When engineers describe the "aliveness" of analog tape recordings, they often credit wow and flutter. The explanation is usually left there: tape speed varies slightly, pitch wobbles, that wobble sounds organic. What gets skipped is that "wow and flutter" names three physically distinct mechanisms operating at completely different frequencies, each producing a completely different perceptual effect. Two of them involve pitch variation. One does not.
Three Frequency Regimes
IEC 60386, which superseded DIN 45507 in 1972, defines the zones of tape speed instability:
Drift covers deviations below 0.5 Hz. This is speed inaccuracy, not wow or flutter. A machine running at 7.4 ips instead of 7.5 ips is drifting. Drift matters for archivists comparing recordings made on different machines, not for the sonic texture of the recording itself.
Wow spans 0.5 Hz to approximately 6 Hz. At these rates, the pitch variation is perceptible as pitch variation. You hear the frequency go up and down. At 1 to 3 Hz, it is indistinguishable from deliberately applied vibrato. The primary mechanical sources are irregularities in the capstan motor, worn bearings, and fluctuations in tape tension as the supply and take-up reels change diameter through a recording.
Flutter spans 6 Hz to approximately 100 Hz. Here the perceptual effect changes completely. At these rates, the auditory system no longer integrates the modulation as pitch variation. Instead, it resolves it as texture: roughness, smearing, a slight "dirt" in the sound. The sources differ too. Pinch roller unevenness, idler wheel irregularities, and periodic vibrations transmitted through the machine's frame all contribute to the flutter band.
Scrape Flutter: The Third Mechanism
Above approximately 100 Hz sits scrape flutter, which is physically and perceptually unlike anything below it.
As tape moves across stationary heads and guides, friction creates mechanical oscillation through a stick-slip process. The tape slows slightly due to friction; tension builds in the tape until it overcomes the static friction; the tape snaps forward, overshooting; the cycle repeats. This happens at rates of 100 Hz to several hundred Hz.
At these frequencies, scrape flutter does not sound like pitch variation. The modulation rate is too fast for the auditory system to resolve as pitch change. Instead, it produces intermodulation: the flutter-induced sidebands fall at harmonically unrelated offsets from the original signal components, creating a subtle but characteristic anharmonic roughness. This is part of the grit in heavily driven analog recordings that is not attributable to magnetic saturation.
Traditional tape machine designs placed mechanical dampers and rollers at specific points in the tape path to interrupt the stick-slip cycle. The tape heads themselves were designed to minimize contact surface area. On poorly maintained machines where guide friction was elevated, scrape flutter appears as one of the most prominent degradation artifacts.
Measurement and Weighting
The standard measurement method uses a 3150 Hz test tone (CCIR uses 3000 Hz) and applies a weighting filter before computing the RMS result. This weighting filter is common across IEC 60386, the older DIN 45507, CCIR 409-3, and AES6-2008.
The filter matters in practice. It emphasizes the frequency range where modulation is most perceptually significant and rolls off at the extremes. A machine with significant energy concentrated in high-frequency flutter can appear in its specification with a lower number than a machine with more evenly distributed flutter, even though the former may sound less stable in use. Two machines with identical weighted measurements can sound different if their spectral distributions differ.
Results are reported as a percentage of pitch deviation from nominal. Above 0.1%, wow is audible to trained listeners. Above 0.2%, it is obvious. Professional machines target numbers significantly lower:
- Studer A820: 0.04% (DIN weighted)
- Studer A80 2-track mastering: 0.05% (DIN weighted)
- Consumer hi-fi cassette (IEC maximum): 0.2%
- Non-hi-fi cassette (IEC maximum): 0.4%
The Ampex ATR-102, introduced in 1976, achieved flutter specifications described by its engineers as essentially non-existent, through servo-controlled reel motors and an unusually large capstan without a pinch roller. Eliminating the pinch roller removed one of the primary sources of periodic flutter.
Why the Spectrum Matters More Than the Number
A sine LFO at 2 Hz sounds like a single-rate pitch wobble. Real tape wow has a more complex spectral signature.
The capstan rotation generates a fundamental frequency determined by the machine's transport speed and capstan circumference. Mechanical imperfections in the motor introduce odd harmonics of this fundamental, since real motor rotation has nonlinear error components. The winding and unwinding of tape on the reels adds low-frequency modulation that shifts gradually as the reels change diameter from start to end of tape. Random mechanical vibration contributes a broadband noise floor across the flutter range.
The result is a spectrum with: a capstan fundamental and its harmonics in the wow-to-low-flutter region, reel-related drift components at sub-Hz frequencies that change slowly with play position, and a broadband stochastic component across the entire flutter band.
A two-parameter "wow rate / flutter rate" model does not reproduce this. Setting a single wow rate and a single flutter rate yields a sound that identifies itself as artificial. The spectral shape is wrong: real wow has a cluster of harmonically-related peaks embedded in noise, not a single peak at one frequency.
DSP Emulation: What It Requires
Tape emulation plugins modeling wow and flutter face a specific technical constraint: the correct implementation is speed modulation applied to a variable-delay line, not pitch shifting applied at fixed timing. A phase vocoder pitch shifter with LFO is not equivalent to a variable-speed playback system, even though both change pitch. Speed variation scales all frequencies uniformly and maintains the temporal relationship between signal and modulation. Phase vocoder processing imposes its own artifacts unrelated to tape mechanics.
More authentic approaches use several modulation sources summed together: a fundamental component in the wow range, harmonics at higher frequencies, a stochastic component in the flutter band, and separate treatment of scrape flutter as an intermodulation effect rather than pitch deviation. Perlin noise, or similar coherent noise algorithms, generates the kind of smoothly varying but unpredictable modulation that real mechanical systems produce. Unlike white noise, which changes direction at the sample level, Perlin noise has a defined correlation time that can be matched to the characteristic wander of a real capstan motor.
The inverse problem, correcting tape wow and flutter from existing recordings, illustrates the same geometric constraints. Celemony Capstan analyzes recordings by detecting pitch variation in the signal, computes a time-varying correction curve, and resamples. This works because wow is primarily low-frequency and bandwidth-limited by mechanics: the same characteristic that makes real tape modulation sound distinctive makes it recoverable.
The Practical Upshot
When you reach for a tape emulation plugin's wow and flutter controls, three separate mechanisms sit behind the knob:
Wow at 1 to 4 Hz adds pitch movement that reads as organic expressiveness. This is the vibrato-like quality that makes a held piano note sound less static.
Flutter at 10 to 60 Hz adds textural roughness. Too much sounds like a cheap cassette player running low on batteries. The right amount sounds like material that was recorded on tape.
Scrape flutter, when modeled, adds anharmonic intermodulation that creates subtle grit distinct from magnetic saturation distortion.
Most plugins blend the first two into a single control. The ones that separate them let you tune the character of the instability rather than just the amount. The fact that scrape flutter requires a different processing model, intermodulation rather than speed variation, means that many emulations leave it out entirely. This is part of why tape saturation combined with wow and flutter still does not always reproduce everything that physical tape does. One of the three mechanisms is usually missing entirely.