by Bruce Bartlett Copyright 2011
To make CDs “hot” or loud, it’s routine to apply peak limiting and normalization. The peak limiting knocks down the level of the transient peaks — leaving the average level untouched. The normalization brings the limited peak levels up to nearly 0 dBFS.
For example, suppose the peaks are limited (reduced in level) by 7 dB, and they are normalized back to 0 dBFS. That’s a gain increase of 7 dB. So the average level also goes up 7 dB, and the mix sounds 7 dB louder because loudness is determined by average level rather than peak level.
Not all limiters sound the same. Some have a slow release time, so they drop the gain of the average-level program during a transient and slowly bring it back. It sounds like the average level ducks down during transients.
Other limiters are more transparent. They respond so quickly, you can hardly hear them working. But all limiters reduce the peak level of transients and reduce transient clarity and focus.
I wanted to compare the sound of four peak limiter/normalizers, also called maximizers. I started with a clean, unprocessed recording of a blues with strong drum hits. Those hits were not compressed or limited.
I inserted one limiter plug-in at a time into the track of the original recording. In each limiter I set the gain boost to 10 dB, peak level reduction 7 dB maximum, and output normalized to -0.3 dBFS to prevent clipping. Next I dropped the gain 10 dB on all the limited/normalized wave files to make them as loud as the unprocessed mix.
A sample of the resulting waveforms is shown below. Note how the maximizer reduces the peak level of the drum hits. They are not clipped.
Left: original signal
Middle: after maximizing
Right: maximized signal -10dB
Below you can click on several mp3 files to compare the sound of the four maximizers. Again, you are hearing the maximized signal after dropping its level by 10 dB.
For quick download on the Web, all the wave files were converted to 320 kbps mp3 files. Doing that smears the transients slightly, but you can still compare the results.
Unprocessed.mp3 This is the original unprocessed mix. Note the sharp clarity of the drum hits.
Boost11.mp3 Cakewalk Sonar Boost 11 maximizer.
Tlsmax.mp3 TLs Maximizer (freeware).
Voxengo.mp3 Voxengo Elephant 2.5 mastering limiter. Limiting mode EL-3. Speed: slow. 20 Hz DC filter.
Ardaz.mp3 Ardaz Maximizer with 20 Hz DC filter.
Although these results show how some maximizers are more transparent than others, you might not even want to use a transparent maximizer. It depends on the effect you want to create.
# # #
AES and Syn Aud Con member Bruce Bartlett is a recording engineer, microphone engineer (www.bartlettaudio.com) and audio journalist. His latest books are “Practical Recording Techniques 6th Edition” and “Recording Music On Location”.