Harmonic Comb — User Guide

Recursive inharmonic comb morph with a sample-by-sample beta trajectory, weighted resonant teeth, random phase modulation, and true dry/effect mixing.

Author: Shai Cohen Affiliation: Department of Music, Bar-Ilan University, Israel Version: 0.5.1 (2026) License: MIT License Repo: https://github.com/ShaiCohen-ops/Praat-plugin_AudioTools

Contents:

What this does Quick start Presets Parameters Comb engine Output & visualization

What this does

Harmonic Comb is a time-varying stack of recursive comb resonators. A randomly chosen fundamental delay anchors a set of teeth whose delays follow fundamental / h^β. The exponent β moves linearly from Beta_start to Beta_end over the sound, so the resonant spacing can move from harmonic to compressed or stretched/inharmonic structures.

The harmonic index h = 1 is fixed at the fundamental delay because 1^β = 1. Higher teeth are weighted by 1/h² and receive independent random phases in the cosine modulation.

Quick start

Select exactly one Sound.
Run Harmonic_Comb.praat.
Choose a preset, or Custom.
Use Beta_start and Beta_end to define the spectral-spacing trajectory.
Set Wet_dry_percent; 0 = dry, 100 = comb-resonance contribution only.
Run the script. The result is named original_comb_Preset.

Presets

Preset	Harmonics	Fundamental delay range	β start → end	Modulation period
Subtle Comb	4	30–80 samples	1.00 → 1.05	1200 samples
Medium Comb	7	20–100 samples	1.00 → 1.30	1000 samples
Heavy Comb	11	15–120 samples	0.85 → 1.50	850 samples
Extreme Comb	16	10–150 samples	0.70 → 2.00	700 samples

Parameters

Parameter	Default	Behavior
Number_of_harmonics	7	Number of recursive comb teeth.
Fundamental_delay_min_samp	20	Lower bound for the randomly selected fundamental delay.
Fundamental_delay_max_samp	100	Upper bound. The two bounds are rounded, clamped to at least 2 samples, and reordered if necessary.
Beta_start	1.0	Initial spacing exponent. β=1 is harmonic; β<1 compresses; β>1 stretches.
Beta_end	1.3	Final exponent; β is linearly interpolated sample-by-sample.
Modulation_period	1000 samples	Period parameter in the cosine term. Harmonic `h` uses `h/period` cycles per sample plus a random phase.
Wet_dry_percent	50	Clamped to 0–100.
Scale_peak	0.95	Final attenuate-only peak ceiling; quiet signals are not boosted.

Comb engine

β(col) = Beta_start + (Beta_end - Beta_start) × (col-1)/(ncol-1) delay_h(col) = round(fundamental_delay / h^β(col)) weight_h = 1/h² if delay_h ≥ 2 and col > delay_h: self += weight_h × self[col-delay_h] × cos(randomPhase_h + 2π × col × h / Modulation_period)

The delayed read is from self, so each tooth is a feedback/recursive comb, not a feed-forward tap. Teeth whose current delay falls below 2 samples are treated as inactive to avoid zero/one-sample self-gain behavior.

The chosen fundamental delay is random on every run unless the minimum and maximum bounds are equal. Every tooth also receives a random phase. There is no exposed random seed, so repeated runs with identical form values need not be identical.

Wet/Dry semantics

The internal processed buffer contains the dry signal plus comb resonance. The final mixer explicitly subtracts the embedded dry component, so the public control behaves as advertised:

output = dry × (1 - wet) + (processed - dry) × wet

Output and visualization

Name: original_comb_Preset.
Duration, sample rate, channel count and time domain: inherited from the selected Sound; the script processes the Sound directly without adding a tail.
Peak handling: only attenuates when the result exceeds Scale_peak.

The visualization contains the original waveform, result waveform, a 1/h² weight chart, and paired start/end delay bars for each comb tooth. Those delay bars show the endpoints of the β morph; the actual delay changes continuously across the sound.