Measuring a Room
One measurement mic and a sweep show the peaks, nulls and decay before any money goes on treatment.
Related
Most rooms concentrate their low-frequency trouble in two or three places, and guessing by ear usually buys the wrong size of trap. The kit is modest: an omnidirectional measurement mic, an audio interface and a sweep.
Tools and signal
Use an omni measurement mic with its calibration file, mounted at ear height in the listening position and aimed between the speakers. Feed it a logarithmic sweep from 20 Hz to 20 kHz, a few seconds long, averaging more runs when the room is noisy. Switch off air conditioning and fans first, and make sure no processing is active in the interface, or the measurement describes the processing rather than the room.
Four plots, four questions
| Plot | What to read | What healthy looks like | Usual trouble |
|---|---|---|---|
| Response | Where peaks and dips sit | A few broad low peaks, a gentle top roll-off | Deep narrow dips are cancellation; do not boost |
| Waterfall | Decay against frequency | Low-frequency tails under a few hundred milliseconds | One frequency that will not fade |
| Reverberation time | Decay per band | Bands close together, bass slightly longer | Much longer bass means too little absorption |
| Impulse | Early reflections | A direct arrival, then a few weak reflections | A strong reflection a few milliseconds late |
Reverberation time comes from integrating the decay curve, following the method in ISO 3382-2. Bass usually decays a little slower than the midrange; a large gap points to insufficient low-frequency absorption, not to a bad EQ setting.
A repeatable procedure
Measure one speaker, then both. Measure the current layout, then move things ten centimetres and measure again. Take at least three points per position: centre, and twenty centimetres to either side. Keep the raw files rather than screenshots. Change one variable per round, otherwise nothing can be attributed.
Where reading the graph goes wrong
A single point turns one seat's null into the room's null. Third-octave smoothing hides narrow notches, while forty-eighth-octave smoothing lets noise back in. Too long a gate folds room reflections into the direct sound; too short a gate loses low-frequency resolution. An uncalibrated mic can be several decibels out at the top. The most overlooked point: the curve describes one microphone position, and moving your head ten centimetres changes it.
When not to measure
If you mix in one spot and have listened in the room for hundreds of hours, a measurement confirms rather than redesigns; take one and file it. In a headphone-first workflow a room measurement cannot answer headphone questions. Nor does it replace listening: a flat curve can still sound dull, because the time structure never appears on the plot.
Listen in the library
All tracks from midicn-lib — inline badge is the license tier (C1 commercial / C2 non-commercial / C3 study only).
Along the same line
Sources
- Everest, F. Alton，Master Handbook of Acoustics（McGraw-Hill）— 房间测量与脉冲响应
- ISO 3382-2 — 房间混响时间测量方法（标准）
- Room EQ Wizard，用户指南（厂商文档）— 扫频与瀑布图设置