/ˈlɪn.i.ər feɪz ˌiːˌkjuː/
Linear Phase EQ is an equalizer that applies frequency boosts and cuts without altering the phase relationship between frequencies. Unlike conventional EQs, it preserves transient integrity at the cost of processing latency and pre-ringing artifacts.
Every time a conventional EQ touches your mix, it quietly rearranges time — nudging certain frequencies ahead or behind others in ways your ears may never consciously detect, but your mix will absolutely feel. Linear phase EQ is the tool that refuses to make that trade.
Linear phase EQ is a class of digital equalizer that applies amplitude corrections across the frequency spectrum while preserving the phase relationships between all frequency components of an audio signal. In a minimum phase equalizer — the architecture underlying virtually every analog hardware EQ and most digital emulations — a boost or cut at a given frequency inevitably introduces a corresponding phase shift: frequencies near the center of the filter are delayed or advanced relative to those outside it. This is an inescapable property of minimum phase filter design, and in isolation it is rarely audible as a distinct artifact. Cumulatively, across a full mix with multiple equalizers in series, it can smear transient clarity, alter stereo imaging, and change the tonal character of the program material in ways that are difficult to diagnose.
Linear phase EQ solves this by implementing equalization through a Finite Impulse Response (FIR) filter rather than the Infinite Impulse Response (IIR) filters used in minimum phase designs. An FIR filter achieves a perfectly flat group delay across the entire audible spectrum — meaning every frequency takes exactly the same amount of time to pass through the processor, regardless of whether it has been boosted, cut, or left untouched. The amplitude correction is identical to what a minimum phase EQ would produce; the phase response is not. The result is equalization that is, in the mathematical sense, phase-neutral: it changes the level of frequencies without repositioning them in time.
The practical consequence is most apparent on program material with sharp transients — drum hits, plucked strings, percussion — and on stereo or mid-side processing where phase coherence between channels directly governs stereo width and imaging stability. When a conventional parametric EQ boosts the low end of a full mix, it delays low-frequency content relative to the high end; the kick drum's attack becomes subtly separated from its body, and the mix can lose the snap and punch that made the raw tracks sound exciting. A linear phase EQ delivering the identical boost leaves the transient structure intact. For mastering engineers, this distinction is the primary reason linear phase processing has become a standard tool since the late 1990s.
However, linear phase EQ carries two significant technical liabilities that prevent it from being a universal replacement for minimum phase designs. The first is latency: FIR filters require the processor to look ahead in time, buffering audio samples before outputting them. This lookahead introduces fixed processing delay — typically between 20 and 100 milliseconds depending on filter length and sample rate — which is manageable in non-real-time rendering but prohibitive for monitoring-latency-sensitive tasks like tracking and live performance. The second liability is pre-ringing: the mathematical mirror image of the post-ringing that characterizes minimum phase filters. Rather than a resonant tail following a transient, a linear phase filter can generate a faint artifact that arrives before the transient, which some engineers find more audible and objectionable than the smearing it was designed to prevent, particularly on extreme low-frequency boosts.
Understanding linear phase EQ, therefore, requires understanding both its genuine strengths and its genuine costs. It is not universally superior to minimum phase EQ — it is a specialized tool with a specific domain where its properties are advantageous. That domain is primarily mastering, stereo bus processing, mid-side equalization, and any context where multiple instances of equalization interact with the same shared signal path. In those contexts, the phase neutrality of a linear phase design can be the difference between a mix that coheres and one that subtly doesn't.
The core mechanism of a linear phase EQ is the FIR (Finite Impulse Response) filter, which is fundamentally different in architecture from the IIR (Infinite Impulse Response) filters that underpin all analog-modeled and most digital minimum phase EQs. An IIR filter operates with feedback — the output at any given moment is computed from both the current input and the filter's own previous outputs. This recursive structure is mathematically equivalent to analog RC circuits and resonant filters, and it produces the phase shift as a direct, unavoidable byproduct of how energy builds and decays through the feedback loop. The shorter the filter, the lower the computational cost, but the more phase shift it introduces at transition frequencies.
A FIR filter has no feedback path. Its output is computed purely as a weighted sum of the current input and a finite number of previous input samples — the filter's impulse response. To achieve a specific frequency response (say, a +3 dB shelf above 10 kHz), the designer constructs a set of coefficient weights that, when convolved with the audio signal, produce exactly that amplitude change. By making this impulse response symmetric in time — identical on both sides of the center sample — the group delay is rendered constant at all frequencies. The symmetry requirement means the filter must extend equally far backward and forward in time, which is why the processor must buffer audio before outputting it: it needs future samples to compute its output for the current moment. A filter with a 2048-sample impulse response at 44.1 kHz needs approximately 23 milliseconds of lookahead; at 96 kHz the same filter length consumes about 10.6 ms, which is one reason high sample rates partially mitigate latency.
Pre-ringing is the artifact that arises directly from this symmetric structure. Because the impulse response extends into negative time (relative to the event), a sharp transient — like a drum hit — will cause the filter to begin responding before the transient arrives, producing a faint echo that precedes the attack. In minimum phase filters, the equivalent artifact is post-ringing: a resonant decay after the transient. Post-ringing is generally less audible to human hearing because the masking effect of the transient itself conceals the tail. Pre-ringing has no such masking advantage. In practice, pre-ringing only becomes problematic with steep, high-gain filters in the low-frequency range (below ~200 Hz), where the artifact's period is long enough to be perceived; gentle shelf filters and bell curves in the midrange and high frequencies produce pre-ringing too short in duration to be audible. Many modern implementations — FabFilter Pro-Q 3, Sonnox Oxford EQ in linear mode, Eventide SplitEQ — offer mixed-phase or frequency-dependent linear phase modes that apply FIR filtering only where it matters while reverting to minimum phase behavior in the low end.
Computational cost scales with filter length. A minimum phase parametric EQ can achieve its entire frequency response with as few as two to five multiply-accumulate operations per sample per band. A high-resolution FIR linear phase EQ may require 4096 to 16384 multiply-accumulate operations per sample — orders of magnitude more. Modern CPUs handle this workload well in offline or non-real-time rendering, and plugin developers have implemented fast convolution algorithms (overlap-add, overlap-save) that make real-time linear phase processing practical within typical DAW buffer sizes. The computational demand is why many producers reserve linear phase EQ instances for final renders or dedicated mastering sessions rather than keeping many instances active during composition and arrangement.
When a DAW compensates for plugin latency automatically — as Ableton Live, Logic Pro, Pro Tools, and Reaper all do — the practical impact of linear phase latency is often invisible during mixing: the DAW delays all other tracks to align with the delayed output of the linear phase plugin. The caveat is live monitoring: if a performer is monitoring their own signal through a chain that includes a linear phase EQ, the lookahead delay becomes an audible monitoring latency that cannot be compensated away, since it requires audio that has not yet happened.
Diagram — Linear Phase EQ: Diagram comparing group delay and frequency response curves of minimum phase EQ vs linear phase EQ, with pre-ringing illustration.
Every linear phase eq — hardware or plugin — operates on the same core parameters. Know these and you can work with any implementation.
In plugins like FabFilter Pro-Q 3 and iZotope Ozone EQ, each band can independently operate in minimum phase or linear phase mode. Use minimum phase for individual tracks where CPU efficiency and low latency matter; switch to linear phase mode on master bus or for steep surgical cuts where phase smear would be audible. Mixed-phase mode — available in some implementations — applies linear phase only to low-frequency bands where pre-ringing risk is lowest and phase coherence benefit is highest.
Longer filter lengths produce more accurate frequency selectivity and steeper transition bands, but increase both processing latency and CPU load. A 4096-sample FIR at 44.1 kHz introduces ~46 ms of lookahead; at 96 kHz the same length costs ~21 ms. For mastering with gentle shelves, 512–1024 samples often suffices. For brickwall or surgical linear phase cuts, 8192 samples or more may be necessary to achieve clean rejection without ripple in the passband.
In linear phase mode, gain behavior is functionally identical to minimum phase EQ — positive values boost, negative values cut. However, extreme gain values (+/- 12 dB or more) in linear phase mode with short filter lengths can produce visible pre-ringing artifacts on sharp transients. Best practice is to keep individual band gains under ±6 dB in linear phase mode; for more aggressive correction, stack multiple gentle bands rather than relying on a single high-gain filter.
Frequency placement in linear phase EQ carries the same psychoacoustic logic as in any EQ, but the pre-ringing concern makes low-frequency placement particularly important. Filters below 100 Hz with high gain will produce pre-ringing artifacts with a temporal period long enough (>10 ms) to be perceived as a faint pre-echo before a drum transient. Above 500 Hz, pre-ringing periods are typically below the threshold of audibility. Most mastering engineers switch low-frequency bands to minimum phase mode when significant correction is needed below 150 Hz.
Wider Q values (lower numbers, e.g., Q 0.5–1.5) spread the amplitude correction across a broad frequency range, distributing the FIR coefficient energy across many samples and reducing pre-ringing intensity. Narrow Q values (Q 3 and above) concentrate energy, increasing the effective filter length required for clean implementation and amplifying the risk of audible pre-ringing. For linear phase mastering work, broad-Q corrections and gentle shelving filters are preferred over narrow notches; reserve tight Q values for offline surgical repair tasks where pre-ringing can be checked in the rendered file.
All major DAWs implement Automatic Delay Compensation (ADC) for plugin latency, but the interaction with linear phase EQ requires verification. In Pro Tools, check the Track Delay Indicator in the I/O section to confirm compensation is active. In Ableton Live, delay compensation is enabled by default under Preferences > Audio. When linear phase EQ is inserted on the master bus only, ADC aligns the entire mix automatically. When inserted on individual tracks mid-session with live monitoring active, the lookahead delay appears as monitoring latency that ADC cannot remove, since it requires audio that has not yet been recorded.
Session-ready starting points. These values assume standard 44.1–48 kHz sessions; at 96 kHz, filter lengths can be doubled without proportional latency penalty.
| Parameter | General | Drums | Vocals | Bass / Keys | Bus / Master |
|---|---|---|---|---|---|
| Recommended Mode | Context-dependent | Min. phase preferred | Min. phase preferred | Min. phase preferred | Linear phase preferred |
| Max Gain Per Band | ±6 dB | ±4 dB (transients) | ±5 dB | ±5 dB | ±3 dB |
| Filter Length | 512–1024 samples | N/A (use min. phase) | N/A (use min. phase) | 512 samples | 2048–4096 samples |
| Q / Bandwidth | 0.5–1.5 (broad) | 0.7–2.0 | 0.7–1.5 | 0.5–1.2 | 0.3–1.0 (broad only) |
| Low-Freq Bands (<150 Hz) | Consider min. phase | Min. phase only | Min. phase only | Min. phase only | Min. phase or mixed |
| Latency Concern | Low (offline OK) | High — avoid tracking | High — avoid tracking | Medium | Low (bounce/render) |
| Pre-Ringing Risk | Low (gentle curves) | High (steep + low-freq) | Low (mid/high focus) | Medium (sub content) | Low (gentle shelves) |
These values assume standard 44.1–48 kHz sessions; at 96 kHz, filter lengths can be doubled without proportional latency penalty.
The conceptual foundation for linear phase filtering predates digital audio by several decades. In 1965, electrical engineer Otto Brune and later the broader signal processing community formalized the distinction between minimum phase and linear phase (all-pass group delay) filter designs within communication theory. The mathematics of FIR filters had been well-understood since the work of James Kaiser and Lawrence Rabiner at Bell Labs in the early 1970s, who published foundational papers on optimal FIR filter design using windowing functions. However, the computational cost of implementing long FIR filters in real time — each sample requiring thousands of multiply-accumulate operations — made the concept impractical for audio hardware through the 1970s and most of the 1980s.
The first commercially significant application of linear phase filtering in professional audio came not from studio EQ but from crossover design. In 1977, Siegfried Linkwitz and Russ Riley described the Linkwitz-Riley crossover topology, which, while not strictly FIR-based, motivated sustained engineering interest in phase-coherent frequency division. The emergence of affordable DSP chips in the late 1980s — particularly the Motorola 56000 family and Texas Instruments TMS320 series — made FIR-based audio processing financially viable for high-end products. The TC Electronic System 6000, introduced in 1999, included a linear phase EQ among its mastering processing suite and became one of the first hardware units to bring the technology to professional studios. Its latency of approximately 42 milliseconds was accepted as a fair tradeoff by mastering engineers who were already working with tape or digital file playback rather than live monitoring.
The software era brought linear phase EQ to a far wider audience. Waves introduced the Linear Phase Equalizer (later Linear Phase EQ) plugin around 2001, bundled in their Platinum and Masters bundles, explicitly targeting mastering engineers. The plugin's graphic interface and preset library introduced the concept to a generation of DAW-based engineers who had never used the TC System 6000. Brainworx (now Plugin Alliance) and Sonnox followed with their own implementations — the Sonnox Oxford EQ with its switchable linear phase mode arrived around 2005 and became a staple on SSL and Neve consoles with DSP cards. The pivotal moment for mainstream adoption came in 2012 when FabFilter released Pro-Q 2, which allowed per-band switching between minimum phase and linear phase modes at zero additional CPU cost in non-linear-phase mode — eliminating the all-or-nothing tradeoff that had characterized earlier implementations and making linear phase a tool producers could reach for selectively rather than committing to globally.
The 2010s saw further refinement through the concept of mixed-phase or frequency-dependent linear phase modes. Eventide's SplitEQ (2021) took a different approach entirely, separating transient and tonal content before applying independent equalization to each, sidestepping pre-ringing by applying linear phase correction only to the tonal component where it causes no artifact. iZotope's Ozone Equalizer introduced per-band phase mode selection in Ozone 9 (2019), and the company's mastering AI systems began using linear phase mode by default on bus processing suggestions. Today, linear phase EQ is less a specialized product category and more a mode available within virtually every serious parametric EQ plugin, with the question no longer being "which plugin is linear phase" but "when and how aggressively should I engage linear phase processing within the tools I already own."
In mastering sessions, linear phase EQ is the default choice for virtually every working mastering engineer. When correcting the tonal balance of a finished mix — adding air above 12 kHz with a high shelf, warming the low-mids with a broad boost around 200 Hz, or reducing harshness in the 3–5 kHz range — the engineer is working on material where every frequency component of every instrument is already present and interacting. A minimum phase boost in the low end would delay bass frequencies relative to treble, subtly separating the kick drum's low-frequency weight from its transient click. Over repeated listening, this separation contributes to a mix that feels less tight. Linear phase processing eliminates this concern entirely. Mastering engineers like Bob Ludwig, Emily Lazar, and Chris Athens have discussed in interviews the role that phase-coherent processing played in maintaining punch and imaging in dense modern masters.
On stereo and mix buses during mixing, the calculus is similar. When a linear phase EQ is inserted on a drum bus that contains kick, snare, overheads, and room mics, any phase shift introduced by a conventional EQ would affect the relationship between the close-miked transients and the room sound captured by the overheads. A low-frequency boost that delays the kick's sub content relative to its attack can turn a tight, punchy drum bus into something that feels slightly bloated. Linear phase mode preserves the phase offset between microphones that was established at tracking time, keeping the stereo image locked and the transient envelope consistent. This is the application most frequently cited by mix engineers as a reason to reach for linear phase — not mastering, but bus processing where multiple microphone perspectives share a single fader.
Mid-side equalization is another domain where linear phase processing provides a distinct advantage. When applying M/S EQ — boosting the mid channel while cutting the sides, for example — minimum phase processing would introduce different phase shifts to the M and S components before they are re-encoded back to L/R. These differential phase shifts can alter stereo imaging in unpredictable ways, widening or narrowing the apparent width at specific frequencies. Linear phase M/S EQ guarantees that the phase relationship between mid and side is maintained post-processing, ensuring that what changes is only amplitude — widening achieved through gain differences rather than phase artifacts. Plugins like Brainworx bx_digital V3 and iZotope Ozone EQ are commonly used in linear phase M/S mode for this purpose.
For individual tracks during mixing — vocals, guitars, synths, individual drum mics — most engineers opt for minimum phase EQ for both pragmatic and sonic reasons. The CPU overhead is lower, the latency is negligible, and the phase shift introduced by a single EQ on a single track is rarely perceptible in context. Some engineers argue that the phase shift of a minimum phase EQ is part of its sound, particularly for hardware emulations where the filter's phase behavior contributes to the character associated with a Neve 1073 or API 550. Linear phase mode on individual tracks is occasionally useful for forensic repair work — removing a narrow resonance from a recorded instrument without altering the transient structure — but for corrective and creative EQ on individual sources, minimum phase designs are the professional standard.
One email a week. The techniques behind the terms — curated by working producers, not algorithms.
Abstract knowledge becomes practical when you can hear it in music you know. These tracks demonstrate linear phase eq used intentionally, at specific moments, for specific purposes.
Bob Ludwig has cited this master as a case study in maintaining bass punch across a broad low-end balance correction. The kick drum and bass guitar retain a coherent transient relationship throughout the track's length — the click of the kick sits precisely atop the low-frequency weight rather than being perceptibly separated from it. Listen through headphones at 0:20–0:40 as the full band enters; the sub-bass movement under the kick hits cleanly without the slight low-end bloom that minimum phase low-shelf boosts can introduce at full-mix mastering stages. The linear phase processing preserves the stereo image width established in the mix, audible in how the guitars spread evenly without frequency-dependent narrowing.
The solo kick that opens the track is a useful reference for bus EQ transparency. The low-frequency weight of the kick (approximately 50–60 Hz) and its transient punch (around 3–5 kHz) are perceptibly phase-aligned — they arrive together rather than in sequence. Colin Leonard has discussed the importance of phase-coherent processing on the master bus for maintaining this relationship on a track where the kick is effectively the entire low-end anchor. Compare the transient sharpness on a streaming service playback versus an MP3 encode; the linear phase processing at mastering preserves transient clarity that compression artifacts in lossy codecs later attack.
This track's characteristic quality — a bass-forward, hyper-close vocal image with precise stereo placement of high-frequency percussion — is a direct illustration of phase-coherent bus processing. John Greenham has noted that the stereo bus chain for the album prioritized phase transparency, using linear phase EQ for the high-frequency air boost above 14 kHz that gives Eilish's voice its intimate presence without widening the stereo field (which minimum phase high-shelf filters often do by differentially advancing treble energy in the outer channels). At 0:05–0:10, note how the hi-hat appears to have an exact stereo position that does not drift with frequency — a hallmark of linear phase stereo bus processing.
Mastered by Bob Ludwig at Gateway Mastering, this album (Kid A) is frequently cited in mastering education as an early example of digital mastering transparency. The tonal balance of the track — rich in low-mid presence around 200–400 Hz with a clean, detailed top end — shows careful broadband EQ that preserves the coherence of the dense synth and vocal stack. The absence of frequency-dependent stereo image anomalies in the mid-high transition (compare 2 kHz vs 8 kHz stereo width on headphones) suggests phase-coherent processing at the mastering stage, which in 2000 would have involved early DSP-based linear phase tools of the era.
The foundational implementation: a fixed-length FIR filter applied uniformly across all frequency bands. All bands operate in linear phase simultaneously, and the latency is determined by the filter length setting. This type offers the most theoretically pure phase response but provides no flexibility to revert individual bands to minimum phase for low-frequency content. Best suited for dedicated mastering chains where the signal passes through offline rather than in a live monitoring context.
Allows each filter band to independently operate in minimum phase or linear phase mode within a single plugin instance. This is the most flexible and practical implementation for mixing and mastering workflows, enabling linear phase processing in the mid and high frequencies where it provides the most benefit while leaving low-frequency bands in minimum phase to avoid pre-ringing artifacts. FabFilter Pro-Q 3's implementation is the current industry standard for this approach, with per-band phase mode selection visible in the band properties panel.
A more radical departure that separates the input signal into transient and tonal components before applying independent equalization to each component. The tonal component — sustained tones, room ambience, harmonic content — receives linear phase processing without pre-ringing risk, while the transient component can be equalized with minimum phase processing that tracks its temporal structure. This architecture sidesteps the pre-ringing problem entirely and represents the current frontier of transparent equalization design, though at significantly higher CPU cost.
Some implementations use shorter FIR filters with predictive algorithms or IIR-to-FIR approximations to achieve near-linear phase behavior at very low latency — sometimes under 5 milliseconds. These designs trade mathematical purity for practicality, producing group delay that is substantially flatter than minimum phase designs without the 20–100 ms lookahead of full FIR implementations. The phase response is not perfectly linear but is linear enough to eliminate audible phase smear while remaining compatible with monitoring-latency-sensitive tasks.
Dedicated hardware units implementing FIR linear phase equalization in standalone mastering processors. These units typically offer fewer bands than software equivalents but are valued for their conversion quality, discrete signal path, and analog-feel workflow. The Weiss EQ1-LP became a reference mastering processor in the late 1990s and 2000s; Bob Katz's writings on mastering discuss its linear phase operation in the context of preserving stereo imaging on wide-release masters. Hardware linear phase EQs are now supplemented or replaced by their software plugin counterparts in most studios.
Frequency conflicts — two instruments in the same range at similar levels — are the root cause of muddy mixes.
These MPW articles put linear phase eq into practice — specific techniques, real tools, and applied workflows.