Why peaking near 0dBFS is unnecessary when mixing audio

Modern digital audio workstations can operate at 24 bit resolution and above, this provides a theoretical dynamic range of 144dB. This in practice is usually limited to around 120dB due to the limitations of the analog circuits that convert the audio to and from digital. The small analog operational amplifiers in the sound cards will produce a noise floor slightly above the digital noise floor. With this in mind it comes as a surprise that many people mix with their DAW meters peaking very close to the top of the scale.

This is unnecessary and has some negative implications, when your audio peaks near the end of the scale there is a higher chance that the audio clips (distorts by exceeding digital zero) because there is very little headroom to avoid any momentary overshoots. The mathematic calculations in the DAW are not infinite and end at 0dBFS. If you overshoot this final value at 24 bit fixed point your signal is not being represented with accuracy anymore. This is distortion, the peaks no longer look rounded but instead square off and look flat.

In addition when a signal approaches 0dBFS it is actually producing a relatively high electrical signal level for a typical home studio sound card to reproduce. Again in the worst case scenario it can mean further distortion but in the analog domain. All signals that are displayed on an audio meter are ultimately represented by an analog voltage which appears at the output of your sound card TRS jack outputs. This electronic voltage is the audio signal in question and it is important how that signal interfaces with other devices be it analog or digital.  

Technical history

Lets look back at how the gain structure of a large format mixing desk would be set up. It would be normal to peak signals up to the 0 Vu mark on the mixer channels signal metering. When a signal is peaking at 0vU on a large professional mixing console it is approximately the equivalent of -12dBFS (on a digital meter like those in a digital audio workstation) So at 24 bit resolution it is absolutely unnecessary to peak your signals anywhere near close to 0dBFS. It worked nicely for professional mix engineers in big bucks studios so why not for you ?

By optimizing your gain structure in this manner you are building in valuable clarity and headroom so you can hear the details of your music mix down better.   
 
Emulations

Another factor is to consider if all of the analog emulation software you have sets it’s own internal gain structure to operate at electrical levels of the original devices. If so and you regularly send signals of -2dBFS you may well be over driving the inputs and creating more THD (total harmonic distortion) than the original devices were meant to. Different software developers may have modeled the over load characteristics of the devices they emulate using the same electrical signals as the original device was meant to be receiving.

Of course you may well wish to do this intentionally for effect, but it is best to understand the principles of audio gain structure and be it’s master rather than it’s slave through ignorance. This way you can accurately manipulate the gain structure by working with nominally sensible signals and adjusting them as you see fit to achieve the sonic characters you wish for.

Once you have completely your mix down you can export the mix down file as a 24 bit stereo interleaved file and either ‘self finalize’ the music in a separate session or send out for professional mastering. In both cases you will find there will be ample headroom in the file to be able to master the track with ease without having to pull down the stereo master fader for any reason.

Summary

 
Gain structure is a very important and fundamental audio engineering concept which all rookie engineers, musicians and producers need to understand in order to produce the best recordings, mixes and end results for their music tracks.

Barry Gardner operates SafeandSound Mastering