The fade issue in mastering
You have probably stumbled upon the difficult task of a simple fade out (or in) in mastering (and/or mixing). It just doesn’t sit right, seems uneven and stumbles in its seemingly easy task of just lowering the volume. It also sounds as if there are high- and a low-pass filters that are slowly closing the frequency spectrum and start strangling the life out of the audio. This is often extra intrusive when there’s vocals present.
A theoretical display of how human experience a fade out
You would think that the best use of a fade out would be a linear fade where the sound pressure level is lowered just as much over the entire fade. This is often completely unusable and it’s more standardized to use a fade that matches the human hearing better like the logarithmic or S-shaped fade. They decrease the velocity at around half of the fade. This makes it more apprehensible though its not quite there and the frequencies are still acting weird.
Finding the frequency problem
This got me thinking of other ways to improve the flaws of the fade. I started to experiment with different solutions but the audio just wouldn’t ”sing” along with the fade. Then I remembered the ”equal-loudness contour” (also known as the Fletcher Munson curves) and started to think that the problem maybe wasn’t the audio but the human perception of it.
This animation shows the perceived loudness of frequencies at different sound pressure levels.
You can test the concept by playing a fade out of a song and at the same time counteract the sound level with the volume knob so that the audio is consistent in sound pressure level (volume). The audio will then sound even in the frequency spectrum though if not listening in 24bit (or better) you will probably also experience the bit depth issue.
Improving the fade
To solve/improve this I put an eq last in the plug-in chain and created a personal eq curve that implemented the difference between the 80dB sound pressure level (a kind of normal listening level) and the hearing threshold (according to the equal-loudness contour). I saved this as a ”fade out trick" preset and automated the eq from a flat eq where the fade started to the ”fade out trick” preset setting at the very end of the fade.
This fade out trick makes the frequencies more even through the entire fade and it really retain the song quality throughout the fade out. I did a fab-filter pro-q 3 preset you can use or you could just use these values in your preferred eq.
- +20dB bell at 50Hz, Q 0.3
- -5dB bell at 3kHz, Q 1
- +5dB bell at 14kHz, Q 1.5
Remember to automate from a flat frequency curve (to these values) when doing the fade out.
When listening to a 100dB sine wave at 20Hz it is perceived as 20dB lower than it actually is but when listening to the same frequency at the hearing threshold it is perceived as 70dB lower than it actually is. This phenomenon is called the equal-loudness contour or sometimes the Fletcher Munson curves (though this is old standards). Read more about this in my blog post about the equal-loudness contour and the importance of a consistent listening level in mastering.