Wow & Flutter Meter
Play the 3150 Hz band of a test record or tape and watch the speed error and weighted wow and flutter change as you adjust the machine, with a spectrum that points to the cause.
Meter
Uses your microphone, line input or the sound you share; it is analysed on your device and is not recorded or uploaded.
Choose the test tone and the machine, press Start and play the test track.
| Reading | 2σ (%) | Peak (%) | RMS (%) |
|---|---|---|---|
| WeightedAES6 curve | – | – | – |
| Unweighted0.2–200 Hz | – | – | – |
| Wow0.2–6 Hz | – | – | – |
| Flutter6–200 Hz | – | – | – |
Mean frequency, drift and the time covered appear here.
Speed over time
Speed error of every millisecond against the nominal speed; the dashed line is the mean.
Flutter spectrum
Size of each regular speed variation, by how many times a second it repeats (unweighted).
- The likely causes of the largest peaks appear here.
Speed around one platter turn
Average speed at each point of a turn: outside the circle is fast, inside is slow. 0° is the moment you started, or the last time you pressed Set 0°.
Test tone and self-test
The tone is at the test tone frequency chosen in the meter: 3150 Hz.
Test this meter sends the chosen signal (a modulated one; the plain tone becomes the 4 Hz test) through the meter above and compares the readings with what the AES6 curve predicts.
| Reading | Expected | Measured |
|---|
Runs on your device. Uses your microphone or the sound of another tab or the screen; the audio is processed on your device. Nothing you type or record is uploaded. Keeps working without a connection while the page is open. Works with touch. Saves the test tone, machine, reading window, input device and generator settings on this device.
Added
How to measure
- A test tone. A test record or test cassette with a 3150 Hz band (3000 Hz on those made to the JIS or NAB methods) is best, because its own speed was checked when it was made. Without one, record the plain tone from Test tone and self-test onto your tape deck and play it back; read Limits first.
- The tone into the computer. Use one of the four ways below. Keep the level bar between its two marks; the status line warns when the tone is weak or the input clips.
- Start. Choose the test tone and the machine, press Start and play the tone. The speed shows after one platter turn, wow and flutter about six seconds after the tone is found.
- Adjust. Turn the speed trimmer until Speed reads close to 0 %. After a change to the drive (a new belt, oil, a cleaned pinch roller), press Restart readings so the figures cover only the new state.
Getting the tone in
- A cable. Connect the line output of the tape deck, the turntable preamplifier or the amplifier’s recording output to the line or microphone input of the computer, then press Start: microphone or line in. With several inputs, a list appears to pick one. No room noise gets in, so this is the most accurate way.
- A microphone. Hold the phone or computer still, close to one loudspeaker, in a quiet room. Room noise raises the unweighted figures more than the weighted one.
- Computer audio. If the test track plays in another program or browser tab on this computer, press Computer audio and share that tab, or the whole screen with its sound. Desktop browsers only.
- A recording. Record the test track at 44.1 or 48 kHz and open it with Open audio file (WAV, FLAC, MP3). A file is analysed faster than it plays.
Reading the results
- Speed: how far the mean tone frequency is from the nominal one; plus means fast. On a turntable it is averaged over the last full turn, so a once-per-turn wobble cancels out, and also shown in rpm; on tape over the last 2 seconds.
- Wow & flutter, weighted 2σ: the main figure of the AES6 standard. The speed variation goes through a weighting filter that gives full weight to a 4 Hz wobble, which hearing notices most, and less to slower and faster ones: about −10 dB for a once-per-turn wobble at 33⅓ rpm (0.556 Hz) and −23 dB at 200 Hz. 2σ is the value the weighted variation exceeds, upwards or downwards, for 5 % of the time.
- Peak and RMS: the largest value in the window and the root mean square. Weighted RMS is the figure of meters built to the Japanese JIS method.
- Unweighted, wow and flutter: the real speed variation from 0.2 to 200 Hz without the hearing filter, then split at 6 Hz into slow wow and fast flutter. Watch these when you look for a mechanical fault.
- Drift: the line under the table gives how much the speed averaged over each turn (each 2 s on tape) moved within the window.
- Readings over: the last 5, 10 or 30 seconds, or the whole run (up to 10 minutes, without the first 4 seconds after the tone is found). The window ends 2 seconds before now, because the filters need a little signal on both sides of every moment.
Finding the cause
A fault that repeats at a steady rate shows as a peak in the flutter spectrum; under the chart the page names the usual suspects, as a guide, not a diagnosis.
- Once per platter turn (0.556 Hz at 33⅓, 0.75 Hz at 45 rpm): a record whose hole is off centre, a bent spindle or a platter that runs out of true. In the plot of one turn the trace bulges out on one side and in on the other.
- Twice per turn: often a warped record. As the arm rides up and down over the warp, the stylus moves a little forwards and backwards along the groove. The trace becomes an oval.
- Faster, steady peaks: the drive. A part that turns, or has a bump that passes, as many times a second as the peak frequency is the first suspect: motor pulley, belt joint, idler wheel, or on tape the capstan and pinch roller. A capstan of diameter d turns v / (π d) times a second at tape speed v: at 4.76 cm/s a capstan 2.4 mm across turns 6.3 times a second.
Record or turntable? Put a small piece of tape on the platter rim. Press Set 0° now as it passes the tonearm and note where the bulge in the plot of one turn is. Lift the record, turn it half a turn on the platter, and repeat with the same mark. If the bulge moved by half a turn, the record is off centre; if it stayed, look at the platter, spindle and bearing.
How the meter works
The input is multiplied by a complex oscillator at the tone frequency fLO and low-pass filtered at 300 Hz, which leaves a slowly turning pointer. Its change of angle Δφ over each step Δt of about 1 ms gives the tone frequency:
f = fLO + Δφ / (2π Δt)
The oscillator follows the mean frequency slowly; the delay of the filter is taken into account. Speed error = f̄ / f0 − 1, where f̄ is the mean frequency and f0 the nominal tone; a turntable’s speed in rpm is the nominal rpm × (1 + speed error). The variation is d(t) = f(t) / f̄ − 1, so wow and flutter are percentages of the mean speed.
The weighting and the bands are applied to d with a Fourier transform, together with a correction for the 300 Hz filter and for averaging over each 1 ms step. The weighting follows table 1 of AES6-2008 at its 17 frequencies and joins them with a smooth curve on a logarithmic frequency scale (monotone cubic interpolation):
| Hz | 0.1 | 0.2 | 0.315 | 0.4 | 0.63 | 0.8 | 1 | 1.6 | 2 | 4 | 6.3 | 10 | 20 | 40 | 63 | 100 | 200 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| dB | −48.0 | −30.6 | −19.7 | −15.0 | −8.4 | −6.0 | −4.2 | −1.8 | −0.9 | 0.0 | −0.9 | −2.1 | −5.9 | −10.4 | −14.2 | −17.3 | −23.0 |
2σ is the 95th percentile of |weighted d|. For a variation that is a pure sine wave of peak A, 2σ = A sin 85.5° = 0.9969 A and RMS = A / √2 = 0.7071 A.
Example. Choose 4 Hz wobble, ±0.1 % and press Test this meter. The curve is 0 dB at 4 Hz, so expect a weighted 2σ of ±0.0997 % and a weighted RMS of 0.0707 %. With Once per 33⅓ turn (0.5556 Hz) the curve gives −9.98 dB (a factor of 0.3171), so the weighted 2σ is 0.2 × 0.9969 × 0.3171 = ±0.0632 %, while the unweighted RMS stays 0.2 / √2 = 0.1414 %.
Limits
- The sound card’s clock sets the time scale, so its error adds straight to the speed reading: a clock 50 parts per million fast makes every tone read 0.005 % slow. Wow and flutter are percentages of the mean, so the clock does not change them.
- A tone you recorded yourself measures recording and playback together. Their speed errors add, and the wow and flutter of both passes combine, so the figure is higher than that of the deck alone.
- The Peak column is the largest value in the window. The older IEC 386 and DIN 45507 meters used a slow quasi-peak needle instead, which AES6-2008 replaced with the 2σ method, so an old meter can read differently on irregular flutter. On a steady sine-wave wobble both read close to its peak.
- Between the 17 table frequencies the weighting depends on the curve used to join them. At 0.5 Hz this page uses −11.46 dB; the analog filter design published with the same table gives −11.48 dB.
- The tone is searched within 10 % of the chosen frequency. A machine that is further off, or a test track that is not a steady tone, is not measured.
- Dropouts: missing samples make the tone jump; a jump of more than 1.5 % within milliseconds is left out and counted under the table.
- Tested on synthetic tones with a known wobble (4, 0.5 and 20 Hz, and 1 % fast) on 14 November 2026: through a file and through a simulated microphone the readings were within 1 % of the expected values and the speed within 0.005 %.
Sources
- Audio Engineering Society. AES6-2008 (r2012), the AES standard for measuring weighted peak flutter. The 3150 Hz test tone, the weighting filter and the 2σ (statistical) reading that replaces the quasi-peak meter.
- AES6-Wow-and-Flutter, an open-source project: notes on the AES6 weighting filter. A transcription of AES6-2008 table 1 (the 17 values above, relative to 4 Hz) and an analog filter design that follows it.
- The same project: overview of its measurements. The 2σ reading as the 95th percentile of the absolute deviation, weighted RMS as the JIS reading, and the same weighting curve in AES6, DIN 45507 and IEC 60386.