Советы по мастерингу. Громкость

Ron Parker

Last revision: 17 October 2004

The task

This document is a definition of loudness in audio mastering and an introduction to techniques for solving the basic problem. I include quotes from two letters as arguments in favor of loud mastering. The album referred to in the first review is a job for which I requested advice from the Linux Audio User mailing list for techniques to achieve loud mastering. For some of us, the ensuing thread concluded that loud mastering is bad.

"Just heard the new CD. This is a phenomenal piece of work. The recording is very crisp and bright. Top of the heap. First rate! No kidding, this sounds much better than the majority of my CDs. It's like my stereo was designed to play it."

The job order that caused this "loud" master basically stated, "Achieve every bit of loudness that you can get". And I did!

After completing the above album, I engineered, mixed and mastered the audio for an 8-minute art film. To design an appropriate mix and master, it was vital to know that the final destination format was to be DVD-Video and the venues for playback would vary from 500-seat theatres to art center installation rooms that fit a couple dozen people.

After asking numerous questions I concluded that the quality of playback systems would range from great to junk. Based on these insights, I designed a mix that lends itself to loud mastering. By minimising dynamics and carefully controlling total bass energy the level of the average soundfloor could be easily increased to produce a loud master. My reasoning is that low volume playback on garbage systems will lose an appreciable amount of information. And the client review:

"Doctor, good job on the film project, sounds good on our system..... nice bright & clear notes, should survive [production house name]"

NOTE: Unlike the album review, the film hasn't hit the streets and passed the acid test. I won't know for a couple months but fully expect a dozen attendees crammed into the corner of a room, listening to a boom box playback system, to hear every single note of the audio track while a raging orgy of ecstatic wine and cheese tasters go bananas in the background.

It's my opinion that to claim loud mastering is bad is the posturing of audio engineers that don't understand the task. In my experience, loud mastering is a style that is the perfect solution for specific challenges. The film soundtrack is an example of a job that requires loud mastering.

[back to top]

Techniques for controlling loudness

I define Mastering as having the single objective of controlling loudness. The challenge is to increase the level of the average sound floor for any mix.

JAMin and every other mastering solution have numerous gain stages; input, EQ, compressor, limiter, boost/softclip, output.

JAMin
Figure 1: JAMin

Assume you have ten control room mixes that aren't loud enough, and that each of them has a max peak of -9.8dBfs. The first opportunity is to increase peaks to about -0.2dBfs. The +10dB of input gain produces a max peak of 0.2dBfs.

Spectrum view of unmastered stereo mix
Figure 2: Spectrum view of unmastered stereo mix

But this only begins to exploit the potential loudness. The opportunity is in creating headroom so the level of the soundfloor average can be increased towards 0.0dBfs. The peak between 25Hz and 100Hz is the loudest frequency while the average is the max response level for the majority of the frequency range which is everything above 100Hz. Equalization can be used to cut 25Hz to 100Hz by 8.db to create a more average soundfloor.

Spectrum view of equalized stereo mix
Figure 3: Spectrum view of equalized stereo mix

Using equalization to cut the dominant frequency range has created about 8dB of headroom. All of the aforementioned gain stages can be used to move the average closer to 0.0dBfs. Obviously, real mixes have numerous "errant" frequency bands. The following image shows incredible loudness.

Spectrum view of average sound floor moved towards 0.0dbfs
Figure 4: Spectrum view of average sound floor moved towards 0.0dbfs

To close the gap even further we could adjust the low range band of the JAMin compressor to apply a 10:1 ratio.

[back to top]

Multitrack sources

The advantage of multitrack sources is a level of control that isn't available when working with stereo files. Even though good mixes don't require a great deal of tampering, common problems are discovered and easily fixed during mastering. Regardless, there is a serious limitation; it's difficult beyond reason to put 10 multitrack sources within one session file, and mastering is the task of making many songs sound great together.

Consider a multitrack source where the snare drum is the loudest instrument and it has +6.0dB transient peaks. When this mix is routed to a stereo bus which in turn routes to JAMin the instrument controlling the potential level for the average soundfloor is the snare drum. When the LADSPA TAP Limiter is used to eliminate the +6dB spikes on the snare track the resulting average soundfloor (entire mix) can be increased +6.0dB without causing clips.

The TAP Limiter as it appears in Bob Ham's JACK Rack
Figure 5: The TAP Limiter as it appears in Bob Ham's JACK Rack

Assuming low frequency energy below the audible range of 20Hz is a cause for peak levels, Fons Adriaensen's JACK Audio Analyser can be used to examine individual tracks for excessive energy and equalisation can be adjusted. Imagine three instrument tracks where -3.0dB equalisation cuts in the low frequency ranges around 20Hz create an accumulated 9.0dB of headroom without having a negative effect on the sound quality. The resulting average soundfloor is a much better mix for the mastering engineer.

The JACK/ALSA Audio Analyser
Figure 6: The JACK/ALSA Audio Analyser

These examples are far from reality but they describe what mastering is and how to use a typical set of tools to achieve the greatest potential for loudness.

[back to top]

Acknowledgments

To a small degree this document is a response to Rip Rowan's Over The Limit article referenced during a thread in which I requested assistance for achieving maximum loudness.

The author also thanks Daniel James for editing this version.

[back to top]