The decibel meter doesn’t capture the variable that matters most.

The frequency-following response

The brain has a well-documented tendency to synchronize its electrical activity to rhythmic external stimuli. This is called the frequency-following response (FFR). If a sound source pulses at a specific frequency, the brain’s neural oscillations tend to match that pulse rate. The same principle underwrites binaural beats, which are used deliberately to induce specific brain states.

The relevant range for sleep disruption is the delta and theta bands — roughly 0.5 to 8 Hz. If a sound source pulses at rates in those ranges, it can interfere with normal sleep staging by driving the brain toward inappropriate rhythms rather than allowing it to cycle naturally through sleep architecture.

Two different frequencies are in play

A point of clarification that confuses lay readers: two different frequencies are involved in this analysis.

1. The audio frequency of the sound itself — the actual hertz of the sound waves. White noise contains all audible frequencies (roughly 20 Hz to 20 kHz). A bass tone is in the low hundreds of Hz. A whistle is in the kilohertz.
2. The frequency of the amplitude modulation — the rate at which the overall volume of the sound rises and falls. This can be very low (one cycle per several seconds = fractional Hz), or fast (dozens of cycles per second).

The brain’s entrainment response tracks the modulation frequency, not the audio frequency. A white-noise source whose amplitude cycles at 4 Hz delivers a 4 Hz entrainment signal even though the sound itself contains no 4 Hz audio component.

This is why a “drone under the fan sound, slowly rising and lowering” is a meaningful observation independent of the apparent loudness or frequency content of the noise itself. The drone is the modulation envelope.

Sources of amplitude modulation in residential environments

Amplitude modulation can arise from innocent mechanical causes or from interactions between multiple devices:

• Bent or imbalanced fan blades. A fan with even a small deformation creates a rotating imbalance that produces rhythmic amplitude variation at the rate of the blade’s rotation. The deformation may be too small to see.
• Off-center motor mounts. Same effect via a different mechanism.
• Beat frequencies between two devices. Two fans running at slightly different motor speeds interact acoustically; the difference between their frequencies produces a slow oscillation in apparent volume. Multiple devices in the same space compound this.
• Two-stroke engines under variable throttle. See the blowers page for the detailed analysis. Backpack leaf blowers operated near windows are the most acutely problematic case for residential exposure because the operator’s hand on the trigger produces an unpredictable amplitude envelope at exactly the frequencies most disruptive to sleep.

Infrasound — below the threshold of conscious hearing

Frequencies below about 20 Hz are below the threshold of conscious human hearing. This does not mean they have no physiological effect. Infrasound at specific frequencies has been associated with:

• Feelings of unease
• Anxiety
• Disorientation
• In some documented cases, visual anomalies

The frequency around 18.98 Hz is sometimes cited in this literature because it is close to the resonant frequency of the human eyeball.

The British researcher Vic Tandy documented this in a paper published in the Journal of the Society for Psychical Research in 1998. He found that a standing wave at 18 Hz in his laboratory was producing feelings of unease and apparent visual phenomena in people working there. The work has been cited in subsequent infrasound-effects literature.

Industrial and commercial sources of infrasound include HVAC systems, large-scale wind turbines, and internal combustion engines — including the two-stroke engines used in commercial backpack blowers.

The pulsed-white-noise risk

Steady-state white noise is generally benign and often beneficial — it creates auditory masking that gives the brain a stable, non-threatening sound floor to rest against. Pulsed white noise is a different thing entirely.

In a sleep-disruption context, a white-noise source whose amplitude cycles at a delta or theta frequency could deliver an entrainment signal to the sleeping brain at frequencies that disrupt normal sleep staging. The sleeper would experience the audio as steady (the eardrum is being driven at audible frequencies), but the modulation envelope would be acting on the neural oscillation system below the threshold of conscious detection.

How to test whether a sound is modulated

A cheap diagnostic procedure for residential equipment:

1. Run one device at a time. Turn off all but one fan, purifier, or other steady-noise source. Listen for the drone.
2. If the drone disappears or changes character when you turn devices off, it was a beat frequency between two of them. Often eliminating one device removes it.
3. If the drone persists with only one device running, it is internal to that device — likely a bent or imbalanced fan blade, an off-center motor, or similar mechanical asymmetry.
4. If the drone persists with all your devices off, it is coming from outside the building — exterior HVAC, a distant industrial source, or something else.

The diagnostic step is cheap and quickly narrows the source.

What the brain does with this

The neurological cost of modulated sound is higher than steady-state sound at the same average decibel level for a specific reason: the brain cannot habituate to unpredictable stimuli the way it can to predictable ones. A steady sound floor lets the threat-detection system disengage. A varying envelope keeps the system armed because the next event is unpredictable. Cortisol stays elevated. Sleep stays light. The damage continues.

This is why the steady drone of a refrigerator compressor — even at a measurably higher decibel level than a more variable sound — is less neurologically expensive than the irregular two-stroke engine cycle of a leaf blower running under variable throttle. The decibel meter does not capture the variable that matters most.