Filters are shape machines

The most common advice for home recording artists is simple: clean up the low end of your mix. You are told to high-pass every track except the kick drum and the sub-bass. You load up an EQ plugin on your vocal, guitar, and synthesizer channels. You apply a steep 24 dB or 48 dB high-pass filter at 120 Hz. But instead of getting cleaner, the mix starts to sound thin and phasey.

Your drums lose their impact, and the vocals sound disconnected. This happens because filters do more than just cut frequencies. They alter the timing of the signals passing through them, and steep curves introduce heavy timing delays.

When you apply a digital filter, you change the frequency response of the signal. But in a standard minimum-phase filter, you cannot change the amplitude of a frequency without also changing its phase.

Steep filter slopes cause phase rotation. Frequencies near the cutoff point are delayed relative to the rest of the signal. This delay smears transients, making the attack of drums or acoustic guitars lose their focus. It pushes the transient back in time, which robs the track of its punch. If you stack these delays across dozens of tracks, the entire mix becomes muddy and loose, even if the spectrum analyzer says the low end is empty.

A filter operates by mixing delayed versions of the input signal back into itself. The steeper the slope, the more delays the filter must apply. A 6 dB per octave filter rotates the phase by 90 degrees at the cutoff frequency. A 12 dB per octave filter rotates it by 180 degrees, and a 24 dB per octave filter rotates it by a full 360 degrees.

This delay is frequency-dependent. The delay is strongest right at the cutoff frequency. The time delay $Delta t$ can be expressed in relation to the phase shift $Delta phi$ and frequency $f$:

$$\Delta t = \frac{\Delta \phi}{2 \pi f}$$

For a steep filter in the low end, this can translate to a delay of several milliseconds for the sub-bass frequencies. This delay separates the fundamental transient from its higher harmonics, making the low end sound soft.

You can hear this transient smear by testing a steep filter on a transient-heavy sound source.

Producers often switch to linear-phase EQs to avoid phase shift. A linear-phase filter shifts the phase of all frequencies equally, which prevents phase smear. But this comes with a physical trade-off.

Linear-phase filters introduce pre-ringing. They generate a quiet, backwards echo before the transient occurs. On drums, this pre-ringing sounds like a soft suction sound right before the hit, which ruins the transient punch.

Another error is filtering out frequencies that are not causing problems. If a vocal has low-end information at 80 Hz, it might be the resonance that makes the voice sound natural. Cutting it off by default makes the singer sound like they are recording in a phone booth.

Choose gentle filter shapes. Use a 6 dB or 12 dB slope high-pass filter for general cleanup. This gradual roll-off keeps phase rotation minimal, preserving transient timing and keeping the mix tight. Reserve steep 24 dB or 48 dB cuts for surgical isolation, such as removing physical room rumble or air conditioning hum. If you must use a high-pass filter on a kick or bass, check the alignment of the transient in mono to ensure the sub-bass has not been delayed.