tell you how they calculate anything, I'm afraid! Essentially though, if you calculate the RMS value of the data in each FFT window, you should end up with the correct result. Which is why, if you increase the size of the window (reduce the FFT size), the value for that window is likely to be greater. It's an RMS value, so it's always going to be positive. There's no normalization at all - you will notice that if you drag the Y axis scale, it can be dragged up and down, so you can centre it where you want. But essentially, it's down to the FFT size you use (and the windowing).

S

supersonicpandaAuthor

Participant

Thanks for your input Steve. I sincerely appreciate it. As I read on the Adobe website, when you hit the "Scan Selection" button, that is equivalent to:

"Scans the entire file or selection, and displays average frequency data in the graph. (By default, the graph displays data from the center point of files and selections.)

I am just thinking, in order to get the dB number that is reported on the y-axis...From the Amplitude Statistics, the Total RMS Amplitude = -23.02dB and the Average RMS Amplitude is -36.08dB. Does this make sense:

plt.semilogx(f, 10*np.log10(Pxx_den)-36.08)

Where Pxx_den is the power spectral density? Even with that, I am still about 20dB shy of the Adobe plot.

SteveG_AudioMasters_

Community Expert

I stopped trying to apply any sort of formulaic sense to what Audition does a long time ago. But as far as your estimate goes, you have to bear in mind that for the Amplitude Statistics, the calculating window size is fixed (although you can vary it), so the results from an RMS histogram and what happens in the Frequency Analysis window won't necessarily correlate at all. Also we have no idea how they calculate all of those RMS figures either, so really you are on a hiding to nowhere with this in Audition, I'm afraid.

Sign up

To post, reply, or follow discussions, please sign in with your Adobe ID.

Sign in to Adobe Community

To post, reply, or follow discussions, please sign in with your Adobe ID.

Scanning file for viruses.

This file cannot be downloaded