I'm With The

Probably the most common example is multiband compression, which typically splits a signal into three to five bands. This is used a lot with drums and program material (e.g., a mixed stereo signal), because single-band compression has limitations. For example, with drums, suppose you feed the drums through a compressor and there’s a strong kick hit. That causes compression to occur, and it affects the high-frequency sounds as well, such as hi-hats. With multiband compression, you could put drastic compression on the low-frequency kick drum, while adding only a slight amount of compression (or none) to other frequency ranges. When used with program material, multiband compression gives a less “squashed,” more transparent sound.

Other Uses

Because a keyboard covers such a wide frequency range, you might want to process one frequency range differently. For example, suppose you’ve loaded a cool imitation-Rhodes preset, and you want the lower notes to “bark” more than the upper notes. Apply multiband distortion, and apply subtle amounts of distortion only to the lower frequencies — say, below 200Hz. Now those lower keys will have a gritty character, while the others sound normal.

With vocals, I often split the vocal into high and low frequencies, then add delay only to the higher frequencies. Thus the sibilant-rich sounds have echo, while the lower notes do not. If there’s a fairly dense instrumental background, this technique avoids muddying up the midrange — the vocal echoes “float” over the other sounds.

How To Do It

Some processors are designed for multiband operation, such as Wave Arts’ MultiDynamics. Steinberg’s Quadrafuzz (a virtual version of a hardware processor I designed in the ’80s) provides multiband distortion, and iZotope’s Trash takes multiband distortion to an extreme. But if you want to do multiband processing with a processor that’s inherently single-band, you have some work ahead.

The simplest hardware method is to use a crossover; but with computer-based recording, your options are limited. The only crossover plug-in I’ve seen is part of NI’s Guitar Rig 2, which allows splitting a signal within the rig. The problem with creating a general-purpose VST/AU/DX crossover plug-in is that you can’t plug it in to one track, but have the outputs appear on several different tracks.

There are two workarounds. BIAS Vbox is basically a plug-in that hosts plug-ins; you would split the signal into as many parallel signal chains as desired, add appropriate filtering at the beginning of each chain to create the bands (more on this later), insert signal processors in the chains, then sum their outs in Vbox.

The other option is to clone a track for as many times as you want bands. Insert an EQ plug-in so each track covers a specific frequency band, then insert your plug-ins into the tracks. You’ll probably want to feed the track outs to a subgroup, so you can use the subgroup level to bring the level of all bands up or down at once.

The most difficult part is setting the EQs to split the bands evenly so that there’s no overlap. This becomes more complex with more bands, as you’ll be using a combination of highpass, lowpass, and bandpass filters. However, you can’t really use a conventional parametric bandpass filter, as its slopes don’t roll off to 0 level. Instead, make up a bandpass filter by using a combination of highpass and lowpass filtering to isolate a specific band of frequencies (Figure 1).

After splitting the bands, mix them together and listen to the unprocessed output to make sure you haven’t introduced any unintended frequency response peaks or dips. It’s not easy to tweak these parameters properly to get a good split — but when you do, the world of multiband processing will open up to you.