openmixerGitHub →

<!-- SPDX-License-Identifier: GPL-3.0-or-later -->

One summing bus

openmixer has exactly one way of adding signals together, and every bus on the console — MAIN, a group, an aux, a matrix — is the same operation applied to a different set of sources. There is no separate "master bus" mechanism sitting beside a separate "aux bus" mechanism; there is one operation, instantiated as many times as the console has buses.

The one operation

A bus is a weighted sum of its sources: each source contributes its signal multiplied by a coefficient, and the bus is the sum of those products, sample by sample.

bus = Σ  coefficient(source) × signal(source)

That is the whole idea. Everything else a bus does — a fader, a send, a mute, a pan, a crosspoint — is a name for one of the two things in that sum:

  • A coefficient is a number that scales one source's contribution to one bus. A fader is a coefficient. A send level is a coefficient. A mute is a coefficient forced to zero without disturbing the number underneath it, so un-muting returns exactly where you left it.
  • A tap is which point in a channel's signal path is being read for a given contribution — the same channel can feed one bus post-fader and another pre-fader, because a bus reads a point, not "the channel" as a single wire.

MAIN is not a special case. MAIN is the bus whose fader realises the coefficient. The motorised fader under your hand is not a different thing from an aux send level; it is the same coefficient, given a dedicated physical control because the desk is built around mixing to MAIN by default. A group, an aux and a matrix use the identical operation with a stored, per-source coefficient instead of one riding the channel fader.

Because it is one operation, the things that look like separate features are consequences of it rather than extra mechanisms:

  • A send is a channel's coefficient into a bus other than MAIN.
  • Sends-on-faders is the fader wall temporarily displaying and driving that coefficient instead of MAIN's — the same fader, pointed at a different sum.
  • A matrix crosspoint is a coefficient where the source is itself a bus rather than a channel: matrices sum buses (and sometimes channels) the same way a group sums channels.
  • A mute never deletes the coefficient it silences; it sets the contribution to zero and remembers what it was zeroing, which is why release restores the fader exactly.

One operation, one place it is implemented, and every bus on the desk — however many the console allocates — is an instance of it.

Power, amplitude and loudness

Three different things get called "level," and mixing them up is the root of most decibel confusion.

Amplitude is the size of the waveform itself — the voltage on a cable, the number stored for each audio sample. It is what a fader coefficient multiplies.

Power is proportional to amplitude squared — it is what heats a loudspeaker coil and what an energy meter reads. Doubling amplitude quadruples power, because doubling a number and squaring the result multiplies it by four.

Loudness is what a listener actually perceives, and it grows more slowly than either — ears are not linear meters. A change that measures as a clean doubling in amplitude or in power does not sound twice as loud; it takes a considerably bigger change than that before most listeners agree something is "twice as loud."

The decibel exists to make one ratio out of these different scales. It is always a ratio, never a quantity on its own — "+6 dB" means nothing without something to be six decibels more than. Because power is amplitude squared, the same physical change reads as the same number of decibels down either scale, provided you square the ratio you drop into the amplitude formula: amplitude uses 20·log10(ratio), power uses 10·log10(ratio), and a doubling of amplitude is a quadrupling of power — so both arrive at the same six decibels. dBFS, the unit meters on the desk read in, is exactly this ratio, measured against digital full scale (the loudest sample value the format can represent) rather than against another signal.

Doubling Reads as
×2 amplitude +6 dB
×2 power +3 dB
×2 perceived loudness ≈ +10 dB

The three doublings

Read against that table, three different-looking rules of thumb turn out to be the same one idea, applied to the thing that is actually doubling:

  • +3 dB doubles power, when two unrelated signals combine — two different musicians' microphones, or any two sources whose waveforms are not alike from moment to moment. Their powers add; their amplitudes do not, because unrelated signals do not consistently reinforce each other.
  • +6 dB doubles amplitude, when two identical signals combine — the same signal arriving twice, in step. This is why a stereo pair carrying mono content (the same signal on both legs, dead centre) sums to twice the amplitude, not the square-root-of-two you would get from two unrelated sources: the two legs are perfectly correlated, so they add like amplitudes, not like powers, and the result is 6 dB hotter than either leg alone — not 3.
  • +10 dB doubles perceived loudness, roughly and empirically. This one is not derived from the sum at all; it is a property of hearing, folded in here only because it is the number people reach for and confuse with the other two.

Why the mono fold and the pan law are fixed numbers, not settings

Two controls on the desk look like they are missing a knob, and neither is.

The mono fold has no level control. When a destination is a single socket — a mono fill speaker, a hearing-assist feed — the desk has to turn a stereo pair into one signal, and stereo content is, in the case that matters, the same signal on both legs (a mix that was built to sit centred). Summing two identical legs is the +6 dB case above: without correction, folding to mono makes a centred mix six decibels hotter than it was in stereo. The fold applies a fixed −6 dB to bring a correlated pair back to the same amplitude it had on either leg alone. That −6 dB is not a mix decision; it is the exact number that undoes a doubling, so there is nothing to set — a control there would only let someone dial in the wrong answer to a question arithmetic has already settled.

The pan law's centre position is −3 dB per leg, for the matching reason on the power side. A pan control distributes one source across two legs by power, not by amplitude, so that a sound panned hard left and the same sound panned hard right are equally loud — and so that sweeping a pan knob across the stereo field does not swell or dip in level as it passes through the middle. At centre, the source's power is split evenly between the two legs: half the power in each leg is −3 dB down from the source's full power, and the two legs sum back to the source's original power. Any other split at centre would either lose level in the middle of a pan sweep or gain it, and both are audible defects, not stylistic choices. Constant-power panning is the one split that keeps a pan move level-neutral, which is why it is the law and not an option.

See also