How to read REW measurements

Half the confusing measurements posted on forums are confusing because of how they were taken, not because of the room. Three checks take a minute and save an evening.

1. 1.Smoothing off. REW defaults to some smoothing in the SPL view, and 1/3 octave smoothing hides exactly the narrow resonances you are hunting for. Set it to None while diagnosing, and only smooth afterwards to see the broad shape.
2. 2.Microphone calibration file loaded. A UMIK-1 without its calibration file reads several decibels wrong at the top and bottom of the range, and you will spend a week chasing a treble dip that is not there.
3. 3.Microphone at ear height, pointing where REW expects. A UMIK-1 uses a 90-degree calibration file when pointed at the ceiling and a 0-degree file when aimed at the speaker. Mixing those up puts a broad error across the top two octaves.

The single most useful thing to know about an in-room frequency response is that its left half and its right half mean different things and are fixed by different tools. The line between them is the Schroeder frequency.

Below it, the room's resonances are far enough apart in frequency to act individually. A peak is one resonance being loud. Above it, resonances overlap so densely that no individual one matters, and what you see instead is interference between the direct sound and early reflections. For a normal domestic room the crossover lands somewhere between 150 and 300 Hz.

• Below about 200 Hz: big peaks and deep dips, tens of decibels. These are real, they are the room, and they are audible. This is where nearly all the improvement is.
• Roughly 200 Hz to 500 Hz: the transition. Peaks are still worth attention; narrow dips are already becoming position-dependent.
• Above 500 Hz: fine ripple of a few decibels. Ignore the ripple entirely. Only the broad trend up here means anything, and it should tilt gently downward, not run flat.

Reverberation time is how long sound takes to decay by 60 dB. REW reports several estimates of it (T20, T30, EDT and Topt) because a small room rarely decays cleanly enough to measure the full 60 dB directly. Topt is REW's own best pick and is the one to read unless you have a reason not to.

Do not read the single number. Read the shape across frequency. A room whose decay climbs steeply towards the bass is the boomy room everyone recognises, and it is a completely different problem from a room whose decay is uniformly too long.

• Decay rising towards the bass, flat above 500 Hz: classic untreated small room. You need thick corner absorption, not more panels on the walls.
• Decay uniformly long everywhere: not enough absorption of any kind. Bare walls, hard floor, glass.
• Decay falling steeply above 2 kHz: too much thin absorption already. Foam and thin panels work above 500 Hz and do nothing below it, so a room full of them ends up dull on top with every bass problem intact.

A waterfall plot is the frequency response repeated over and over as time passes, stacked back into the screen. Its job is to show you which peaks are resonances that keep going after the signal stops, and which are simply loud.

That distinction matters, because a peak that decays at the same rate as everything around it is a level problem, and a peak that hangs on for hundreds of milliseconds after its neighbours have gone is a resonance. Both look identical on the frequency response.

This is the question behind most forum posts, and there are three cheap tests that answer it without any theory.

1. 1.Move the microphone. Take a second measurement half a metre away. Anything that changes substantially is the room. Anything that stays put is the speaker.
2. 2.Compare left and right. Two identical speakers in a symmetric room measure nearly the same. Where they diverge, the room is asymmetric: usually one speaker is closer to a side wall, or there is an opening on one side.
3. 3.Check the frequency against the room's dimensions. Divide 343 by twice each dimension in metres. Those three numbers are your fundamental resonances, and their multiples account for most of what happens below 200 Hz.

That third test is what this site automates, along with the boundary arithmetic that explains most of the deep dips. Feed it a measurement and the dimensions, and it will name the cause of each feature rather than leaving you to match numbers by hand.