Measuring and aligning sound
Sound takes time to travel and loses level on the way. Nearly every decision a system engineer makes follows from those two sentences.
The physics underneath
The inverse square law, and why it is 6 dB
A source radiates into an expanding sphere. Double the distance and the same energy is spread over four times the area, so intensity drops to a quarter. A quarter of the power is −6 dB, and that is the whole rule: every doubling of distance costs 6 dB.
This is why a spec sheet figure is meaningless without its measurement distance. "This box does 100 dB" is not a claim until someone says at what distance — 100 dB at 1 m is 70 dB by 30 m.
Because sound is slow
How to time-align a delay speaker
Sound travels about 343 m/s at 20 °C. A delay tower 40 m downstage of the main hangs is therefore hearing the mains about 117 ms after they fire — and if the tower plays at the same instant, the audience under it hears the tower first and the mains as an echo.
The fix is to delay the nearer speaker by the difference in travel time, so both arrive together. Measure the physical distance between the two sources, divide by the speed of sound, and that is your delay time.
Temperature matters more than people expect: the speed of sound rises roughly 0.6 m/s per °C. An alignment set in a cold empty room at 10 am can be measurably out by the time a full, warm house is in.
- Aligning at one seat misaligns at another. Alignment is a compromise across the coverage area, so the measurement position is a decision, not a convenience.
- Many engineers add 5–15 ms beyond the physical figure. The precedence effect then pulls the image back towards the stage rather than the nearby tower. That is a taste decision, not an error.
- Processing delay is not free. Every converter, DSP and network hop adds latency of its own — budget it with the latency tool rather than assuming the electronics are instant.
The measurement itself
How a transfer function is actually taken
Dual-channel FFT measurement — Smaart, Open Sound Meter or REW — compares two signals: a reference taken from the signal you are sending, and a measurement taken from a microphone in the room. The software divides one by the other, and what is left is what the system and the room did to the signal — magnitude, phase, and impulse response.
The third trace is the one to trust first. Coherence says how much of what the microphone heard is actually related to the reference. Where coherence is low — because of noise, reflections, or too little energy at that frequency — the magnitude trace above it is not evidence and should not be equalised against.
- Take the reference before the processing you want to measure. A reference tapped after the EQ measures everything except the EQ.
- Low coherence means "do not act on this". It is not a display problem to be smoothed away.
- The microphone position is the measurement. Two engineers three metres apart get different and equally correct answers. Class 1 measurement microphones are specified by IEC 61672-1.
- Fix time before frequency. Find the arrival time, set the delay, then look at the magnitude — an unaligned measurement shows comb filtering that no EQ can remove.
The room is the biggest processor in the system
Reverberation time, and what each use wants
Every venue has a decay time, and it is doing more to the sound than any device in the rack. RT60 is how long a sound takes to fall by 60 dB after it stops, and it is set by two things only: the volume of the room, and how much absorption is in it.
Sabine put it as RT60 = 0.161 × V / A — volume in cubic metres, absorption in square metres of equivalent open window. So a big room is a long room, and adding absorption is the only lever. An audience is absorption, which is why an empty venue sounds nothing like a full one and why a soundcheck flatters everything.
What is good depends entirely on what the room is for, and the requirements genuinely conflict. Speech needs the room to get out of the way. Orchestral music needs the room to be part of the instrument. A room asked to do both will do neither well unless something in it can change.
- These are mid-frequency targets and a single number hides a lot. A room can measure 1.2 s at 1 kHz and 2.4 s at 125 Hz, and the second number is what makes it muddy. Ask for the curve, not the figure.
- An audience is absorption, and a variable one. Design for the condition you actually perform in, and expect the empty-room rehearsal to be a different venue.
- You cannot subtract reverberation. A PA can add level and change coverage; it cannot remove what the room does. Directivity that keeps energy off the surfaces is the only real tool, which is a large part of why pattern control and array design matter so much.
- Variable acoustics exist for a reason. Retractable banners, rotating panels, coupled chambers and electronic enhancement all exist because one room genuinely cannot serve speech and symphony at the same RT.
Which box, and why
Point source against line array
A point source radiates into a sphere: energy spreads in every direction and you get the full 6 dB per doubling. A line array, while it is long compared with the wavelength it is reproducing, radiates more like a cylinder — the energy spreads sideways but much less vertically, so the loss is closer to 3 dB per doubling in that region.
Prediction software — Soundvision, EASE Focus, ArrayCalc — models this before anything is hung. That is the entire reason arrays exist for long throws: the front row and the back of an arena can be brought within a survivable level range. It is also why arrays are wrong for short rooms — the cylindrical behaviour needs distance to develop, and a short array in a small venue is just an awkward point source.
The behaviour is frequency-dependent, too. An array that is long compared with a 200 Hz wavelength may be short compared with a 40 Hz one, which is why array low end behaves differently from array high end.
Speaker delay with temperature correction, latency budget, speaker impedance and noise exposure dose are all on the field tools page and work offline. Measurement software is indexed under software.
2 questions on the mechanism rather than the trivia. Nothing is scored and you can answer again — the explanation under a wrong answer is the point of it.
-
1A hall measures RT60 of 2.2 seconds. What is it well suited to?
-
Right. Choral and organ repertoire is written for long reverberation — the tail is part of the sound, and 2.2-3.0 s is the band it sits in.
-
Drama wants roughly 0.8-1.1 s. At 2.2 s consonants get buried in the tail of the syllable before.
-
Speech intelligibility wants around 0.6-0.9 s. This hall would need heavy treatment or a very directional system.
-
Around 1.0-1.4 s suits amplified work. At 2.2 s the reverberant tail smears rhythm and intelligibility.
-
-
2A full audience arrives and the room sounds noticeably drier than in rehearsal. Why?
-
The room itself has changed, not the system gain.
-
Air absorption is real at high frequencies and over long distances, but it is a much smaller effect than the bodies themselves.
-
Temperature does change the speed of sound slightly, which affects timing rather than reverberation time.
-
Right. RT60 is inversely proportional to total absorption, and a few hundred clothed bodies are a large number of sabins. This is why an empty-room measurement flatters the room.
-
Answers are kept in this browser on this device. Nothing is sent anywhere.
- How do I measure a system?
- How do I time a delay tower?
- Point source or array?
Where to go next
These pages are a chain rather than a library — the next one along usually answers the question this one raised.
Everything on this page, in the index
Each of these has its own entry with sources, ports, gotchas and an open gap where a field is missing.