A simple approach to mixing a rock song: Part 5

By mic on 7/6/2007

To be sure, contemporary home studio equalization uses digital equalizers. A couple of years ago, when I was trying to design a couple of equalizers, I finally understood what they actually do, so perhaps before we go into how we use equalization we should talk a bit about digital equalizers.

As I wrote in my previous post, the sound of a track probably contains a very complex sound wave. In other words, there are many frequencies that may exist in a single sound. The problem then is that it is difficult to tell what frequencies are actually contained in a sound wave. An equalizer must change the amplitude of certain frequencies without changing others and so an equalizer must "know" what frequencies are actually contained in a sound and what the frequencies’ amplitudes before equalization are. Alternatively, even if the equalizer does not know explicitly (by value) what frequencies are contained in a sound, the equalizer must at least know how to isolate certain frequencies.

In the digital world, there are a couple of ways to work with frequencies: one can use Fourier transforms to decompose a complex sound wave into separate frequency pieces or one can use digital filters to increase or decrease the amplitudes of ranges of frequencies. Оne way or another, this is a very computationally intensive process, which means that your digital signal processor must do a lot of computations. Тhe only reason this is important is that it makes difficult making digital equalizers work during runtime, without first processing the underlying track. That makes experimentation a very slow process.

The more important thing is that all computations are imprecise. In the range of human hearing (say, 20 Hz to 20 KHz) there are infinitely many frequencies. The Fourier transform cannot decompose a sound into infinitely many pieces, but must instead work in small steps. If you work in the steps of 20 Hz, 25 Hz, 30 Hz, and so on, you will be missing 21 Hz, 22 Hz, and so on (this is just an example). Digital filters are also imprecise. If you attempt to filter out frequencies between 5 KHz and 6 KHz, you may end up getting a good handle on 5.5 KHz, you will filter out 4 KHz completely, but you may end up not filtering out 4.9 KHz completely (depending on how you design the filter), and you may end up unintentionally boosting up 5.1 KHz. This means that you can even run your sound through an equalizer in which all bands are set to 0 dB, and still end up with something different than your original sound. Moreover, the more precise an equalizer is, the more computationally intensive it usually is. This means that it is even more difficult to make a useful real time equalizer.

In short:

  1. Digital equalizers are not precise; and
  2. Digital equalizers will mangle your sound to some extent.

I am guessing that problems (1) and (2) above apply to analog as well as digital equalizers, but I cannot really be sure. I have never designed an analog equalizer. This should not discourage you from using equalizers, however. A properly designed equalizer may just do what you need it to do. But what should that be? I, for example, have only really been successful in using equalization to make minor adjustments to tracks. In general, if a track is really off it is better to just re-record it. Below are some examples of where we have used equalization.

We tend to equalize muddy vocals by adding anywhere from 2 to 5 dB above 4 KHz. As per my previous post this means that we probably add a bit of presence to the vocals (frequencies around 4 KHz) and a bit of sibilance (6 KHz to 10 KHz). We have occasionally tried to open my vocals by also dropping 2 dB at around 100 Hz and around 500 Hz in addition to adding presence and sibilance. Occasionally we have tried to add a bottom end to my vocals by adding anywhere up to 4 dB at around 270 Hz to 300 Hz. The last couple of things are probably specific to my vocals.

Acoustic or electric guitars actually work similarly to vocals, but you can actually be more drastic as changes to timbre are less noticeable than when working with vocals. I actually normally equalize steel string acoustic guitars up above 5 KHz (an after fact of using the SM57). Nylon string acoustic guitars are actually more interesting as they have bell like overtones at these frequencies and equalizing them up gives them a different ring. Electric guitars, since they are a lot more processed, actually change timbre significantly, but that just means that we end up with a different sounding guitar, which may be a good thing.

Recently we recorded drums on an old drum set on which the crash was broken and the hihat was old, the end result of which was that both sounded thick and the high end of the song was basically missing. We added 2 dB above 5 KHz to both tracks to add some sibilance and shimmer.

About the most drastic changes that we have ever made was to lose some of the brightness of the bass, which you can usually do if you drop a lot of dBs above 1 KHz. We have gone as far as dropping 10 dB, which also tends to remove the attack of the pick, but in general that is something we can live with.

We stumbled onto all of the above examples by chance. I used to open the CoolEdit 30-band equalizer and pull sliders up and down until I got the sound that I wanted. This was before I started designing a couple of equalizers and learned a few things about where instruments actually lie. We still experiment a lot, but now, knowing what we know, our experiments are more targeted. The two most important things I have learned so far from all of this is that first, digital equalization should be well controlled and not drastic to avoid major changes in the instrument’s timbre, and two, equalization should control overtone / undertone qualities of the sound to properly place an instrument within the mix.

authors: mic


Copyright 2006 by Kaliopa Publishing, LLC