REW Car Audio Guide: Beginner to Advanced Measurement

REW (Room EQ Wizard) is the free measurement software that turns a $109 USB microphone and a laptop into a competition-grade car audio analyzer. Frequency response, impulse response, group delay, waterfall decay, and distortion plots are all available from one interface. This guide walks the full workflow I use on every customer tune at the shop, from first sweep to advanced time-alignment work.

What REW Is, and Why Every Serious Car Audio Tuner Runs It

REW, short for Room EQ Wizard, is acoustic measurement software written by John Mulcahy and given away free at roomeqwizard.com. It generates a logarithmic frequency sweep, captures the microphone response through a calibrated USB or interface input, and produces frequency response, impulse response, group delay, distortion, and decay plots. None of these plots are unique to REW. What is unique is having them all in one application that costs nothing and runs on the laptop you already own.

A phone RTA app shows you a real-time bar graph of pink-noise content in roughly third-octave resolution. REW does swept-sine measurement, which separates the speaker's direct sound from the room's reflections with a resolution a real-time analyzer cannot match. The difference matters most below 200 Hz and in the impulse-response domain, which is exactly where car audio measurement lives.

REW is step zero of the workflow we cover in the complete car audio DSP tuning guide. You cannot apply parametric EQ correctly without first measuring what the cabin is doing. You cannot set time alignment correctly without seeing the impulse response. REW is the instrument; the DSP is the tool that acts on what REW shows you.

The Minimum Hardware Stack for Car Audio Measurement

Three pieces: a laptop running Windows, macOS, or Linux; a calibrated measurement microphone; and a way to play the sweep through the car's source unit. Most builds use a 3.5 mm aux cable or a USB stick with a generated WAV file loaded onto the head unit.

The microphone is the only part you cannot cheap out on. The miniDSP UMIK-1 is the entry-level industry standard: $109, USB-powered, ships with an individual calibration file matched to its serial number. Loaded in REW, that cal file pulls the mic to ±1 dB flat across the car audio band. The iSEMcon EMX-7150 is the step up at roughly $300 with USB or XLR options and tighter off-axis response, which is why it shows up at EMMA and IASCA judging tables. The Dayton Audio EMM-6 sits between them at around $80.

Phone RTA apps run on uncalibrated MEMS mics with rolled-off bass and unpredictable response above 8 kHz. They are useful for spotting a 10 dB problem; they are useless for setting parametric EQ to the nearest dB.

First-Run Setup: Calibration File, Sweep Length, and Output Routing

Install REW from roomeqwizard.com. On first launch, three settings matter and the defaults are wrong for car audio. Get them right once and you can save them to a preferences file.

Load the mic calibration file. Open Preferences, then Mic/Meter. Click "Browse" next to "Cal file" and load the .txt or .frd file that came with your UMIK-1 (download from miniDSP using the serial number printed on the mic body). Without this file loaded, your trace shows the mic's response curve added to the car's, and every EQ correction is wrong by 1 to 3 dB at the extremes.

Set sweep length to 256k or 512k samples. REW's default is 256k, which is fine for full-range work. For bass-only measurements where you need 1 Hz resolution to spot a port-tuning issue or a room mode, push it to 512k or 1M. Longer sweeps take longer to complete but resolve more bass detail.

Set sweep output level to −12 dBFS. The default of 0 dBFS will clip the DSP input on most systems. Drop to −12 dBFS for the first measurement, raise only if the captured signal is below the noise floor. A working sweep peaks at 75 to 85 dB SPL at the listening position. Anything louder is fatigue without benefit.

Taking Your First Measurement (Beginner Workflow)

The first sweep takes ten minutes start to finish. Get this part right and the rest of the work in REW is just looking at different views of the same captured data.

1. Position the mic at the driver's ear. Use a stand or boom set to your seated ear height, mic capsule pointed straight up (omnidirectional, so orientation matters only for high-frequency off-axis response). The center of the headrest is the wrong spot. Move it forward to where your actual head sits when you drive.
2. Run an SPL calibration. Click the SPL meter icon in REW. Play the calibration signal at the source. Set the SPL meter to read 75 dB on your handheld SPL meter or phone SPL app. Now your trace's y-axis reads in actual dB SPL, not arbitrary level.
3. Set the sweep range to 10 Hz to 20 kHz. For sub-only measurements drop the upper limit to 200 Hz so the sweep doesn't waste resolution on frequencies you are not measuring.
4. Hit Measure. Stay still and quiet for the duration of the sweep, roughly 8 to 15 seconds depending on length. Door slam, AC fan, or a passing car ruins the measurement and you have to redo it.
5. Save the file. Use a naming convention like vehicle_date_position_speaker.mdat. You will reference these later when comparing before/after EQ. For the EQ work that follows the first measurement, our car audio DSP step-by-step EQ guide picks up here.

Reading the Frequency Response Trace (Intermediate)

A raw trace from a car cabin is ugly. Below 80 Hz you will see 10 to 15 dB of cabin gain. Across 100 Hz to 1 kHz you will see narrow peaks and dips from reflections off the dash, glass, and door cards. Above 5 kHz you will see steep rolloff from speaker directivity off-axis. None of this is a problem with the speakers. It is the cabin.

Set smoothing to 1/6 octave. Right-click the graph, then choose 1/6 smoothing. The trace becomes readable without losing the features that EQ needs to address. Use 1/3 for big-picture decisions, 1/12 for distortion hunting, and no smoothing only when you specifically need to see raw measurement noise.

Use a downward target curve. Human hearing expects a tilt of roughly −0.5 to −1 dB per octave from bass to treble. A flat trace sounds thin and harsh. REW's target curve overlay (Controls panel) lets you load a reference like the Harman in-car curve and visually align your trace to it. Correct deviations of 3 dB or more from the target. Leave smaller features alone.

Cuts before boosts. A 6 dB peak at 250 Hz gets a 6 dB parametric cut. Do not raise surrounding frequencies to flatten the peak from below; that costs amp headroom and pushes drivers harder. The right Q value comes from REW's auto-EQ filter generator: narrow Q for narrow peaks (Q 6 to 10 for sub-200 Hz resonances), wide Q for broad shelves (Q 0.7 to 1.4 for vocal-band tilt).

Impulse Response and ETC for Time Alignment (Advanced)

Time alignment is the highest-value adjustment in any car audio DSP, and REW is the only tool that shows you exactly how far off you are. Measure each speaker channel separately (mute everything else in the DSP), then switch to the Impulse view in REW's All SPL graph.

The impulse response shows the arrival time of the sweep at the mic, in samples or milliseconds. Hover the cursor over the first peak of each channel's impulse trace and REW displays the time at the cursor in the bottom-right corner. Subtract the closer driver's arrival time from the farthest driver's arrival time and you have the delay value in milliseconds that needs to go on the closer driver.

The farthest driver is the reference at 0 ms. Closer drivers get delayed to match it. If the right tweeter arrives 1.2 ms before the left tweeter at the driver's ear, the right tweeter gets 1.2 ms of delay added in the DSP. Subwoofer-to-front alignment uses the same logic with one extra step: invert sub polarity, sweep again, and use the polarity that produces the deepest null at the crossover frequency. That null tells you the two drivers are 180 degrees out of phase, which means the correct polarity is the opposite of the one that produced it. For the math without the measurement, the car audio time alignment calculator works out the millisecond targets from speaker distances alone.

Use the ETC (energy-time curve) view for late reflections. The first peak is the direct arrival. Peaks 1 to 5 ms after it are early reflections off the dash and pillar. Peaks 10 ms or later are noise. ETC is what tells you whether a treatment material on the A-pillar is doing anything measurable, beyond what you can hear.

Group Delay, Waterfall, and Distortion (Advanced)

Three more views in REW that pay off once frequency response and time alignment are done. None of them are first-pass tools; all of them catch problems the FR trace alone cannot.

Group delay shows time arrival vs frequency on a single trace. A flat or gently rising line across the audible band is correct. A sharp spike at a crossover frequency means the high-pass and low-pass slopes are not aligned in phase. Drop the spike below 5 ms and the imaging tightens audibly. This is one of the few places where adding a small (0.1 to 0.3 ms) all-pass delay to one side of the crossover can correct what looks like an EQ problem but is actually a phase issue.

Waterfall plots amplitude vs frequency vs time, showing how quickly each frequency decays after the signal stops. Long ridges below 200 Hz are room modes and panel resonances. A 100 Hz peak that persists 200 ms after the signal ends is a door card resonance you can hear as boom. The fix is mechanical (mass-load the panel or apply CLD damping); EQ cannot correct it without overshooting.

Distortion measurement requires the same sweep, viewed under the Distortion tab. REW separates the fundamental from 2nd, 3rd, and higher harmonics. A driver running 3% THD at its rated wattage at 60 Hz is fine. A tweeter producing 5% 3rd-order distortion at 4 kHz at 90 dB SPL is mechanically stressed and one volume click from failure. Distortion data is also how you find drivers that are out of spec from the factory before you tune around their flaws (Crutchfield).

Mistakes That Make REW Lie to You

Running with the cal file unloaded. The most common first-time error. The trace looks plausible but it bakes in 2 to 3 dB of mic error. Always verify Preferences > Mic/Meter shows the active cal file before measuring.

Measuring with doors or windows open. Car audio measurement is a sealed-cabin measurement. Open a door and the cabin gain disappears, the trace shifts 10 dB at 40 Hz, and you will EQ in bass that vanishes when the doors close. Doors shut, windows up, engine off, climate fan off.

One-position EQ. A single mic position gives you the curve at that exact point in space. Move the mic 4 inches and a 6 dB null can shift 30 Hz. For final EQ work, take three measurements (left ear, right ear, between them) and average them in REW's Vector Average function. Tune to the average, not the worst-case single trace.

Frequently Asked Questions

REW is the measurement leg of the workflow; the rest of the tune lives in the DSP. For the complete picture of how the measurements feed into gain structure, crossovers, time alignment, and EQ, the complete car audio DSP tuning guide walks every step in the same order I run them on a customer car. If you want me to bring the mic and the laptop to your build, reach Scott through the contact page.