TECHNICAL DEEP DIVE  ·  SIGNAL PROCESSING  ·  AUDIO ENGINEERING

How Gamma40 Tracks
Are Built
From First Principles

Every track is assembled from mathematics, not samples. This is the physics and engineering behind the 40 Hz pulse — from a single sine wave to a validated 60-minute audio file.

SINE WAVESNYQUIST THEOREMPULSE ENGINEERING DUTY CYCLEFFT VALIDATIONLUFS NORMALISATION PSYCHOACOUSTIC COMPRESSIONBROWN NOISE
THE FULL PICTURE

Two layers. One pipeline.
144,000 verified pulses.

Every track is built from two independent audio streams mixed, shaped, validated, and exported. Layer 1 is synthesised entirely from equations — nothing borrowed from any recording.

LAYER 1
40 Hz Pulse Train
+
🌧️
LAYER 2
Ambient Soundscape
⚖️
MIX
RMS Balance −2 dB
📈
SHAPE
Fade-in · LUFS
VALIDATE
FFT · Jitter Check
💾
EXPORT
MP3 · M4A · WAV
§ 01 · THE BUILDING BLOCK

The sine wave —
the universe's favourite shape

Everything starts here. A sine wave is the simplest possible oscillation. It describes a pendulum, a guitar string, a tuning fork. And the 40 Hz pulse at the heart of every Gamma40 track.

y(t) = A × sin(2π × f × t)
A = amplitude (loudness height)
f = frequency in Hz (cycles/second)
t = time in seconds

At 40 Hz, the wave completes 40 full cycles per second, each lasting 25 ms. At 1,000 Hz (the carrier inside each pulse) a full cycle takes just 1 ms. The 1 kHz wave oscillates 25× faster than the pulse rate.

Why 1,000 Hz? The mid-range where the human ear is most sensitive. One 1 kHz cycle also fits exactly inside one 1 ms pulse burst, which is not a coincidence.
WAVEFORM COMPARISON · LIVE ANIMATION
§ 02 · DIGITAL AUDIO

Sampling — turning continuous
waves into numbers

A computer cannot store a continuous wave. Instead it takes 44,100 measurements per second, each called a sample, and records the amplitude as a number. This is the CD standard.

Sample rate: 44,100 Hz
Duration: 60 min = 3,600 s
Samples (stereo): 317,520,000

Why 44,100? The Nyquist theorem states you must sample at least twice the highest frequency you want to capture. Humans hear up to ~20,000 Hz, so 44,100 provides a safe margin above the 40,000 Hz minimum.

Precision: The entire pipeline runs in 64-bit floating point (values −1.0 to +1.0) for maximum arithmetic accuracy. Only at the final export step are values rounded to 16-bit integers: 65,536 distinct loudness levels.
SAMPLING VISUALISATION · 44,100 Hz
§ 03 · THE PULSE TRAIN

Not a tone —
a knock

The 40 Hz stimulus is not a continuous 40 Hz hum. Instead it is a pulse train: a rapid series of brief sound bursts, each just 1 ms long, repeating 40 times per second.

Each 40 Hz cycle = 25 ms total:
|← 1 ms ON →|← 24 ms silence →|

Duty cycle: 4% (ON only 4% of the time)

This matches the Wang et al. (PNAS 2026) stimulus exactly: "1 ms duration at a frequency of 40 Hz."

Why pulses beat continuous tones: The auditory brainstem responds most powerfully to transients (rapid onset and offset events). Each pulse generates a precise Auditory Brainstem Response (ABR), driving neural synchrony far more effectively than a steady hum at the same frequency.
PULSE TRAIN · 40 Hz · 4% DUTY CYCLE
§ 04 · PRECISION TIMING

The fractional accumulator —
exact 40.0000 Hz

At 44,100 samples/second, one 40 Hz cycle spans exactly 1,102.5 samples. Not a whole number. Rounding to 1,102 or 1,103 accumulates timing error. Over 144,000 pulses this becomes measurable.

44,100 ÷ 40 = 1,102.5 samples/cycle

Solution: carry the 0.5 forward every cycle.
The fractional remainder never disappears.

A running counter tracks position with sub-sample precision. When it reaches 1,102.5 it resets, but the 0.5 carries forward. Over an entire 60-minute session, zero phase error accumulates.

Measured result: frequency error <0.001 Hz (50× better than the ±0.05 Hz target). Timing jitter: 0.008 ms RMS (62× better than the 0.5 ms target).
FRACTIONAL ACCUMULATOR vs NAIVE INTEGER
§ 05 · EDGE SMOOTHING

Cosine micro-fades —
eliminating the click

Switching a sine wave on and off abruptly creates a click. Mathematically this is a rectangular envelope applied to the carrier. A rectangle has infinite-bandwidth harmonics in the frequency domain.

A cosine fade is applied to the first and last 4 samples (≈0.09 ms) of each burst, smoothing amplitude from 0→1 at onset and 1→0 at offset.

fade(i) = 0.5 − 0.5 × cos(π × i / n), n = 4

Result over 4 samples (i = 0 … 3):
0 → 0.15 → 0.50 → 0.85 → [1.0 steady]
Result: the pulse sounds like a soft, rounded transient, closer to a gentle tick than a sharp electrical pop. The therapeutic signal is fully preserved; the discomfort is eliminated.
PULSE EDGE SMOOTHING · ZOOM ×40
§ 06 · AMBIENT SYNTHESIS

Brown noise — from a
random walk

The Brown Noise track contains no recording. It is synthesised from mathematics. Starting from white noise (pure randomness at every sample) and integrating it into something warm and natural.

White noise: random at every sample
white[n] ~ Gaussian(0, 1)

Brown noise: running sum (integration)
brown[n] = brown[n−1] + white[n]

Integration in the time domain produces a −6 dB/octave rolloff in the frequency spectrum: low frequencies dominate and highs fade. The result sounds like distant thunder or a low waterfall: deep, warm, enveloping.

After synthesis: a high-pass filter removes sub-20 Hz energy (below human hearing but capable of speaker damage), then the signal is normalised to the target LUFS level.
RANDOM WALK → BROWN NOISE + SPECTRUM
§ 07 · MIXING

Combining two layers —
RMS balance

RMS (Root Mean Square) is the standard measure of average signal power. It represents what the ear hears as sustained loudness, independent of momentary peaks.

RMS = √( mean(samples²) )

pulse_RMS = ambient_RMS × 10^(−2/20)
≈ ambient_RMS × 0.794

The pulse is scaled so its RMS sits 2 dB below the ambient. It carries about 79% of the ambient's average power. Clearly audible, always present, but the ambient stays dominant for comfortable 60-minute listening.

Stereo rule: the pulse is always identical in both channels (centre-panned). Any timing difference between ears would create binaural beating, which is a completely different and unvalidated mechanism.
LAYER MIX · RMS BALANCE
§ 08 · VALIDATION

FFT validation — proving the
physics is correct

Before any track is exported, the pulse train is analysed by a Fast Fourier Transform. The FFT decomposes the signal into its frequency components — a spectrum showing exactly how much energy is present at each frequency. If either check fails, the track is not exported.

What the spectrum reveals

The pulse train — a 1 kHz sine burst repeating at 40 Hz — produces a characteristic fingerprint:

40 Hz — the fundamental repetition rate of the pulse

80, 120, 160 Hz — harmonics from the 4% duty cycle shape

1,000 Hz — the carrier sine wave inside each burst

960, 1,040 Hz — sidebands at 1 kHz ± 40 Hz

Measured accuracy: 40.0000 Hz ±<0.001 Hz
Timing jitter: 0.008 ms RMS (target: <0.5 ms)
BY THE NUMBERS

The quantities that define a session

40.0000
Hz exact pulse frequency
ERROR <0.001 Hz
144,000
Pulses per 60-min session
40/SEC × 3,600 SEC
1 ms
Duration of each pulse burst
4% DUTY CYCLE
44,100
Samples per second (CD quality)
NYQUIST × 2.2
317M
Total samples in stereo track
317,520,000
0.008
ms timing jitter RMS
62× BETTER THAN TARGET
−18
LUFS integrated loudness
ITU-R BS.1770-4
~3 min
Pipeline time, all 9 tracks
60-MIN SESSIONS EACH
§ 09 · EXPORT

Does compression
disturb the pulse?

A common concern: does MP3 or AAC encoding alter the 40 Hz timing? No — and here is why. Lossy codecs operate by removing spectral content the ear cannot perceive given simultaneous louder sounds. They do not alter timing. The 40 Hz pulse envelope (the on/off pattern) survives encoding intact.

This has been confirmed empirically: FFT analysis of exported MP3 and M4A files shows the same 40.0000 Hz peak as the source WAV.

MP3
320 KBPS · LAME ENCODER
The float64 master encoded at 320 kbps. Perceptual loss is inaudible at this bitrate. File size: ~144 MB per 60-minute track. Compatible with every device and media player.
M4A
256 KBPS · AAC · MODERN CODEC
AAC achieves equivalent perceptual quality to MP3 at lower bitrate. At 256 kbps it is indistinguishable from the source. Preferred for the Apple ecosystem. File size: ~115 MB per session.
WAV
LOSSLESS · 16-BIT / 44.1 kHz PCM
Float64 master scaled to 16-bit integers (×32,767) and written as a lossless WAV file. Bit-perfect reproduction, zero information discarded. File size: ~300 MB per session.
IMPORTANT — STREAMING LOUDNESS NORMALISATION

Streaming platforms apply loudness normalisation that alters the carefully calibrated pulse-to-ambient ratio. Direct download and local playback is always recommended for Gamma40 tracks. The −18 LUFS master is calibrated precisely — streaming adjusts this without your control.