That Vocal Sound

The human voice is a fascinating instrument. The physics of how sound resonates in the larynx and vocal cords and projects through the throat and mouth, combined with the resonances of the chest and sinuses, creates a staggeringly rich and complex signal. And every last one of these instruments is different from the next.

Now, factor in the dimensions and reflectivity of the room or booth the singer is tracked in, the sonic imprint of the microphone, the singer's positioning on the mic and the microphones polar response --nope, we're not done yet. Don't forget the fingerprint of the mic preamplifier, the behavior of any vocal compression, the tone of the AD converters or tape media... No two vocal sounds are going to be the same.

For sake of discussion, let's say there are two starting points when mixing vocals. The first is, "the recorded vocal already sounds good." This means the tracking engineer (which might be you!) did his job well, and you're trying to preserve a good thing. In this scenario, you're in "Do No Harm" mode, and will use far less processing.

The second starting point is, "The vocal doesn't sound good (yet)"... and therefore needs some help. In this case, the complex system described above introduced some undesirable components to the vocal signal that we want to clean up. Hopefully, we can do some quick surgery, and then find ourselves back in "Sounds Good" mode.

Personally, if I'm trying to clean up a vocal, I think about removing the unwanted artifacts (resonances) from the signal. Generally, I'll use several narrow cuts with a parametric EQ to accomplish this. Perhaps skewer out some ~220-250Hz to remove proximity effect, cut something in the 300-500Hz range to defeat room/booth tone buildup, perhaps notch out something around 700-900Hz that might be sinus or head resonance. This is clean-up, so I'll do this prior to any additional processing (compression, effects, etc). As always, it's essential to use your ears to figure out what's needed.

A quick note on subtractive EQing: A great technique for finding troublesome frequencies to notch out with a parametric EQ is to select a narrow Q (bandwidth), boost heavily, and sweep the frequency of the EQ band to find the ugly component of the sound. Then, keeping the narrow Q, cut that frequency to taste (could be a 3-12dB cut). You may then wish to adjust the Q as needed to make sure you're scooping out as much or as little as desired. You'll discover there's a trade-off between the depth of the cut and the bandwidth. Experiment to see which suits the material better.

 Vocal tracks are frequently compressed to tape. And heavy subractive EQing can rob a compressed signal of fullness and life. (Listen for this; it has a sound.) So be careful what and how much you're cutting, and always A/B your EQ'd track to the original by flipping the EQ in & out of bypass. Did you make it sound better?

If we've succeeded in EQing out some of the "nasties" from the vocal, we can determine if additional processing is needed. If the mix is dense, or we want the vocal to be very up front and "in your face", some compression may help. We'll talk about this shortly...