Mastering Headroom

  • Thread starter Thread starter bouldersoundguy
  • Start date Start date
bouldersoundguy

bouldersoundguy

Well-known member
I see a lot of wave file mixes mastered to 0 dBFS peak. That has the potential for producing peaks above 0 dBFS. In fact, it's virtually guaranteed given modern audio production methods. There are two reasons for this.

First is how digital data is converted to analog. DACs have to process the audio with a low pass filter during conversion to properly reconstruct an analog waveform. As a result, the analog waveform can "overshoot" the peaks represented by the digital samples ("intersample peaking"). This will cause some converters to distort on peaks. The "True Peak" measurement in LUFS meters is designed to predict this level. I find that a digital peak level of -0.3 dBFS gives sufficient room to prevent excess True Peak levels from a wave file in most cases.

But there's a second problem, which is the effect of compressed file formats. I've found that converting to mp3 results in an increase in a file's peak levels. It varies a bit depending on the file, but more than 0.5 dB is typical. In the attached images, you'll see that the True Peak level increased 0.8 dB and the LUFS integrated level decreased by 0.2 dB as a result of converting a file from wave to 128 kbps mp3.

This is why I master files with a full 1.1 dB of True Peak headroom. It's also probably why the meter turns red for anything above -1 dB True Peak (see red marks above the level trace).

LUFS wave.webp


LUFS mp3.webp
 
You might just like this guy's channel. Been following for a while and he is thorough with the technical side of things. Very, very different from the old school days.

 
One gripe so far (not done watching). DACs that clip with intersample peaks aren't necessarily "cheap," they probably just meet the Red Book standard, which specifies output voltages per digital level. In the early days of digital audio, processing was done in analog and converted directly to the end product. If there was no clipping in the ADC, there would be no clipping in the DAC. Digital processing wasn't available so it wasn't taken into consideration.
 
One gripe so far (not done watching). DACs that clip with intersample peaks aren't necessarily "cheap," they probably just meet the Red Book standard, which specifies output voltages per digital level. In the early days of digital audio, processing was done in analog and converted directly to the end product. If there was no clipping in the ADC, there would be no clipping in the DAC. Digital processing wasn't available so it wasn't taken into consideration.
What's that old expression, it's not the destination but the journey? Hard to watch this guy and not learn a few tricks. Don't have to agree with the outcome. A few weeks back he did a long back and fourth with another engineer on loudness vs dynamics. With my background, I'm a big dynamics guy but tools are just tools and it is up to you as to how you use them.

If nothing else, it has been fun to watch him evolve in front of my eyes. It's also like having a guy that will read the manual and explain it all to you.
 
I thought the technical info was fine, it was more of a minor nitpick.

DACs were designed a certain way to provide consistent performance, output level per digital level.

Making DACs so they don't clip is fine, but it's really pretty simple to make audio files that don't clip regardless of the converters. And there are lots of legacy DACs out there, so why not?
 
Thanks for the heads-up @bouldersoundguy! I'll be keeping all this in mind next time I do some renders in various formats.
 
Harmonic distortion is hard to fix

I have learned how to fix it (and have fixed it for my clients) but it's better to not go there in the first place :)
 
A couple of thoughts I've had on this ...

I was on the "mastered headroom" bandwagon for a long time, until I realized that intersample clipping is mostly an audiophile issue - most people don't notice it at all. And the majority of true peak limiting that I've auditioned sounds worse than just regular limiting with over-sampling enabled (there are zero intersample peaks when I've tested this method against true peak). I don't see any value in True Peak limiting as a result when you can get a better sounding result by doing regular limiting with oversampling. So I stopped overly caring about True Peak limiting as well as leaving headroom on the table for streaming. Honestly, the only place I still consider leaving headroom these days is for Apple Music releases (even though the Apple Masters gimmick is just that - a gimmick).

Instead, I just make sure my masters sound as good as possible on my set-up and everywhere I audition them.

Part of my auditioning process is that I convert my masters to MP3 and other formats to test the sound out. When I do it, there is no clipping even when I master to -0.02dbFS because I'm using an excellent downsampling filter and the highest-quality conversion settings in my DAW for each bitrate.

Spotify and Youtube etc are using crappy conversion methods, and I've decided I'm not going to let that dictate leaving almost 6db of perceptual loudness off the table ;)
 
Last edited:
"True Peak" is a kind of measurement, not a kind of limiting. A limiter might specify a target in terms of TP, but it can achieve that by whatever means. Heck, you can just decrease the gain and not change the dynamics at all.

I'm not sure if there are any modern limiter plug-ins that don't oversample.
 
Back
Top