With lowpass or highpass filters, the approximate frequency above or below where signal attenuation begins.
Passes frequencies below a particular cutoff frequency. Above that frequency, response diminishes progressively with higher frequencies. This response is most useful for reducing hiss and excessive brightness on individual tracks. Using it on program material is seldom recommended, because the way it attenuates high frequencies can “muffle” the sound. A shelving response (see later) tends to be more “musical.”
Passes frequencies above a particular cutoff frequency; response diminishes progressively with lower frequencies. The low-cut switches found on consoles, which generally tailor mic response, are highpass filters. This is useful because with “plosive” (“P,” “B”) sounds, mics can produce significant low-frequency energy. The more low frequencies you can cut below the range of the material being recorded, the better. Highpass filters with very low cutoff frequencies (e.g., 10–20Hz) are common for removing subsonic energy and/or DC offset from audio that would otherwise use up available bandwidth. Except for DJ uses, highpass filters are seldom used to change the “tone” of program material.
Passes frequencies around a selected resonant frequency, with response diminishing progressively both above and below this frequency. This is different from a parametric filter (see later) where at some point above and below the resonant frequency, the response flattens out. The most well-known bandpass filter application is the wah-wa pedal. Bandpass filters are also essential for “telephone voice” effects.
The shelving response occurs at the high and/or low ends of the frequency spectrum, and boosts or cuts. The process gets its name because the response first rises or falls, then flattens, like a shelf. This response is useful for general tone-shaping on individual tracks or program material. Also, a little high-frequency shelving boost can help give a track more definition and “snap.” Shelving is a broad, gentle type of equalization effect that if used properly, can create a subtle yet significant improvement.
A parametric equalization stage has a variable resonant frequency, boost/cut amount (typically ±12 to ±18dB), and width (the range over which signals are boosted or cut). This is considered the most flexible equalizer stage, yet practically speaking, it can be less useful in some applications than shelving or other modes. The parametric excels as an insert effect for individual channels, as you can boost or cut very specific parts of the signal to solve problems. For example, if one tom in a drum part seems too loud, you can zero in on just that frequency and reduce it a bit. However, using parametrics with program material requires restraint. Generally, broad, gentle changes work best. A little upper midrange boost can add intelligibility to solo instruments and vocals, while a slight dip in the 300–500Hz range can reduce “muddiness” and give a track a crisper sound.
A notch filter is usually a variation on a parametric, except that the width is extremely narrow, the response only cuts (not boosts), and the cut can be –60dB or more. The classic notch application is to cut out a tiny slice of signal to get rid of 50 or 60Hz hum.
Some filter modes are simply one of the types above with a different number of poles. The term comes from hardware filters, whose basis is often a resistor/capacitor combination that “tunes” the filter and provides one pole of filtering. However, analog filters don’t just stop the signal past a certain point; instead, the response rolls off gently. A basic 1-pole analog filter rolls off at 6dB/octave. In other words, the response is down 6dB an octave above the cutoff frequency, 12dB two octaves above the cutoff frequency, etc. Sharper rolloffs require cascading multiple stages. A 2-pole filter combines two 1-pole filters and rolls off at 12dB/octave, a 3-pole filter at 18dB/octave, and a 4-pole at 24dB/octave. Cutoffs steeper than 36dB/octave are very difficult to implement with analog circuit designs, but are feasible with digital technology. Lowpass, bandpass, and highpass filters can all be multi-pole types.
There are several types, but the most common is a lowpass filter with a resonance control. This adds a resonance peak in the vicinity of the cutoff frequency. You can use this peak to boost the response at a specific frequency, but this precludes sweeping the filter, as the boost’s frequency will change.
Gets its name from the response curve, which has so many dips it looks like a comb. The flanging process creates a comb filter, then sweeps the frequencies of these peaks.
Designed to mimic the filtering characteristics of the human mouth, format filters impart vowel- or vocal-type sounds onto a signal. They typically use filters with multiple resonant peaks.
These don’t roll off the signal gently, but rather, the signal “falls off a cliff”—a brickwall rolloff of 90dB/octave is not uncommon. These are used primarily at the input of digital systems to prevent signals from interfering with the clock frequency, and at the output to filter out any clock signal from the audio (as well as smooth out stairsteps in the output waveform resulting from the digital-to-analog conversion process). It is extremely difficult to design analog brickwall filters, and although it’s easier to design digital types, making them sound good is not easy.