Fluctuation strength quantifies the perception of slow amplitude modulations in a sound.
While roughness captures rapid modulations (15 – 300 Hz), fluctuation strength addresses
modulation frequencies below approximately 20 Hz, with peak sensitivity at about 4 Hz. At
these low modulation rates, the listener perceives a rhythmic "fluctuating" or "throbbing"
character rather than a continuous rough texture.
Theory
Low-Rate Modulation Detection
The auditory system resolves slow envelope fluctuations as distinct temporal events. When the
modulation rate is sufficiently low (below ~20 Hz), each modulation cycle is perceived
individually, producing a sensation of temporal variation. As the rate increases beyond 20 Hz,
the individual cycles merge into a continuous roughness percept.
Modulation Transfer Function
The fluctuation strength model applies a modulation transfer function that is band-pass in
nature, peaking at 4 Hz. This function reflects the temporal resolution limit of the auditory
system:
Here ΔL(z) represents the level variation within each critical band due to modulation,
and fmod is the modulation frequency in Hz. The denominator reaches its minimum at
fmod = 4 Hz, producing the maximum fluctuation strength.
Relationship to Roughness
Fluctuation strength and roughness are complementary metrics that together cover the full
range of temporal modulation perception. The crossover region lies near 15 – 20 Hz, where
both models produce low values. In practice, most real-world sounds exhibit modulation
components in both regimes, and both metrics should be evaluated together.
Parameters
Parameter
Options
Description
Block Size
200 ms – 2000 ms
Longer blocks are needed compared to roughness to resolve modulations at 2 – 4 Hz
Overlap
0% – 75%
Overlap between consecutive blocks; higher overlap provides smoother time traces
Field Type
Free field / Diffuse field
Ear transfer function for critical-band analysis
Modulation Frequency Range
0.5 Hz – 20 Hz (fixed)
Effective modulation frequency range captured by the model
Output Quantities
Quantity
Unit
Description
F
vacil
Overall fluctuation strength
F(t)
vacil
Fluctuation strength vs Time trace
Fluctuation Strength vs Time
Time-varying fluctuation strength F(t) reveals how low-frequency modulation evolves over the
course of a recording. Typical applications include:
Detecting idling-speed irregularities in combustion engines where cylinder misfires
produce ~4 Hz modulation patterns
Evaluating HVAC duct pulsation that may produce a "breathing" or "surging" quality at
low blower speeds
Assessing the rhythmic beat frequency when two closely spaced tones produce audible
beating at 2 – 8 Hz
Interpretation Guidelines
Reference value: 1 vacil is defined as the fluctuation strength of a 1 kHz
tone at 60 dB SPL, 100% amplitude-modulated at 4 Hz.
> 0.7 vacil — Strong fluctuation; dominant temporal pattern, likely to cause
annoyance if sustained.
Practical Application Notes
In automotive applications, fluctuation strength is relevant for idle quality assessment.
Irregular combustion at idle produces modulation near 4 Hz, which maximizes the fluctuation
strength and is perceived as engine "lumpiness."
Fluctuation strength is less commonly specified as a stand-alone target metric than
loudness or sharpness. However, it is frequently used as an input to composite sound quality
indices (e.g., Aures pleasantness model).
Due to the long analysis blocks required (200+ ms), temporal resolution of F(t) is
inherently lower than for roughness. Rapid transients may be smoothed out.
Beat phenomena between two tonal components separated by a few Hz will produce high
fluctuation strength. This is a common source of complaint in multi-compressor
installations or multi-fan systems.
Standard References
E. Zwicker & H. Fastl, Psychoacoustics: Facts and Models, 3rd ed., Springer, 2007,
Chapter 10: Fluctuation Strength
H. Fastl, "The Psychoacoustics of Sound-Quality Evaluation," Acustica / acta acustica,
vol. 83, 1997
W. Aures, "Ein Berechnungsverfahren der Rauhigkeit," Acustica, vol. 58, 1985
(companion roughness model for comparison)