Understanding Phase Coherence
Phase coherence describes how faithfully a monitor preserves the relative timing of frequency components as it reproduces a signal. A perfectly phase-coherent system delays all frequencies equally (or not at all), so a transient arrives intact. Real monitors introduce frequency-dependent phase shift, quantified by phase response and group delay. The main sources are crossovers, physical driver offset and ported enclosures. Most of these behave as minimum-phase systems, meaning magnitude and phase are linked, while linear-phase DSP can flatten phase at the cost of pre-ringing and latency. This article explains the theory, where coherence is lost, how it's measured via impulse and step response, and what is genuinely audible.
What Phase and Phase Coherence Mean
Any signal can be decomposed into sinusoidal components. Each component has an amplitude and a phase, its position within its cycle at a given instant. A complex waveform's shape depends not just on the amplitudes of its components but on their relative phases. Shift those phases independently and the waveform changes shape, even if the magnitude spectrum is identical.
- Phase
- The position within a cycle of a sinusoidal component, measured in degrees or radians. Relative phase between components determines the shape of a composite waveform.
- Phase coherence
- The degree to which a system preserves the correct relative phase (timing) of all frequency components. A phase-coherent reproduction keeps the relationships present in the original signal, so transients and complex waveforms are reproduced with their shape intact.
A system is phase-coherent for a signal if it imposes either no phase shift or a phase shift that corresponds to a constant time delay across frequency. A constant delay simply shifts the whole waveform later in time without distorting its shape. Problems arise when different frequencies are delayed by different amounts, because the waveform then smears.
Phase Response and Group Delay
Two related measurements describe a system's phase behaviour. Phase response plots phase shift against frequency. Group delay re-expresses the same information as time. It is the negative derivative of phase with respect to angular frequency, and it tells you how long a narrow band of frequencies is delayed as it passes through the system.
- Phase response
- A plot of the phase shift a system imposes as a function of frequency. A flat (zero-slope after removing pure delay) phase response means all frequencies share the same timing.
- Group delay
- τg(ω) = −dφ/dω, the time delay experienced by the envelope of a narrow band of frequencies. Constant group delay across frequency means no phase distortion. Varying group delay means different bands arrive at different times.
A constant group delay is benign, because it is just latency. It is the variation of group delay with frequency that constitutes phase distortion. For example, a ported monitor exhibits a group-delay rise in the low bass around the port tuning frequency, meaning the lowest notes lag behind the rest of the signal. Whether that lag is audible depends on its size and the frequency, discussed below.
Suggested diagram: impulse and step response, coherent vs incoherent
Two columns. Left, a phase-coherent monitor, with a clean, narrow impulse response and a step response that rises sharply and settles. Right, an incoherent monitor, with an impulse smeared in time showing separate woofer and tweeter arrivals, and a step response with a stepped or sloped leading edge. Annotate the time-axis differences between high- and low-frequency arrivals.
Where Monitors Lose Coherence
A monitor's coherence is degraded at a few well-understood points.
Crossovers
Splitting the signal between drivers requires filters, and analog (IIR-type) crossover filters impose phase shift around the crossover frequency. A standard fourth-order Linkwitz–Riley crossover, for instance, sums flat in magnitude but rotates the phase by 360 degrees through the crossover region. The drivers are in phase at the crossover frequency, but the through-system phase still rotates with frequency, so the composite waveform's timing is altered around that band.
Physical driver offset
If the woofer's and tweeter's acoustic centres are at different distances from the listener, as on a flat baffle, their outputs arrive at different times. This acoustic offset is a non-minimum-phase effect. It's a genuine time-of-arrival difference, not just a magnitude anomaly, and it can't be corrected by EQ alone. Designers address it with sloped or stepped baffles, careful crossover design, or DSP delay. A single full-range driver, or a coaxial with coincident acoustic centres, avoids the offset entirely.
Ported enclosures
A bass-reflex port is a Helmholtz resonator that extends low-frequency output but adds phase rotation and a group-delay peak near the tuning frequency. A sealed enclosure has a simpler, more gradual roll-off with less low-frequency group delay, which is one reason sealed designs are often described as tighter or more time-accurate in the bass.
Minimum phase and the rest
A minimum-phase system is one whose phase is uniquely determined by its magnitude response (related by the Hilbert transform). Most individual driver resonances and room modes are minimum-phase, so fixing the magnitude with EQ also fixes the phase. Driver time-of-arrival offset and some crossover behaviour are non-minimum-phase and require delay or FIR techniques, not just EQ.
Minimum-Phase vs Linear-Phase Correction
Once you can measure phase, you can attempt to correct it with DSP. There are two broad approaches, with different trade-offs.
- Minimum-phase system
- A system with the least possible phase shift for its magnitude response, where magnitude and phase are linked. Correcting the magnitude of a minimum-phase anomaly with EQ simultaneously corrects its phase.
- Linear-phase system
- A system with phase that varies linearly with frequency, that is, constant group delay, so all frequencies are delayed equally. Achievable with symmetric FIR filters, at the cost of latency and potential pre-ringing.
| Minimum-phase (IIR) | Linear-phase (FIR) | |
|---|---|---|
| Latency | Low | Higher (filter length dependent) |
| Pre-ringing | None | Possible (artefact before transient) |
| Fixes magnitude | Yes | Yes |
| Fixes phase independently | Only as linked to magnitude | Yes, can flatten phase |
| Good for | Driver and room minimum-phase anomalies | Flattening through-system phase, crossover linearisation |
| Risk | Can't fix non-minimum-phase offset | Audible pre-ringing if misapplied |
The practical lesson is that linear-phase correction is capable but not free. Long FIR filters add latency (a problem for live tracking) and, applied carelessly, can produce pre-ringing, which is energy that appears before the transient and which the ear finds unnatural. Minimum-phase correction is latency-friendly and matches most real loudspeaker and room anomalies, which is why a lot of monitor and room DSP is minimum-phase by design.
Measurement and Audibility
Phase behaviour is measured from the system's impulse response, captured with a calibrated mic and dual-channel FFT or log-sweep tools (such as REW). The impulse response contains the full magnitude and phase information. The step response is a useful visual check of coherence, because a time-aligned, coherent system produces a clean step while driver offset shows up as a stepped or sloped leading edge.
- Impulse response
- The system's output to an ideal impulse. It fully characterises a linear system, because its Fourier transform gives both magnitude and phase response, and it reveals reflections and time smearing directly.
Audibility depends on the type and magnitude of the distortion. Steady-state relative phase of harmonics is often inaudible in isolation (Ohm's acoustic law), but group-delay distortion on transients is more readily heard. Classic studies (Blauert and Laws) put group-delay audibility thresholds at a few milliseconds, rising in the low bass where the ear is less sensitive to timing. In practice, the most audible consequences in monitoring are crisp transient reproduction, a stable and well-focused stereo image, and clean behaviour when a stereo mix is summed to mono, all of which benefit from good coherence and all of which matter for mixing decisions.
Applying this at the desk
You don't need to chase a perfectly flat phase curve. Prioritise designs that avoid gross group-delay distortion (sensible crossover, time-aligned or coincident drivers, sealed bass where tightness matters), measure your system's impulse response, and judge transients, imaging and mono behaviour by ear against references.
Common Misconceptions
Phase coherence is inaudible audiophile nonsense.
Steady-state harmonic phase is often inaudible, but group-delay distortion on transients, and the imaging and mono consequences of incoherence, are real and measurable. The effect is genuine. The overstatement is in treating tiny, inaudible phase curves as critical.
A flat magnitude response guarantees good phase.
Only for minimum-phase anomalies. Non-minimum-phase effects like driver time-of-arrival offset and some crossover behaviour can leave phase distortion even when magnitude is flat.
Linear-phase correction is strictly better than minimum-phase.
It can flatten phase, but it adds latency and risks pre-ringing artefacts. Minimum-phase correction is latency-friendly and matches most real loudspeaker and room anomalies. Each has appropriate uses.
EQ can fix any phase problem.
EQ corrects minimum-phase magnitude (and its linked phase). It cannot, on its own, fix a genuine time-of-arrival offset between drivers, which needs delay or FIR techniques.
Frequently Asked Questions
What is phase coherence in a studio monitor?
It's how faithfully the monitor preserves the relative timing of frequency components. A coherent monitor delays all frequencies equally (or not at all), so transients and complex waveforms keep their shape. Incoherence smears timing, blurring transients and destabilising the stereo image.
What is the difference between phase response and group delay?
Phase response plots phase shift versus frequency. Group delay is the negative derivative of phase with respect to angular frequency, expressing the same information as a time delay per frequency band. Constant group delay is benign latency, while varying group delay is phase distortion.
Why do crossovers affect phase?
Splitting the signal between drivers needs filters, and analog (IIR-type) crossover filters impose phase rotation around the crossover frequency. Even a Linkwitz–Riley crossover that sums flat in magnitude rotates the through-system phase, altering waveform timing around that band.
Does a sealed monitor have better phase than a ported one?
In the low end, typically yes. A port adds phase rotation and a group-delay peak near its tuning frequency, so the lowest notes lag. A sealed enclosure rolls off more simply with less low-frequency group delay, which is why sealed designs are often described as tighter or more time-accurate.
What is a minimum-phase system?
One whose phase is uniquely determined by its magnitude response (linked via the Hilbert transform), with the least phase shift possible for that magnitude. For minimum-phase anomalies, correcting the magnitude with EQ also corrects the phase. Most driver resonances and room modes are minimum-phase.
Can EQ fix phase problems?
It can fix minimum-phase anomalies, because their magnitude and phase are linked. It cannot fix non-minimum-phase effects like driver time-of-arrival offset on its own, which require delay alignment or FIR (linear-phase) techniques.
What is linear-phase correction and what's the catch?
Linear-phase correction uses symmetric FIR filters to impose constant group delay, flattening phase. The catches are added latency (a problem when tracking) and possible pre-ringing, which is energy appearing before a transient, if applied carelessly. It's a trade-off, not a free improvement.
How is phase coherence measured?
From the system's impulse response, captured with a calibrated microphone and dual-channel FFT or log-sweep software like REW. The impulse response yields both magnitude and phase, and the step response is a quick visual check, showing a clean step for a coherent system and a stepped edge when drivers are offset.
Is phase distortion actually audible?
It depends. Steady-state harmonic phase is often inaudible (Ohm's acoustic law), but group-delay distortion on transients is more readily heard, with thresholds of a few milliseconds (higher in the bass). The clearest practical consequences are transient clarity, image stability and mono behaviour.
How does point-source design relate to phase coherence?
A single full-range driver has no crossover and one acoustic centre, so it's inherently time-aligned and avoids crossover-region phase rotation and driver offset. Coaxial point sources place the tweeter at the woofer's acoustic centre to achieve much the same coincidence. Both improve coherence by construction.
Should I aim for a perfectly flat phase response?
No. The goal is to avoid gross, audible group-delay distortion, not to obsess over a flat phase curve. Choose well-engineered crossovers or time-aligned and coincident drivers, measure your impulse response, and judge transients, imaging and mono compatibility by ear against references.
Conclusion
Phase coherence is the preservation of relative timing across frequency, and it's best understood through phase response and group delay. A constant delay is harmless, but frequency-dependent delay smears transients and destabilises imaging. Monitors lose coherence mainly at the crossover, from physical driver offset, and from ported low-end behaviour, and whether those are correctable depends on whether they're minimum-phase (fixable with EQ) or not (needing delay or FIR). Linear-phase DSP can flatten phase but trades latency and pre-ringing for it. Measured from the impulse response and judged against realistic audibility thresholds, coherence is not mysticism. It shows up as transient clarity, a stable image and clean mono, the qualities that make a monitor trustworthy for mixing.
Glossary
- Phase
- The position within a cycle of a sinusoidal component. Relative phase shapes composite waveforms.
- Phase response
- Phase shift as a function of frequency.
- Group delay
- −dφ/dω, the time delay of a narrow frequency band. Constant is benign, varying is phase distortion.
- Minimum-phase
- A system whose phase is fixed by its magnitude, so EQ that fixes magnitude also fixes phase.
- Linear-phase
- Constant group delay (all frequencies delayed equally), achievable with FIR filters at the cost of latency and pre-ringing.
- Impulse response
- A system's output to an impulse, which fully characterises its magnitude and phase.
Sign up to hear more when we release something new
