Linear-phase and minimum-phase EQ

Two ways to build a filter

Every filter changes two things: the amplitude (how much of each frequency passes) and the phase (how much each frequency is delayed).

• Minimum-phase EQ (IIR, like analog EQs and almost all standard EQs): the delay changes around the frequency being touched. Frequencies close to the point of intervention come out slightly shifted in time relative to the rest. There is no latency and the sound is the “classic” EQ sound.
• Linear-phase EQ (FIR): all frequencies are delayed by the same amount, so the waveform is not distorted in time. In return the plugin introduces latency (from a few milliseconds to over 100 ms) and produces pre-ringing: a faint echo that precedes the transients, longer the narrower, deeper and lower in frequency the filter is.
• Natural phase / analog phase (e.g. FabFilter Pro-Q “Natural Phase”): a middle way that reproduces the phase response of an ideal analog EQ while reducing digital artifacts near Nyquist.

What you hear

• The phase rotation of a minimum-phase EQ on a single track is not heard as a flaw: it is the behavior of all historical EQs.
• It becomes a problem when two correlated signals are equalized differently: two microphones on the same instrument (kick in/out, snare top/bottom, DI and amp), a parallel track (parallel compression with a different EQ on the branch), bands separated by a crossover, Mid and Side treated differently. In these cases the sum changes and cancellations can occur.
• The pre-ringing of linear phase is heard on sharp transients (kick, snare, plucks, piano) when using steep filters or narrow cuts in the low end: a “suck-back” before the hit that softens the attack.

When to use each

Linear-phase parameters

• Resolution / latency (Low, Medium, High, Very High, Maximum in Pro-Q): the higher the resolution, the more precise the low and narrow filters, and the longer the latency and the pre-ringing. For mastering with wide EQs, Medium or High is enough; for narrow cuts below 100 Hz you need a high resolution, accepting more pre-ringing.
• Latency compensation (PDC): the DAW handles it on tracks; in FL Studio check that “Use plugin delay compensation” is enabled in the Audio settings. On live tracks or with automation synchronized to the millisecond, the latency can cause problems.
• CPU: high-resolution linear filters are heavy: use them on the master and on buses, not on 40 tracks.

Comparison procedure

1. Duplicate the track, insert the same EQ in minimum-phase mode on one and linear on the other.
2. Invert the polarity of one of the two: what remains is the difference introduced by the phase.
3. Listen to the transients (kick, snare) with both versions soloed, at matched volume. If you hear a pre-echo or a softer attack, the linear mode is too aggressive for that correction.
4. On multi-mic setups: compare the sum of the microphones with the two modes; choose the one that keeps more low end and more attack.

Common mistakes

• Using linear phase “because it is more precise” on every track: latency, CPU and pre-ringing with no benefit.
• Steep HPs (48 dB/oct or brickwall) in linear phase on the master: pre-ringing on the sub and a blunted kick.
• Different minimum-phase EQs on the two branches of a parallel compression: the sum loses body around the frequency being touched.
• Linear-phase EQ automation with high latency on parts that need to be time-accurate.
• Forgetting that the filters of a multiband crossover also create phase rotation: use the linear mode of the multiband on the master when transparency matters.