Tantrum Audio

Phase in NEXUS and Omnissiah: What We Correct and Why

Omnissiah's phase tuner: a needle for each measured snapshot on a dial from minus 180 to plus 180 degrees, above a strip of bars across frequency coloured by the phase difference between the channels, in degrees.
Omnissiah's phase tuner. Each needle is a measured snapshot, and the strip below shows where across frequency the snapshots' phases agree. After alignment, the needles sit together on the dial.
NEXUS DSP speaker processor, front panel
NEXUSDelay, polarity and all-pass alignment on every output, measured by Omnissiah and run in hardware.
See NEXUS →

A single speaker's phase at your seat is mostly the time its sound takes to reach you. A delay shifts the whole waveform later without changing the waveform's shape, so the ear receives the same sound, later. A delay is audible only against something else, like a second speaker or your own playing. The phase difference between two sources that share a frequency band determines whether their magnitudes add or cancel. NEXUS and Omnissiah measure each speaker against the other at your seat and match their phase responses with delay, polarity and optional all-pass filters. Flattening a single isolated speaker's phase is left out by design. The goal is for every speaker to match arrival times at the listener, at every frequency the speakers share. Optional all-pass filters close the remaining phase gap without increased latency or filter pre-ringing.

LearnWhy You Can't Hear Absolute Phase
01

What NEXUS and Omnissiah Correct

  1. 1Peaks and broad imbalances are corrected with minimum-phase IIR filters. A minimum-phase problem gets a minimum-phase solution: for a room mode's peak at the seat, a cut matched to the mode's frequency, depth and Q fixes the peak's timing too. A deep null in a single speaker's response, where a room reflection cancels the direct sound, is left alone, because a boost raises the direct sound and the reflection by the same amount. A deep null where two speakers arrive 180 degrees apart in phase is corrected by aligning the speakers.
  2. 2Each output is given a delay and a polarity, measured at the listening position. By default the sub is aligned to the mains and left and right to each other. Set Time-Align Tops to Subwoofer only to align just the sub.
  3. 3Phase through the shared band. With Phase Correction on, Omnissiah adds all-pass filters to bring the speakers' phase closer through the band they share.

A minimum-phase filter's phase follows from its level, so a minimum-phase cut that cancels a minimum-phase peak's level also cancels the peak's delay. In a modelled 50 Hz mode with a +12 dB peak and a Q of 15, a matched cut takes the group delay at 50 Hz from 72 ms to 0 ms and the time to fall 60 dB from 636 ms to 52 ms.

Two stacked plots of a modelled 50 Hz room mode. Top: phase from 20 to 100 Hz. The mode swings the phase from plus 37 degrees just below 50 Hz to minus 37 degrees just above it. The cut filter, at 50 Hz, minus 12 dB, Q 15, is a dashed line swinging the opposite way, the mirror image. Mode plus cut is a flat line on zero. Bottom: level at 50 Hz after a note stops, falling 60 dB in 636 ms before correction and in 52 ms after the cut.
A modelled 50 Hz room mode, +12 dB with a Q of 15, before and after a minimum-phase cut with the same frequency, depth and Q. The cut's phase is the mirror image of the mode's, so the two cancel and the ringing shortens. An idealised model: one mode, drawn with the arrival time removed. A measured room adds the travel time to the microphone and several overlapping modes.
02

Aligning The Speakers In Time

Two speakers playing the same octave sum at your seat, and if one arrives late they partly cancel. Through a 4th-order crossover at 80 Hz, a sub 2 ms late costs 1.5 dB, 4 ms costs 6.5 dB and 5 ms costs 11.7 dB. A sub one metre further back than the mains is about 2.9 ms late before its own filters are counted. Acoustic treatment changes reflections, not the arrival time of each speaker's direct sound, so treatment leaves the crossover dip in place.

A crossover cancellation is a broad dip, about an octave wide. Bücklein (1981) found broad response deviations audible at about 1 to 2 dB and narrow ones from about 3 to 5 dB, with dips harder to hear than peaks of the same size. A 3.2 dB dip an octave wide sits above that broad figure, though 80 Hz lies below the bands Bücklein found most sensitive, 200 to 600 Hz and 2 to 6 kHz.

Summed response of a sub and mains through an 80 Hz crossover. Aligned is flat, a sub 2 ms late shows a 1.5 dB dip, and a sub 4 ms late shows a 6.5 dB dip near 90 Hz.
Summed level through an ideal 4th-order 80 Hz crossover with the sub aligned, 2 ms late and 4 ms late. Alignment removes the dips in the 2 ms and 4 ms curves.

Fixing the dip costs only delay. The speaker that arrives first is delayed to match the two speakers' phase curves, so a sub 3 ms late adds 3 ms to the mains. NEXUS holds up to 8.5 ms per output. If a room needs more than 8.5 ms, alignment is skipped with a warning.

A sub's impulse is a slow swell with no sharp peak, so judge alignment by how closely the two phase traces overlap through the crossover rather than by lining up impulse peaks.

Making each speaker linear phase does not fix the dip. Modelled with linear-phase filters, the same crossover with the sub 2.9 ms late leaves the same 3.2 dB dip, because the dip comes from the timing between the two speakers. Aligned, the normal crossover already sums flat with its phase rotation left in.

Two stacked plots from 20 to 500 Hz. Top: the mains' phase through a normal 4th-order Linkwitz-Riley crossover at 80 Hz falls from about 320 degrees to near zero, and the same crossover in linear-phase form is a dashed line flat on zero. Bottom: the summed level of sub and mains. Aligned, the normal and linear-phase crossovers both sit flat on 0 dB. With the sub 2.9 ms late, both crossovers dip to minus 3.2 dB at 97 Hz, and the two dipping lines lie on top of each other.
A 4th-order Linkwitz-Riley crossover at 80 Hz, modelled with normal filters (solid) and linear-phase filters (dashed). Aligned, both versions sum flat. With the sub 2.9 ms late, both versions leave the same 3.2 dB dip at 97 Hz.

By default left and right are polarity-checked and time-aligned to each other. Two identical speakers at equal distances already match. At higher frequencies, the remaining phase difference between the two speakers comes from reflections, and a room's reflections change with listening position.

03

Why an All-Pass Cannot Flatten a Speaker's Group Delay

All-pass filters change phase without changing level, so they look like the natural tool for flattening a speaker's group delay at low latency. Two exact results show that they cannot.

  1. 1A stable all-pass filter only ever adds delay: the filter's group delay is zero or positive at every frequency. An all-pass filter therefore cannot pull late bass forward, and can only hold back everything above the bass until the bass catches up.
  2. 2Each second-order all-pass section adds the same total amount of delay, wherever it is tuned and however sharp it is. Add up the delay it gives each frequency across the whole audio band and the total is always one second-hertz. Lifting a band of B hertz by D seconds needs at least D times B sections.
FeatureDelay to add (D)Band to lift (B)Sections needed
Two-way crossover, 2 kHz0.2 msAbout 2 kHz0.4
Sealed box, 50 Hz roll-off5 msUp to 16 kHz79
Ported sub plus 80 Hz crossover25 msUp to 16 kHz400
All-pass sections needed to flatten each feature (D x B)

The crossover row is feasible, and a 2 kHz crossover's 0.2 ms already sits below the threshold of about 1 ms measured at 2 kHz. The bass rows need 79 and 400 sections, and NEXUS has 15 filters per output. Optimised cascades of four sections change the bass delay little.

An FIR filter could flatten the bass group delay, at 19 ms for an 80 Hz crossover and up to 82 ms for a ported box, with pre-ringing wherever the correction's phase departs from the inverse of the acoustic phase. Linear-Phase Room Correction: The Trade-Offs covers the latency and pre-ringing costs, and the mechanism is in Where the Latency in Linear-Phase Room Correction Comes From.

04

Limits of the Correction

  • Single-speaker group delay. NEXUS does not currently linearise a single speaker's group delay. Removing the delay does not fix a dip between misaligned speakers, and costs 31 ms of FIR latency for a sealed box at 50 Hz, 82 ms for a ported box at 35 Hz, or 79 to 400 all-pass sections. Listening tests have measured audible group delay only above 500 Hz, mostly on headphones with clicks, so no published test we found shows whether a speaker's bass group delay is audible in a room.
  • Room phase at high frequencies. A room's response changes with position, and at short wavelengths a small head movement changes the phase of each reflection, so a correction fitted at the microphone does not hold at your ears.
  • Modal ringing as a target. A minimum-phase cut matched to a room mode shortens its decay as a side effect, but nothing in the optimiser aims for decay, and the app does not yet show a decay view to confirm the shorter decay.
  • Seats other than the measured one. A quick calibration measures 5 microphone positions and a complete one measures 7 by default.

Phase Correction is the pass that brings the tuner's needles closer together after alignment, and it makes calibration take longer. Turn Phase Correction off to save calibration time when the needles already sit together after alignment.

05

How to Check the Alignment in Your Room

Re-measure after applying. In the tuner, show the sub and mains snapshots together. The two snapshots' needles should sit side by side and the verdict should read Locked. The summed response should no longer show a dip around the crossover. If the needles move apart after applying, something in the signal chain differs from what was measured.

Rules of Thumb

01Align speakers that share a band before correcting anything else.
02Hold back the speaker that arrives first.
03A metre of path difference is about 2.9 ms.
04At high frequencies, leave the room's phase alone.
05Removing a sealed box's bass group delay costs 31 ms of FIR latency.
06Re-measure after applying. The tuner's needles should sit closer together than before.

Frequently Asked Questions

Does NEXUS do phase correction?

NEXUS matches speakers' arrival times at the listener with per-output delay, polarity and optional all-pass filters through the shared band. NEXUS does not currently linearise single speakers. Matched arrival times between speakers are the goal, and delay and all-pass filters reach that goal without FIR latency or filter pre-ringing.

Why doesn't Omnissiah use FIR filters?

The goal is matched arrival times between speakers, and delay and all-pass filters reach that goal without an FIR filter. Linearising each speaker with FIR would cost 19 ms at an 80 Hz crossover and up to 82 ms for a ported box, with pre-ringing on transients, and a linear-phase crossover still dips as far as a normal one when the speakers are misaligned.

Which speaker is delayed?

The one that arrives first. A sub is usually late at the seat, so the mains are held back to meet it. The added latency equals the sub's lateness and applies to the delayed outputs only.

What does Time-Align Tops do?

With the default, All speakers, the sub is aligned to the mains and left and right are time-aligned to each other too. With Subwoofer only, just the sub is aligned and left and right are polarity-checked. Keep All speakers when left and right sit at noticeably different distances from you.

Should I leave Phase Correction on?

Yes, unless you are short on time. Check the tuner after your first calibration, and keep Phase Correction if it pulled the needles in.

Does alignment add latency?

Only the alignment delay, on the outputs that receive it. NEXUS runs at 0.85 ms analog in to analog out, plus any alignment delay on the outputs that receive it. Aligning a sub that is 3 ms late adds 3 ms to the mains.

What if my sub is more than 8.5 ms late?

NEXUS holds up to 8.5 ms of delay per output. Beyond that, alignment is skipped with a warning. Moving the sub closer brings it into range at 2.9 ms per metre. Through an 80 Hz crossover, flipping its polarity reaches an equivalent setting about 5.5 ms shorter, at about 1 dB.

Conclusion

Speakers that share a band are aligned with delay and polarity, measured at your seat, with an optional all-pass pass through the crossover. Level problems are corrected with minimum-phase filters. The goal is matched arrival times between speakers at the listener. A single speaker's group delay is not currently linearised: flattening the group delay does not fix a dip between misaligned speakers, an all-pass cannot flatten group delay, and an FIR would cost 19 to 82 ms and ring before transients. The processing stays at 0.85 ms analog in to analog out, plus any alignment delay on the outputs that receive it, and the tuner shows whether the needles came together after alignment.

Sign up to hear more when we release something new

Tantrum Audio logo in footer
Tantrum Audio Instagram Share Widget
Tantrum Audio Facebook Share Widget

© Tantrum Audio 2026

Content on this site may not be used to build competing software.