Make your own free website on
  Using Multiband Compression for Dance Music Mastering

Multiband compression can bring out, as well as fix, important details in a mix

By Craig Anderton

Mastering is about creating the finest possible showcase for an artistís tracks. Usually this means trying to maintain state-of-the-art quality and transparency, but with some tracks, the "finest possible showcase" may have different priorities.

For example, take dance mixes. As DJs spin recording after recording, a consistent level is important - you don't want to segue into a radical level drop, particularly if you're building to the peak of the evening. As a result, dance musicians agree to a de facto standard reference level: put as much average level as is technologically possible on a recording.

With CDs, this means compressors, equalizers, "curve-stealing" software that analyzes a tune's spectral response and superimposes it on a different file, and so on. However, I don't necessarily like to squash the dynamic range to a bare minimum; dynamics account for much of a song's emotional impact. Therefore, mastering for dance music walks a fine line between maintaining enough dynamics to be musically interesting, while being loud enough to hold its own when bookended between two cuts that were engineered for flat out, maximum level. Fortunately, today's plug-in tools go a long way toward optimizing a mix with as few compromises as possible.


Traditionally, the mainstays for mastering have been high-quality dynamics control, and flexible equalization. However, evolution of the art has brought us the multiband compressor, which combines filtering and compression. The way it works is you specify a range of frequencies to be affected by an associated compressor. Meanwhile, a separate range of frequencies can be affected by a different compressor. This allows for tricks such as tightly controlling the bass response through limiting, while adding midrange compression in, for example, the 1 kHz region to improve "snap" and allow instruments like guitar to "speak" a little better.

My first brush with multiband dynamics was the Dolby 740, a fine "spectral enhancer" that remains unchallenged in the analog domain. When it comes to plug-ins, you can choose from a variety of multiband compressors. My current favorite is Spectral Design's plug-in that's part of the Steinberg Mastering Edition software package. (Frankly, part of why I like it is because the interface is very similar to the Quadrafuzz, a multiband distortion unit I specíed for Spectral Design. So, there was zero learning curve.) However, some software packages (such as Sound Forge) include multiband compression as a standard feature, and there are other multiband compression plug-ins. These all work fairly similarly, so whatís described here should be easy to translate to whatever you're using.

In most cases, you won't need multiband compression; subtle EQ and compression will do the job. However, for tough cases, sometimes the only thing that will work is multiband compression, so let's focus on that.


Multiband compressors typically include 3 to 5 stages. Often, 3 stages (for roughly low, mid, and high) is enough; any more can just complicate matters. Many times what you really need instead of more stages is simply a stage of traditional parametric EQ. However, for some "problem cases" where it's crucial to apply a specific amount of dynamics processing to a specific frequency range, 4 or 5 bands can come in handy.

Let's look at a real-world example of setting up some mastering plug-ins, then zero in on the multiband compression. My most recent mastering project was for the new Rei$$dorf Force CD (titled "Smart Dust"), which is slated for release on EMI in Europe. Although I sometimes play with the group over in Germany, I wasn't able to make the mixing sessions, so they asked if I wanted some input into the mastering. Because the music was recorded with different people, using different samples, over different periods of time, the various cuts had differences that needed to be ironed out. As a result, I used lots of different setups, as each tune required its own particular set of mastering tools (the complete story will end up in EQ magazine, and be posted on this site).

However, one tune in particular, with the working title "Abdullah," was a particularly tough case. The bass range was almost overwhelming, as you'd expect from a club mix. The high end was weak in comparison to the bass, and the midrange was in the low end's shadow. Another problem that needed to be addressed was a midrange prominence that gave a sort of "honking" quality.

For this tune, I used four different mastering plug-ins with WaveLab 3.0 arranged in the following order:

  • Pre-multiband compressor EQ
  • Multiband compressorn
  • Post-multiband compressor EQ
  • Loudness maximizer

The first EQ in the chain was the simplest fix, as it simply backed off a bit on the prominent midrange component around 1200 Hz. With that out of the way, the compressor could work more naturally. The second EQ was for some final trimming: about -1 dB of bass shelving starting at 200 Hz, a bit of added "air" (1.5 dB of shelving starting at 9 kHz), and a tiny midrange cut to smooth out the response a bit. The loudness maximizer at the end merely added a 1.5 dB boost to the whole piece to bring up the average level a tad. The main action happened with the multiband compression.



Using a multiband compressor involves isolating the specific frequencies that need work, then processing them with the compressor. With Steinberg's multiband compressor, there are two "panes" in the window: you can adjust each band's width and amplitude in the left pane, and the input/output transfer function (which determines the shape of the compression, expansion, or limiting) in the right pane. When you solo a band, the line representing its transfer function becomes brighter than the others. Clicking on this line provides "break points" which you can drag to alter the transfer function curve.

First, I soloed the lowest, bass band, and adjusted it to include the desired bass range. This meant going as high as I could without picking up the midrange; an upper limit of about 100 Hz sounded right.

Next came the treble frequencies. Because of the weak high end, I initially set the low end of the high range to around 4 kHz. This picked up the brightest frequencies, while again leaving the midrange mostly alone. However, when I then soloed the midrange, the sound was a bit too bright. Moving the treble crossover point down to 2 kHz produced the desired results: The tune's bottom and power was in the bass band, the brightness and sheen in the treble band, and the definition in the midrange band.

Now it was time to address each problem through dynamics processing. Fixing the high end simply involved adding a small amount of compression, but starting at a low threshold. This acted like a treble boost, with the compression bringing up some of the lower-level high frequencies. This compression was so effective at lifting the brightness that it was necessary to bring down the band's amplitude a bit in the left pane.

The midrange needed the same kind of treatment to increase intelligibility;adding the midrange compression brought up some background vocals and instruments which had almost been buried.

Bass was the most interesting situation. The kick/bass combination had an extremely long sustain, so the low end was not well-defined. In some ways this sounded really cool when you cranked the level, but it also obscured the rest of the track.

The fix for this band was to add expansion rather than compression. Signals below about -10 dB were expanded downward to make them softer than normal. Signals above -10 dB were treated more or less normally. This created more peaks and variations in the bass dynamics, but the loudest peaks were just as loud as before. Doing this opened up the whole tune; the bass no longer overwhelmed the rest of the track, but because the peaks were still plenty loud, it didn't sacrifice that all-important club bass sound. However, it was necessary to increase the overall level of this band a bit, because the expansion lowered the average bass level.

Incidentally, on some other tunes I've used expansion in the treble range to tame an overly bright high end. Doing high-end expansion typically allows transients, like a closed hi-hat strike, to come through just fine, which preserves a songís percussive nature. However, lower-level bright sounds fall in level more quickly, so they "get out of the way" of the rest of the song.

Of course, finding these frequencies and calling up just the right amount of compression is time-consuming. But when you finally nail the sound, hit bypass, and confirm that the mastered version slams the original, any time spent seems very worthwhile.