Measuring amount of distortion in sound files

Hi all,

Bit of a weird one, but here goes.
I am working with sound recordings from field work in a windy environment, which has resulted in large portions of distortion when wind speeds are high.
I am working with over 500, 1 hour recordings, and I need to come up with a numeric value to determine the % or total time or distortion events within each 1 hour recording.

The results are hopefully going to be used to publish data on the efficacy of songmeters in windy environments, but I have reached a standstill with this at present.

Any suggestions of plugins, features or other software is welcome. Please keep in mind, I do not want to spend copious amounts of time on this analysis as it is just to determine one value for each recording.


I can think of two ways to tell when your recording is trash. View > Show Clipping.

Screen Shot 2020-09-28 at 12.32.15 AM.png
Any red bands in the performance indicate places where the wind had pushed the recording past overload into distortion.

Also Effect > Amplify. If both numbers are 0 the show is gone.

I don’t know that there is a more gone or less gone. If both of those indicators fail, the show is not recoverable. The show may be partially recoverable for information or surveillance, not for entertainment.

These only work if you have the raw sound files, and it only works cleanly if you’re in WAV or one of the other uncompressed formats. MP3 or a compressed format will scramble the results. Any filtering, effects, or corrections will mess this up. You can’t turn the volume down later and get the show back. There is no show to get back. The digital system stopped following the sound.

Classic Distortion Analysis is a controlled system. You put a carefully generated pure tone into a system and measure the system output. Anything that comes out that’s not pure tone is distortion and can be measured in percent. You can’t start with a wild feed or field recording and do that. You have to know—accurately—what sound went in.

Field recording is not for the easily frightened. It’s the realm of wind socks, high pass filters, and Dead Cats. There are Hollywood sound people who do this.

This Rode microphone control deals with this problem. The top two settings are rumble filter and reduce the volume so it doesn’t overload.

Screen Shot 2020-09-28 at 12.56.02 AM.png
Dead Cat is an affectionate term given to a fur wind screen because it looks like…you know.

Screen Shot 2020-09-28 at 1.03.40 AM.png
What’s this product you’re working on?

I know of one microphone system that works in two different volumes at the same time to avoid problems like this, or at least greatly increases the chances of walking away with a working production. But still, for all practical purposes you can’t be a New User.


I assume that you are referring to the “Song Meter wildlife recorders”?

It is not possible to measure the “distortion” unless you also have a recording of exactly the same sound that has zero (or at least “much less”) distortion.
Distortion is measured as the difference between the “distorted” sound, and the undistorted original. As you don’t have an undistorted original to compare against, it will not be possible to measure the “distortion”.

One thing that you can do, is to measure the low frequency noise.
“Wind blast” creates a lot of low frequency noise. In a quiet and peaceful natural environment you would expect there to be very little low frequency noise (unless you are near to a road, quarry, waterfall, or some other low frequency noise source).

To get a single figure for low frequency noise, first install the “rms” plug-in. It is shipped with Audacity 2.4.2 but is disabled by default. It is listed in the Plug-in Manager as “rms”. When installed, it appears in the Analyze menu as “Measure RMS”.

  1. Normalize the recording to 0 dB (this provides a standard level for comparison)
  2. Resample the track to 300 Hz sample rate (this removes all frequencies of 150 Hz and above)
  3. Measure the RMS level.