REW Analysis Tutorial: Measuring Room Response, Waterfall Plots, and Treatment Verification

Room EQ Wizard waterfall plot display showing modal decay behavior before and after bass trap installation.

Room EQ Wizard, known universally as REW, is a free software application that transforms a calibrated measurement microphone and a standard audio interface into a comprehensive room analysis system. It generates test signals, captures the room's response, and produces frequency response plots, waterfall diagrams, spectrograms, energy decay curves, and impulse response data. For anyone treating a listening room, REW provides the objective evidence needed to determine whether treatment is working or whether additional intervention is required. Guessing without measurement leads to wasted money on unnecessary panels and persistent problems that treatment could have solved if properly targeted.

The measurement chain consists of a calibrated measurement microphone such as the Dayton Audio EMM-6 at $74 or the miniDSP UMIK-1 at $99, an audio interface with at least one mic preamp with 48V phantom power, and REW running on any modern computer. The microphone connects to the interface's mic input, the interface connects to the computer via USB, and the interface's line outputs connect to the speaker system. A USB measurement microphone such as the UMIK-1 combines the microphone capsule and audio interface into a single device, simplifying the setup at a small premium in cost.

Calibrating the Measurement System

Before capturing room data, REW must know the frequency response of the measurement microphone so it can correct the measured data. Every calibrated microphone ships with an individual calibration file, typically a text file listing frequency and sensitivity offset pairs. In REW, navigate to Preferences, then the Soundcard Mic/Loopback tab, and load the calibration file. This step ensures that the frequency response plot displays the actual room response rather than the combined response of the room and microphone.

The audio interface latency must also be configured correctly. In REW's Preferences, set the output and input devices to the measurement interface, and adjust the buffer size to 256 or 512 samples. The timing reference signal, which REW uses to align the measurement capture with the test signal output, should use either a loopback cable from the interface output to a second input or the built-in acoustic timing reference available in REW version 5.20 and later. The acoustic timing reference plays a brief pulse before the measurement sweep, allowing REW to determine the acoustic delay and correctly time-align the impulse response.

Running a Sine Sweep Measurement

Sweep Duration and Resolution Settings

The sine sweep measurement is the most accurate method for capturing room impulse response. REW generates a logarithmic sine sweep from 20Hz to 20kHz over a duration of 256,000 or 512,000 samples, played through the speakers at approximately 75 to 80dB SPL at the measurement position. The measurement microphone captures the room's response, including the direct sound, early reflections, and the decaying reverberant field. REW then deconvolves the captured signal with the original sweep to produce the impulse response, from which all other analyses are derived.

Set the sweep length to at least 256k samples for adequate frequency resolution at low frequencies. A 256k sweep at 48kHz sample rate lasts approximately 5.3 seconds and provides frequency resolution of approximately 0.19Hz. For bass mode analysis, the 512k sweep (approximately 10.7 seconds, 0.09Hz resolution) is preferable because it separates closely spaced modes that would otherwise blur together in a shorter sweep. Set the output level so that the SPL at the measurement position is between 75 and 85dB, loud enough to overcome the room noise floor but not so loud as to cause speaker distortion or nonlinear room behavior.

Multi-Position Measurement Protocol

A single measurement at the exact listening position provides useful data but does not represent the listening experience across the full head-movement area. I recommend capturing measurements at five positions: the center listening position, and positions 15cm and 30cm to the left and right. Averaging these five measurements produces a spatially smoothed frequency response that better represents what the listener perceives during normal listening. The variation between positions reveals the spatial consistency of the room's response, with large variation indicating poorly damped room modes.

Position the microphone at ear height, approximately 1.2m above the floor for seated listening. Use a microphone stand that does not reflect sound toward the microphone capsule. The microphone should point vertically upward, not toward the speakers, because measurement microphones are typically calibrated for 0-degree incidence and the diffuse field response in a room approximates random incidence conditions. The miniDSP UMIK-1 is calibrated for 0-degree incidence and performs adequately in rooms when pointed upward, capturing a representative mix of direct and reflected sound.

Interpreting the Frequency Response Plot

The frequency response plot displays the sound pressure level at each frequency measured at the microphone position. Below the Schroeder frequency, the response is dominated by room modes and varies dramatically with position. Above the Schroeder frequency, the response reflects the speakers' direct output plus the room's average absorption characteristics. The target for a well-treated room is a response that remains within plus or minus 4dB from 80Hz to 8kHz, with no single-frequency peaks or dips exceeding 6dB relative to the average level in the surrounding frequency range.

Pay particular attention to the region between 60Hz and 200Hz, where room modes produce the most severe response irregularities. A peak of 8dB or more at any frequency in this range indicates an undamped mode that requires additional bass trapping at the corresponding pressure maximum. A dip of 6dB or more may indicate a listening position at a mode node, which can be partially corrected by repositioning the seat, but may also indicate a boundary cancellation that requires speaker repositioning relative to the nearest wall.

Reading Waterfall Plots and Energy Decay Curves

The waterfall plot displays the room's frequency response over time, with the third axis representing decay duration. Peaks that persist long after the initial impulse indicate resonant modes with high Q factors that continue ringing. In an untreated room, the waterfall shows prominent ridges at each axial mode frequency, extending 500 to 800 milliseconds beyond the impulse. After bass trap installation, these ridges should shorten to 250 to 400 milliseconds, indicating that the traps are effectively damping the modal energy.

The Energy Decay Curve (EDC) derived from the impulse response shows the cumulative decay of energy at each frequency. The slope of the EDC at any frequency indicates the decay rate: a steeper slope means faster decay and better modal damping. The EDC is particularly useful for identifying frequencies where decay is non-uniform, such as when two coincident modes produce a double-slope decay curve with an initial fast decay followed by a slower tail. This pattern indicates that one mode is adequately damped while the coincident partner remains under-treated.

When I teach room measurement at MIT, the first thing I tell students is that the frequency response plot tells only half the story. The waterfall plot and EDC reveal the time-domain behavior that determines whether bass feels tight and controlled or bloated and smeared. Two rooms can have identical frequency response curves but dramatically different waterfall plots, and the room with the shorter decay times will consistently produce better mixing results because the engineer hears the source material without the room's resonant tail coloring every bass note.

Using the All RT60 Function

Frequency-Band RT60 Analysis

REW's All RT60 function computes the reverberation time across frequency bands from the impulse response. This feature provides an immediate overview of the room's decay behavior without requiring manual analysis of individual frequency bands. The resulting plot shows RT60 values at each octave or third-octave band center frequency from 63Hz to 8kHz. In a well-treated room, this plot should be relatively flat, with values between 0.25 and 0.40 seconds across all bands.

A rising RT60 curve at low frequencies indicates insufficient bass trapping. The degree of rise quantifies the treatment deficit: if RT60 at 63Hz is 0.80 seconds and RT60 at 500Hz is 0.30 seconds, the 2.7-to-1 ratio indicates substantial low-frequency energy accumulation. Adding corner bass traps and membrane absorbers should reduce the 63Hz RT60 toward 0.40 seconds, flattening the curve. Re-measure after each treatment addition to track progress and avoid over-treating.

Generating Before-and-After Comparison Reports

Overlay and Export Workflow

REW's overlay feature allows multiple measurements to be displayed on the same graph, making before-and-after comparison straightforward. Capture a baseline measurement before installing any treatment, then repeat the measurement after each treatment phase with the microphone in exactly the same position. Overlay the frequency response plots to see the change in spectral balance. Overlay the waterfall plots to see the change in decay times. The visual difference provides immediate feedback on treatment effectiveness and guides further intervention decisions.

For documentation purposes, REW can export measurement data as PNG images, CSV data files, or REW measurement data files (.mdat) that preserve all analysis results. Maintaining a dated series of mdat files creates a treatment history that shows the cumulative effect of all interventions and serves as a reference for future room modifications. This documentation also proves valuable when consulting with acoustic professionals who can analyze the data remotely and provide specific treatment recommendations based on measured behavior rather than descriptions.