Adventures in Frequency-Conscious CompressionLike most other people, engineers are creatures of habit and convention. When we think of frequency-conscious compression, the first thing that pops into our minds is de-essing vocals. Not surprising, as this is arguably the most common application for the technique. But there are many others, and if we think outside the box, we can discover ways to use frequency-conscious compression to tame boomy acoustic and electric bass guitars and cello, blaring vocals and ear-splitting wind instruments. Much of this is familiar to veteran engineers, but let's look at some refinements that should at the very least jar your memory and tune up your chops.
First, a brief refresher for newcomers to the technique. To perform frequency-conscious compression, you'll need to patch an equalizer into the sidechain of your compressor; of course, the track you want to process gets patched through the audio path via the I/O. While listening to the audio track, determine what frequencies you want to tame. Then simply boost those frequencies on your equalizer to make the compressor more sensitive to them. For example, you might boost 7 kHz to 13 kHz by 6 to 12 dB in order to de-ess a female vocalist. When searing sibilance rears its ugly head, the audio in the compressor's sidechain - which is "hearing" an exaggerated response in this frequency band, courtesy of the EQ boost - will exceed the threshold and make the compressor reduce the level of the audio signal.
To be sure, the compressor attenuates not only the high frequencies, but the whole enchilada. The trick is to set up the compressor so that it releases and returns the track to unity gain immediately after the sibilance has ceased. Usually, an attack time of around 50 musec (.05 ms) and a release time of 50 to 60 ms gets the compressor in and out fast enough to attenuate the sibilant portion, yet passing the rest of the vocal through unchanged.
The quality of the equalizer is not critical for this purpose, as it is not in the audio path. For de-essing male or female vocals, I typically start by boosting all frequencies above 5 kHz. Such a broad-stroke approach is usually okay, because there's generally not enough energy above the sibilance band to unintentionally trigger the compressor, even with 12 dB of equalizer boost. This way, I don't need to fuss with the high-frequency cutoff, and I know any and all sibilance will be nabbed. In addition, I usually cut all frequencies below 5 kHz to decrease the compressor's sensitivity to those lower frequencies so that it does not kick in when the spectral balance is okay - such as during vowel sounds and nonfricative consonants.
As for compression ratios, I usually start with u:1 to really hear the effect and then back it off to where it sounds right. I might even go as low as 3:1 to perform subtle de-essing on a track that just needs a pinch of control.
BLARING VOCALSSooner or later, every engineer is faced with a vocalist who sounds like a trumpet. These singers usually sing primarily from the throat, rather than use chest or head resonances. You try everything to warm up the sound - proximity effect, tube mic, tube pre, tube compressor or even a Kenner Easy-Bake Oven - but such vocalists still sound like an ear-splitting siren on high notes. The solution? Frequency-conscious compression.
For this application, I generally boost somewhere in the 2.5 to 4 kHz range on the sidechain equalizer, cutting other frequencies as needed to further sensitize the compressor to the desired band. The affected band may, in fact, need surgical treatment, rather than a broad-stroke approach. For instance, you may need to boost 2.5 kHz by 6 dB, boost 3.15 kHz by 12 dB and leave 4kHz at unity. It all depends on the singer. In any case, the compression ratio usually needs to be set to a more moderate value than one would set for de-essing, typically in the 4:1 neighborhood.
The above example illustrates the superiority of using a 11/43-octave graphic equalizer in the compressor's sidechain. Although parametric and shelving filters will do the job for many frequency-conscious compression applications, I prefer having 31 bands for those situations where multiple frequencies and slopes need to be shaped in the sidechain. This allows the compressor to do multiple tasks, such as keeping a lid on low highs and de-essing, simultaneously.
To set up for this latter application, I might slightly boost the 2.5 to 4 kHz band (keeping in mind the high ratio I'm using for de-essing purposes), cut all else below 2.5 kHz by 12 dB and boost heavily from 5 kHz on up. Set this way, the compressor de-esses vigorously during sibilant portions of the performance and reduces peaks a bit during hard "a" vocal sounds (such as during the ubiquitous rock exhortation, "ba-a-a-a-a-by!"). A variation on this setup would be to keep the 2.5 to 4kHz band at unity gain on the equalizer, making the compressor act like a non-sidechained limiter on these frequencies only. Just keep in mind that the release time must be kept very short in deference to the de-essing also being performed - unless you want to give the singer a lisp!
WIND INSTRUMENTSSmall wind instruments can also sometimes benefit from frequency-conscious compression. Pennywhistle can sound incredibly piercing, especially in difficult multitrack ensemble situations where it may need to be miked somewhat closely. In this case, I often boost the compressor's sidechain EQ in the neighborhood of 1.6 to 2.5 kHz. The ratio generally ends up being around 10:1, the threshold set high, the attack around .5 ms and the release roughly 200 ms.
Frequency-conscious compression can also be put to good use on effect sends. I once noticed that my Lexicon PCM 70's input level has a tendency to momentarily spike 12 to 18 dB when a certain client's wood flute is sent out to it. To keep the reverb's input level consistently high for the best resolution, I placed a compressor before the PCM 70's input. Patching an equalizer into the compressor's sidechain, I boosted the nasty midrange frequencies (315 to 500 Hz, in this case) that were causing the spike and robbing me of headroom. Problem solved.
DEBOOMINGDebooming is the basement buddy of de-essing, standing guard over bass frequencies rather than highs. Electric bass guitar can be deboomed to put a lid on isolated, boomy notes. The offending frequency is usually the fundamental of the boomy note being played. For debooming duties, set the compressor's attack time to less than 1 ms. The release time should usually be set for no faster than 1/10 ms to avoid distorting the waveform. (At quicker release time settings, the compressor's envelope gets superimposed on a single cycle of the longer wavelengths of bass frequencies, often causing audible distortion.)
The same technique works for cello, but for a slightly different reason. Many times, it is not the fundamental but rather the formant of the cello that is causing the boominess. No matter. The formant is being excited by whatever fundamental pitch is being produced at the time you hear the boominess. Remember, unlike with splitband compression, you are not attenuating select frequencies. You only need a trigger - any trigger - to compress the entire signal when the problem occurs. Boosting the fundamental in the sidechain equalizer will result in the compressor kicking in whenever the boomy note is played, reducing the excited formant that occurs along with it. Fundamental pitches are easy to identify quickly. If you'd rather go hunting, however, you can sensitize the compressor's sidechain to the formant, and the results will usually be roughly the same.
Up to this point, we've only discussed the processing of monophonic or melody lines. Debooming acoustic guitar can be a whole other ball game, because broadband spectral content is involved. That is, it's difficult to attenuate boomy bass notes without pulling down the level of higher strings that are ringing simultaneously. This is made all the more difficult because of the slower release times that debooming requires. Lower ratios are a must when debooming acoustic guitar. I generally find that 1 to 3 dB of gain reduction is all you can get away with before the compression starts punching "holes" in the track.
Likewise, the broad spectral (and often percussive) content of stereo program mixes presents an insurmountable challenge to frequency-conscious compression. For such tasks, as well as for debooming the most stubborn acoustic guitar tracks, your best recourse is to use splitband compression. But that's a subject we'll have to leave for another article!