Audio Equalization: Filter Types and Voice Frequencies

The equalizer is the most used tool in any mix, and also the one with which it is easiest to do damage. It lets you raise or lower the level of specific frequency regions, fixing flaws or shaping tone. To use it well you need to know the filter types, the three main parameters and the map of frequencies you work on most often.

The fundamental parameters

A parametric equalizer, the most common type in DAWs, offers three controls per band:

• Frequency: the center of the region you are working on, in hertz.
• Gain: how much you boost or cut, in decibels.
• Q (or bandwidth): how narrow the affected region is. A high Q acts on a thin slice; a low Q on a wide area.

Filter types

High-pass and low-pass

A high-pass filter lets frequencies above a threshold through and progressively attenuates those below. It is the most useful filter for voice: it removes rumble, desk vibrations and traffic noise that contribute nothing to the tone. The slope is expressed in dB per octave: 12 dB/octave is gentle, 24 dB/octave steeper.

A low-pass does the opposite and is useful for removing very high hiss or darkening a background sound.

Shelving

A shelf filter raises or lowers all frequencies above (high shelf) or below (low shelf) a certain point, like a step. A gentle high shelf above 8-10 kHz adds "air" to a voice; a low shelf can reduce excess bass without removing it entirely.

Bell (peak)

A bell filter acts around a center frequency, with a shape that depends on Q. It is the tool for targeted corrections: removing a room resonance, reducing harshness, adding presence.

Notch

A notch is a very narrow, deep bell used to eliminate a precise tone, such as a hum or a whistle, while touching the rest as little as possible.

Voice frequency map

These regions vary a lot from voice to voice: a deep male voice has its body lower than a female or child's voice. The frequencies are starting points for searching, not recipes.

The sweep method

To find an annoying resonance there is a simple technique:

1. Create a narrow bell, with high Q, and boost the gain by 8-10 dB.
2. Sweep it slowly across the frequencies while listening.
3. When the problem becomes obvious and almost unbearable, you have found the frequency.
4. Reverse the gain: cut by 2-5 dB, widening the Q a little.

Use it sparingly: with a big boost every frequency sounds like a problem, and you risk "carving" the voice in too many places.

Principles of good use

• Cut before you boost: a voice often seems to lack presence not because highs are missing, but because there is too much low-mid. Cutting at 300 Hz may work better than boosting at 4 kHz.
• Small moves: corrections of 2-3 dB are already audible. Moves beyond 6 dB often point to a recording problem.
• Listen in context: a voice EQ'd in solo can seem perfect and disappear under the music. Decisions should be made listening to the whole mix.
• Compare at equal volume: boosting frequencies raises overall level, and a louder sound always seems better. Use the output control to match levels when comparing with the original.

Linear-phase equalizers

Ordinary equalizers introduce small phase shifts around the frequencies they modify. Linear-phase equalizers avoid this effect, but add latency and can produce slight pre-echo on transients with very steep filters. For voice and podcasts a traditional equalizer is almost always the right choice.

In my work on spoken voice I almost always use the same sequence: high-pass, a slight cut in the "box" region if the room calls for it, a touch of presence only if needed, and a de-esser instead of a fixed cut on sibilance.

In short

• Frequency, gain and Q are the three parameters of every band.
• The high-pass is the most useful filter for voice; shelves and bells shape the tone.
• Voice regions are reference points to start from, not fixed recipes.
• Cut before boosting and prefer moves of a few decibels.
• Always compare at equal volume and listen in the context of the mix.