Generating 864 MIDI Drones

Generating 864 MIDI Drones

How the SuperScales companion drone system was built — one drone per mode per key, the acoustic engineering behind drone design, and why 864 is the right number.

The SuperScales book includes 864 MIDI drone files — one for each combination of MMA mode, transposition, and practice context. These are not incidental accessories. The drone is the pedagogical anchor of the entire SuperScales ear training system: without a constant, accurate pitch reference, key center hearing cannot be developed. The 864 drones are the infrastructure that makes the SuperScales system practicable rather than merely theoretical.

Where 864 Comes From
12 keys × 7 modes × ~10 drone variants = 864 total drone files
Each key and mode combination has multiple drone variants: different octaves, different durations, different metronome subdivisions. The 864 covers every practice context in the book.

Why One Drone Per Mode, Not Just Per Key

A key center drone for C stays on C regardless of which MMA mode is being practiced. So why 7 different drone files per key rather than one? The answer is that the drone file for each mode includes not just the root tone but mode-specific context: the characteristic intervals of the mode sounding as sustained tones underneath the practice melody, giving the ear an immediate reference for what the mode’s scale degrees sound like against the tonal center.

For the Lydian Dominant mode, the drone includes a soft #11 (F#) sounding alongside the root — so the student’s ear immediately hears the Lydian Dominant’s characteristic interval as part of the background. For the Altered scale, the drone includes the b9 and b13 as soft background tones alongside the root, giving the student an immediate sense of the Altered scale’s sound world before they play a single note of the exercise.

This mode-specific drone design is more sophisticated than a simple root tone, and it requires more production work — 7 distinct drone files per key rather than one. But it dramatically accelerates the ear training process: rather than asking the student to construct the mode’s sound world mentally from scratch with each practice session, the drone gives them the sound world immediately.

Drone Engineering Specifications

Root Pitch Accuracy
±2 cents in equal temperament (A4 = 440Hz)
Critical — even small inaccuracies create beats with the exercise material that distort key center perception
Drone Timbre
Soft organ / filtered sine wave
Minimal overtone competition with the exercise material. Mode characteristic tones at -12dB relative to root
Duration
4 minutes per drone file
Covers a complete 5-10 minute practice session with loop capability. Loopable at phrase boundaries.
Metronome variants
60, 80, 100, 120 BPM per drone
Four tempo variants per mode per key. The click is spectrally distinct from the drone to avoid masking
Octave variants
Low, mid, high register
Three register variants for practicing different range areas. Low register for bass, high for upper voice clarity
File format
MIDI + MP3 (320kbps)
MIDI for tempo flexibility; MP3 for immediate playback without a MIDI player or software

The Generation Pipeline

Each of the 864 drone files was generated using a SuperCollider script that synthesized the drone sound from scratch — no recorded audio, no samples. SuperCollider’s additive synthesis allowed precise control of the timbre: the root tone was a combination of the fundamental and a carefully attenuated second harmonic, creating a full-bodied sound without the overtone density that would compete with the exercise material.

The mode-specific characteristic tones (the #11 for Lydian Dominant, the b9 and b13 for Altered, etc.) were synthesized at -12dB relative to the root — present enough to color the sound but not loud enough to interfere with recognition of the exercise material. The metronome click was synthesized as a short, spectrally bright burst (centered around 2kHz) that cut through the drone without masking it.

Python orchestrated the batch generation: one script looped through all 12 keys, all 7 modes, all tempo and register variants, called SuperCollider for each combination, captured the rendered audio, normalized it, converted it to MP3, and organized the output into folders by key and mode. Total generation time for all 864 files: approximately 6 hours of computer time, versus the weeks it would have taken to record each one manually.

Why Drone Quality Matters — Four Acoustic Reasons
1Pitch accuracy: An out-of-tune drone creates beats with the exercise notes that distort the student’s perception of scale degree intervals. Even 5 cents of drone inaccuracy introduces a rhythmic interference pattern that the brain processes as pitch ambiguity.
2Timbre neutrality: A drone with a rich overtone profile generates pitch content that competes with the exercise material. The student’s ear must separate the drone’s overtones from the exercise notes — cognitive load that should be spent on scale degree recognition.
3Mode-specific context: A drone that only provides the root pitch gives the student less context than one that includes the mode’s characteristic intervals softly in the background. More context = faster acquisition of mode sound world.
4Availability: A drone that requires special software or hardware to play is a drone that doesn’t get used. MP3 format means the drone plays on any device, always available, zero friction. Friction in practice infrastructure kills practice habits.

Engineering and ear training — the same design principles

Parametric generation

One SuperCollider synthesis script, 864 parameter combinations, 864 output files. The same parametric generation principle used for the 9,240 trichord permutations — one template, many instances. Scale of output through systematic parameterization.

Mode-specific context

Including characteristic mode intervals in the drone (at -12dB) is an engineering decision with pedagogical consequences — it accelerates the formation of the mode’s sound world in the student’s ear by providing context rather than just reference.

Timbre as engineering choice

The filtered sine wave drone timbre is an engineering choice optimized for one purpose: minimal overtone competition with exercise material. Every engineering decision in the drone system is in service of the pedagogical goal — faster key center perception development.

Friction elimination

MP3 format, immediate playback, no special software required — every friction point eliminated from the practice workflow. Engineering in service of behavior: reducing friction increases the likelihood that the student actually practices.

The next article examines the 2,521 trichord pair études generated for the SuperScales book — how MMA’s subset structure produced this specific number, and how the études connect the SuperScales system to the Sound Cells hexatonic trichord pair series.

SuperCollider additive synthesis documentation: supercollider.github.io/SuperCollider/Help/Classes/SinOsc.html. MIDI specification: midi.org. The 864 SuperScales drone files are included with the SuperScales book at muse-eek.com.

Leave a Reply

Your email address will not be published. Required fields are marked *