The View from 2005: This is a place where technology has really changed things for me, and some insights gained from Tom Bates have lead me to a very powerful new approach to EQ. See my article from 2004 for more information about that. However, all of what I’ve said here still goes – the verities continue to hold true. There is an awful lot of useful information here. Enjoy. Then go onto my article about Tom Bates, written in 2004.

Overview

A couple of months back I wrote a couple of articles called Spectral Management , which were about the nature of the audio spectrum and the underlying principles of equalization as a way to manipulate tone quality or timbre. Now, I’d like to look a little more deeply into the use of parametric equalization to treat voice. bass and drum tracks. It is a little difficult to do this without specific audio examples, but I hope you’ll pick up enough of the idea to permit you to develop your own techniques and skills using parametric equalization.

First, a comment about industrial-strength creative equalization. Studying such eq can be very tiring and confusing, because you usually spend an inordinate amount of time listening to very small snippets of sound and micro-details of timbre, taken out of context. You will find, as you get really into it, that you will have a tendency to ever-eq everything – in fact, I think you should freely over-equalize as a starting point. So, as you first study this, watch out for fatigue, confusion and problems with levels, and be mentally prepared to cut way back on equalization as you progress. In the end, you’ll find, a little equalization will go a long way.

About Parametric EQ

Traditional analog on-console eq usually limits your choices to a groups of eq bands, each with a selectable array of center frequencies and variable boost or cut, usually to plus or minus 18 dB. The notion behind parametric equalization is to permit a greater range of possibilities regarding spectra. Not only do you have variable boost/cut over a +/- 18 dB range, but you usually have continuously variable frequency selection over several octaves of the audio spectrum (for any given band of the equalizer) and control over the width of the band being boosted or cut. The benefit of parametric equalization is, of course, that you have lot more freedom and flexibility to tailor the response curve to precisely fit your needs. The cost is that you have a lot more decisions to make, which is what makes it all so confusing and tiring.

Most attributes of equalization are pretty intuitive and well understood by musicians, producers and beginning recordists. The most confusing issues have to do with the frequency of musical spectra vs. frequency of pitch (they aren’t the same) and management of levels: boosting a band of frequencies increases the overall signal level significantly while cutting a band of frequencies doesn’t change the overall signal level much at all. The other major difficulty comes with the ability to control the equalized bandwidth (sometimes called “Q”) which is being boosted or cut. This quality, bandwidth, is the most distinctive feature of parametric eq and it is also the one that is hardest to hear out and to understand in the heat of mixdown.

Bandwidth

Q, or bandwidth, may be expressed via a single number (i.e. “The EQ has a Q of 50”) or in octaves (“The bandwidth is about 1/6th of an octave.”). The formal definition of Q has to do with the ratio between the center frequency (i.e. the frequency that is boosted the most) of a resonant circuit and the range between frequencies on either side of that frequency that are at half power (i.e. -3 dB from the center frequency),

	The definition of Q. If the center frequency is 1000 Hz. and the side frequencies are 700 and 1300 Hz., then the Q is 1000 divided by (1300-700), or 1.67. The bandwidth is a little less than an octave.

so that an equalizer tuned to 1,000 Hz. that is 3 dB down at 300 and 1700 (the range is 1400 Hertz) has a Q of .7 (i.e. 1000/1400) or approximately 3 octaves. If the same equalizer were 3 dB down at 950 and 1050 hertz., the Q would be 10 (1000/100) or approximately 1/6th of an octave.

	Low Q, wide bandwidth. If the center frequency is 1000 Hz. and the side frequencies are 300 and 1700 Hz., then the Q is 1000 divided by (1700-300), or .7. The bandwidth is a little less than an three octaves.

	High Q, narrow bandwidth. If the center frequency is 1000 Hz. and the side frequencies are 950 and 1050 Hz., then the Q is 1000 divided by (1050-950), or 10. The bandwidth is a little less than 1/6 octave.

There is actually a good bit more to this, but we’ll leave the details to the propeller heads for the time being. What you need to really glom onto is the notion that high Q represents single frequencies or very narrow frequency regions, and therefore single harmonic components of a sound, while low Q represents very broad bands of spectra. Even fairly large changes in amplitude at high Q can be fairly hard to hear, while small amplitude changes at low Q are usually quite easy to hear.

Formants - ya gotta understand formants to get into this

With that in mind, lets look at the acoustical behavior of sound sources, oh say an acoustic guitar for instance. Guy/gal plucks a string, sound comes out. Now, there are several parts to this. The string vibrates at a bunch of (mostly) harmonically related frequencies (called the fundamental and overtones). So, the string itself may be thought of as a bank of extremely high Q filters that convert broadband noise energy (the plucking) into a tuned set of harmonically related frequencies (the musical note). This set of frequencies is then sent, via the bridge, to the body of the guitar, which serves as a mechanical transformer that converts the mechanical motion of the string into mechanical motion over the surface of the body of the guitar, which is then emitted acoustically into the surrounding environment, through which it makes its way via multiple paths to the ear of the listener. Whew! How do you like that sentence?

The hollow chamber of the guitar acts as a broadband, low Q filter, and the physical guitar body itself also vibrates in a band of medium Q frequency sets that are related to the thickness, density and tension of the wood, etc. Thus the behavior of the guitar body and chamber can be thought of as a fixed filter set (like a 10-band equalizer) that has some Low Q and some medium Q boosts (and cuts) that are applied to the high Q vibrating string. Interestingly, the high Q filter set that is the guitar string is frequency-shifted up and down by changing the length of the string, which is usually done by a finger-tip on the left hand pressing the string against different frets. Musicians call this playing a melody! Equally interestingly, while the high Q resonances of the string shift up and down, the Low Q and medium Q filters of the chamber/body remain constant, so that the resulting total timbre of the sound changes from pitch to pitch, string to string, etc. Wheeze! I’m really outta shape for this kinda writing!

The response curve of the fixed-frequency filter-set we’re calling “chamber and body of the guitar” is usually referred to as the formant of the instrument. And this is why I’ve been wading through all this: when we use an equalizer to modify the sound of an instrument, it is the formant of the instrument that we are modifying, rather than the spectra of the vibrating strings, which move up and down as a function of the changing notes being played. And this is important, because of the way that we hear and identify timbre.

The psychoacoustics of it all

It turns out that when we listen to a sound, we identify its timbre by the fixed character of its formant, not the harmonic structure generated by the vibrating strings. We identify the individual guitar or voice, not the note being performed (as in “That guitar really sounds good” rather than “I know I’ve heard that A-flat somewhere.”) The waveshape created by the vibrating element is comparatively unimportant to our hearing system for the identification of timbre.

At the same time, our hearing system has some interesting quirks that are relevant to this discussion. We detect spectrum in bands approximately 1/3 of an octave wide (a Q of about 6). These are called critical bands and they define the range within which we can discriminate individual frequencies. Because of the limits of these bands, 1/3 octave equalization is generally all we need for modifying timbre. However, the parametric equalizer’s ability to create narrower bandwidths (1/12 of an octave, or a Q of 20) still comes in very handy, as we shall see.

As I mentioned above, Floyd Toole has published some interesting data which indicates that our hearing is comparatively insensitive to extremely narrow, high Q resonances, while extremely sensitive to broad, low Q resonances. What this means is that broadband low Q changes will have a big effect on the tone quality, while high Q changes will have comparatively little effect. There is, however, an exception to this. When the high Q equalizer is tuned to a specific frequency at which a musical overtone is occurring in the recording, that tone will jump out and become highly audible. Ditto for resonant frequencies in the reverberation, natural or electronic. So the presence of significant energy at a specific frequency will become very audible when it is boosted with high Q. When cutting high Q settings, however, the effect of the eq simply isn’t audible.

This means that there are two ways to approach the handling of Q:

• The first is to start with a very low Q tuned to the approximate frequency range you think you want to change, and then to very gently boost or cut until you think you’ve got approximately the right tonal color, and then finish off by trimming the bandwidth down to just the width you want.
• The second approach is to start with a high Q frequency band boosted grossly (even if you ultimately want to cut it) and then sweep across the frequency range until you find the exact frequency that your ear to says to change. Once you’ve located it, then start trimming the amount of boost (or start cutting if that’s what you want to do), and at the same time start increasing the Q (increase the bandwidth) until you’ve got just the right timbral quality.

Personally, I prefer the latter technique for working on individual tracks and the former for more generalized eq on submixes or stereo recordings.