Aligning More Than Three Speakers

With two speakers there is a pair to match. With six there are fifteen pairs, and aligning each to each would give contradictory answers. The practical method picks one reference and brings every channel into agreement with that. Since each channel matches the reference, any two channels are consistent through it, within the limit set out below.

A speaker cannot be made to arrive sooner, so the delay goes to the earlier ones. For speaker i at distance d, the delay is (the largest distance minus d) divided by 343 m/s. The farthest speaker gets zero. In the example at the top of this page the subwoofer is 1.10 m nearer than the farthest speakers and receives 3.21 ms, the left surround receives 2.04 ms and left and right receive none.

Each output has a maximum delay. The largest delay any speaker needs is the spread between the farthest and the nearest, which is 1.10 m or 3.21 ms in the example. If the spread exceeds the limit, the nearest speakers cannot be aligned to the farthest. The remedies are to move the nearest speaker back or to accept a partial alignment.

Processors differ in how much delay they allow per output, so check the processor's specification against the spread you measure. Work in metres first, since each metre of spread is 2.9 ms, and compare that figure with the limit.

A channel is matched to the reference where the two overlap in frequency. A stereo pair of tops shares the whole range above the crossover. A subwoofer shares a narrow band around the crossover with the tops, and a rear surround may share almost nothing with it.

Two channels aligned to the same reference are therefore consistent with each other only in the band where both overlap that reference. A sub and a rear surround that share no band are not directly constrained to each other, and nothing in the method makes them agree. For most listening that is fine, because speakers that play different frequencies do not add or cancel.

Phase can be matched with all-pass filters, which add delay that varies with frequency. In tools that may put all-pass filters on the reference as well as on the other channel, the reference has a fixed number of filters. When the pairs are fitted one after another against a shared reference, the first pair can use the reference's filters for its own best fit. Every later pair is then fitted against a reference that has already been reshaped for the first pair.

The result is order-dependent. In a probe of Omnissiah's fit, with the sub pair first the sub's residual fell from 44 to 3 degrees and the tops' from 64 to 61. With the tops first, the tops fell from 64 to 4 degrees and the sub from 44 to 23.

Fit the pair with the widest overlap first. A stereo pair of tops shares the whole range above the crossover, so a phase error there is audible across a wide band. A sub shares only a narrow band around the crossover, so there is less to lose by fitting it second. If your tools let you set the order, put the tops before the subwoofer.

Each sub aligned to the tops through the crossover can land on a different rung of the ladder, a different delay or a different polarity, each with its own small summation cost. The article on polarity, phase and delay covers those equivalent choices. Two subs that each match the tops well can still differ in phase from each other.

Check the subs' sum below the crossover on its own. Two subs whose delays differ by 12.5 ms cancel at 40 Hz, because 12.5 ms is half a period there, and a gap of 5.5 ms cancels at 91 Hz. Measure the subs together with the tops silent and look for a dip. One order is to align the subs to each other first and then align the pair of subs to the tops.

Every delay above comes from measurements at one position. Below about 500 Hz, where wavelengths are 69 cm or longer, the differences between speakers survive head movements of a few centimetres. A move to another seat changes the delays needed by up to 2.2 ms, as the table below shows. Above 1 to 2 kHz the phase there is dominated by reflections and changes with a few centimetres of head movement.

A large rig is more sensitive to where you sit than a stereo pair is, because its speakers surround the seat, so one move lengthens some paths and shortens others. Moving 0.5 m to the left of the measured seat in the example changes the delay each speaker needs by a different amount.

The stereo pair alone is already 1.44 ms out at this seat, and the surrounds and sub add errors in both directions, up to 2.2 ms. Align for the seat that matters most and treat every other seat as a compromise. If several seats matter equally, no alignment suits every one of those seats, and the delay choice is a compromise judged by listening.

NEXUS has three outputs and allows 8.5 ms on each, which is about 2.9 m of path. That covers a stereo pair and a subwoofer, so the pair-order and multi-sub points above apply to larger systems than NEXUS itself. When the spread of arrival times is larger than 8.5 ms, alignment is skipped and a warning is shown.