What Your EQ Is Doing to Time: Minimum Phase, Linear Phase, and the Artifacts Nobody Talks About
Every EQ makes a decision about time. Most engineers never think about it in those terms because most EQ decisions feel like frequency decisions: boost here, cut there, done. But every filter has a time-domain consequence, and in mastering, where the material is finished and the margins are small, those consequences show up. Sometimes subtly. Sometimes with a client in the room asking what that flutter is.
This is the piece I wish existed when I first started asking these questions.
What Phase Actually Is
A filter does not just change the amplitude of frequencies. It also changes their phase relationships and effective timing. Run a complex signal through an EQ and the frequencies you boosted or cut emerge with altered phase relationships and slightly different effective timing than they would have without the filter. That phase behavior, measured in degrees, is an unavoidable consequence of how filters work.
This is not a flaw. It is physics. The question is not whether your EQ introduces phase shift, because it does, but what kind of phase shift and what the audible consequences are.
Minimum Phase: What Every Analog EQ Does
Virtually every conventional analog EQ ever made is minimum phase. The MEA-2, the ITI MEP-130, the Manley Mastering Stereo Pultec, the API 5500. All of them. This is not a design choice in the sense of something that could be otherwise. It is a consequence of building filters from passive and active analog components. The physics of capacitors and inductors produce minimum phase behavior inherently.
What minimum phase means precisely: the phase response is the minimum amount of phase shift required to produce a given amplitude response. The two are mathematically linked. You cannot have one without the other. Boost 100Hz by 3dB and you will get a predictable phase rotation centered around 100Hz. The shape of that rotation is determined by the shape of the amplitude curve. Change the Q, change the boost amount, change the center frequency, and the phase response changes accordingly, always in the minimum phase response compatible with that amplitude curve.
The phase shift from a minimum phase EQ is also causal: it happens after the signal, not before it. Energy is delayed, not anticipated. This is the natural behavior of physical systems. A guitar string does not vibrate before you pluck it.
The diagram above shows a 4dB boost at 200Hz on a minimum phase EQ. The top curve is the amplitude response. The bottom curve is the phase response. Notice that the phase shift is centered in the same processed frequency region as the boost, and that its shape is mathematically tied to the filter response. The phase rotation tracks the EQ move rather than existing independently of it. This is the characteristic shape of minimum phase behavior: the phase and amplitude are always paired, always proportional, always tied to each other.
What Minimum Phase Sounds Like in Practice
The phase rotation from a minimum phase EQ is not random. It is program-dependent, frequency-dependent, and directly linked to the amplitude change you made. Because the shift is always at the frequency you processed, it tends to integrate with the material rather than fight it. Your ear has been hearing minimum phase behavior from acoustic instruments and analog electronics its entire life. It is not confused by it.
What minimum phase EQ does to transients is more nuanced. A sharp transient, like a snare hit or a plucked string, contains energy spread across a wide frequency range. When a minimum phase EQ processes that transient, different frequency components arrive at slightly different times. The transient shape changes. This is sometimes described as smearing, which implies it is always bad. It is not always bad. The rounding and integration that minimum phase behavior adds to transients is part of what gives analog EQ its character, particularly on low end and midrange, where the phase shift from a boost adds a quality of weight and cohesion that is distinct from simply turning up the level.
Linear Phase: The Digital Alternative
Linear phase EQ is a digital-only option. It does not exist in the analog world because it requires non-causal filter behavior implemented through lookahead and latency: to preserve relative phase relationships across frequencies, a linear phase filter has to look ahead in time, introducing latency so that the delay applied to all frequencies can be equal. Every frequency is delayed by the same amount, so no frequency arrives earlier or later relative to any other. Phase relationships between frequencies are preserved exactly.
This sounds like a strict improvement. In some situations it is. But it introduces an artifact that has no analog equivalent and that the ear finds distinctly unnatural: pre-ringing.
Pre-ringing is energy that appears before the transient that caused it. In the time domain it looks like a decaying oscillation that precedes the main event. The filter at the center of a linear phase EQ is symmetric in time, which means whatever it does after a transient, it also does before it. A sharp boost at 100Hz will produce oscillations at 100Hz both before and after the transient passes through.
The ear does not process pre-ringing as a resonance or a tone color. When audible, it often registers as something unnatural. Sound in the physical world decays after an event. It does not appear before it. Pre-ringing presents the ear with behavior that does not normally exist in the physical world, and the auditory system tends to treat it accordingly. On dense material with few sharp transients, it is often inaudible. On sparse material with clear, isolated percussive events, it can be immediately obvious.
Post-Ringing: The Minimum Phase Version
Pre-ringing gets all the attention. Post-ringing is the quieter problem, and in mastering it shows up more often.
Post-ringing is the time-domain consequence of a steep, narrow filter in a minimum phase EQ. When a high-Q filter makes a deep cut or sharp boost, it oscillates at its center frequency after the signal passes through. The narrower the Q and the deeper the cut, the more the filter rings. The ringing decays, but it decays at the filter's own frequency, which means it adds energy at that frequency in the moments after a transient.
The reason post-ringing is less discussed than pre-ringing is that the ear tolerates it better. Sound decays after an event. That is acoustically normal. Post-ringing reads as resonance or coloration rather than as a causality violation, so it tends to be absorbed into the character of the EQ rather than flagged as an artifact. But absorbed into the character is not the same as absent. In mastering, on the right material, it is audible.
I know this because of a Peals session.
The Session That Taught Me This Properly
Peals is a duo from Baltimore, William Cashion of Future Islands and Bruce Willen of Double Dagger, making quietly radiant instrumental music from tessellated guitars, synths, and percussive toys. I was mastering The Compound 76 for Thrill Jockey, a live recording from 2016 that is minimal and spacious and full of sharp transient contrasts.
The record had a hum problem. 60Hz electrical interference, the kind that shows up on live recordings in certain rooms. The standard tool for this is iZotope RX De-Hum, which uses narrow notch filters to target the fundamental and harmonics: 60Hz, 120Hz, 240Hz, 480Hz, and so on up the series.
The problem with narrow notch filters stacked at harmonically related frequencies is that each one is ringing. After every transient, each filter oscillates at its own center frequency. On a dense mix with constant activity, those rings are buried. On a minimal percussive record with space between events, they were not buried. They were audible as a flutter, a spectral shimmer that trailed each hit. Not the hum. The cure.
The client heard it before I could point to it.
We tried switching between minimum phase and linear phase processing in RX. Both modes had artifacts, just different ones. Linear phase gave us pre-ringing: the shimmer moved to before the hit. Minimum phase kept the post-ringing: the shimmer stayed after. Switching modes was not the answer.
The actual solution was simpler and more fundamental: fewer harmonics tracked, wider bandwidth on each notch. Less aggressive filters meant less ringing in both directions. We found the point where the hum was sufficiently attenuated and the ringing was no longer audible, and we stopped there. The phase mode was downstream of that decision. A gentle filter in either mode usually behaves well. An aggressive filter in either mode causes problems, just different ones.
The lesson: phase mode is not the variable. Filter aggressiveness is the variable. The phase behavior follows from that.
When Linear Phase Is the Right Call
Linear phase EQ is not wrong. It tends to be the wrong tool on material with sharp transients and sparse texture, where pre-ringing has nowhere to hide. It tends to be the right tool for broad tonal adjustments on dense material, where the transient content is not isolated and the pre-ringing is not audible against the surrounding energy. A gentle high-shelf boost on a busy rock mix. A broad low-end roll-off on material that has no sharp subsonic transients to reveal the artifact.
The Weiss EQ1 switches the entire plugin between modes globally. FabFilter Pro-Q gives you the choice per band and lets you blend them. Ozone 12 gives you both modes with a switch. The choice is meaningful, and reaching for linear phase by default because it sounds technically superior is the wrong instinct. Ask first what the material contains, and specifically whether it contains the transient structure that will reveal pre-ringing.
Plugins That Behave Like Analog
Not all digital EQ is linear phase. The Sontec MES-432D9D plugin from Make Believe Studios uses Metric Halo's State Space Model Extraction Process to capture the actual analog circuit behavior of the hardware MES-432, including its minimum phase response. It is not a plugin with a Sontec-shaped frequency curve. It is a plugin that behaves in the time domain the way a Sontec behaves in the time domain. That distinction is the entire difference between a modeled EQ and a clone.
Chris Gehringer, who has used the hardware extensively, said of the plugin: "This doesn't sound like a plugin. This sounds like a Sontec." One reason it sounds like a Sontec is that it preserves the phase behavior alongside the frequency response, not just the shape of the EQ curve.
The Weiss EQ1, when run in minimum phase mode, behaves the same way your analog EQs do. The MEA-2 and the ITI MEP-130 and the Pultec on your desk are minimum phase by nature, always have been, and the character attributed to each of them, the weight of the Pultec low end, the presence of the ITI midrange, the precision of the MEA-2, is partly a frequency response story and partly a phase story. The two are inseparable in analog. In digital, you can pull them apart, which means you have choices that Leif Mases and George Massenburg never had. That is useful, if you understand what you are choosing between.
Q Is a Time-Domain Decision
The practical takeaway from all of this is not a setting or a mode. It is a way of thinking about EQ moves.
Every time you narrow the Q on a parametric band, you are also lengthening and concentrating the filter's ringing behavior in time after a transient. Every time you make a deeper cut or a sharper boost, you are increasing the amplitude of that ringing. The frequency response display on your EQ shows you one dimension of what the filter is doing. The time domain is the other dimension, and it is invisible on the screen in front of you.
On dense material this rarely matters. On sparse material, or material with isolated transient events and spaces between them, it matters every time. The Peals session was an extreme case, multiple stacked notches at harmonically related frequencies, but the same principle applies to a single tight parametric cut on a resonant snare. The cure can introduce its own color in the decay. Whether that color is acceptable depends on the material and how carefully you are listening.
Linear phase is not universally superior. It trades one class of artifacts for another. Minimum phase is not a limitation. It is a characteristic, and the best analog EQs produce minimum phase behavior that has been refined over decades into something musically coherent. Understanding which you are working with, and what the material you are processing will reveal, is the difference between reaching for a tool and knowing why. That, and knowing when to leave the Q alone.
