Synth Science

Synthesis

What synthesis actually does

Every synthesis method is a strategy for generating a spectrum that changes over time. Knowing which strategy you are using tells you which knob to reach for.

Updated 2 min read 11 citations Evidence strength 3/5

The shared problem

A musical sound is a spectrum that evolves over time. Every synthesis technique is a different strategy for producing one efficiently and controllably. The families differ in where complexity is introduced.

Three strategies
MethodStrategyIntuitive to controlAwkward to control
SubtractiveStart harmonically rich, filter awayBrightness, movement, envelopePrecise harmonic relationships
FM / phase modulationGenerate sidebands by modulating phaseBell, metallic and evolving timbresPredicting the result from parameters
Physical modellingSimulate a vibrating systemPhysically meaningful gesturesSounds no object could make
WavetableInterpolate between stored spectraMorphing and motionPrecise spectral targets

Physical modelling in particular remains an active research area — work on simplified and controllable models of mode coupling [1] addresses exactly the tension between physical accuracy and playable control.

Why subtractive dominates

Not because it is most powerful — it is not — but because its parameters map onto perceptual dimensions people already have words for. Filter cutoff maps to brightness. Resonance maps to a nasal, focused quality. Envelope maps to attack and decay. You can predict what a change will do before you make it.

FM is more capable and far less predictable. Changing a modulator ratio can move a sound from bell to bass to noise with no perceptual gradient in between, which is why FM programming has a reputation for being learned by memorisation rather than by reasoning.

Synthesis and control mapping

Control is the real design problem

Research into digital instruments — including optical and gestural controllers [5] and sound generation in mechatronic systems [4] — is largely about mapping: how a performer's gesture connects to a synthesis parameter. That mapping determines expressiveness far more than the underlying method does.

The production equivalent is modulation. A static patch from a capable synthesiser is less interesting than a simple patch with well-chosen modulation, because what makes a sound feel alive is change over time, not spectral complexity at any instant.

Common questions

Which synth should I learn first?
A subtractive one, because its parameters map onto qualities you can already describe. Everything else is easier afterwards.
Is analogue better than digital?
Different, and the differences are mostly in nonlinearity, drift and interface. Neither is categorically superior for producing a given sound.
Why do my patches sound lifeless?
Usually insufficient modulation. Change over time matters more than spectral complexity.
Do I need hardware?
No. Hardware changes how you interact with parameters, which changes what you make — that is a workflow argument, not a sound-quality one.

References

Every citation below links to the original peer-reviewed record on PubMed or via DOI. Nothing here is a substitute for medical advice.

  1. A simplified and controllable model of mode coupling for addressing nonlinear phenomena in sound synthesis processes Poirot S, Bilbao S, Kronland-Martinet R · EURASIP Journal on Audio, Speech, and Music Processing · 2024 · Journal article DOI
  2. Signal Processing for Image Morphing in Digital Image Synthesis Dr. Elena Martinez · American Journal of Signal and Image Processing · 2024 · Journal article DOI
  3. Sound event triage: detecting sound events considering priority of classes Tonami N, Imoto K · EURASIP Journal on Audio, Speech, and Music Processing · 2023 · Journal article DOI
  4. Exploring Sound Generation and Processing in Mechatronic Systems through Digital Signal Processing: A MATLAB-Based Investigation Wahhab M · Engineering and Technology Journal · 2023 · Journal article DOI
  5. Optical Digital Theremin with Audio Synthesis and Graphic Interface Ramos dos Santos M, Coca Salazar A · International Journal of Circuits, Systems and Signal Processing · 2021 · Journal article DOI
  6. Digital audio signal watermarking using minimum‐energy scaling optimisation in the wavelet domain Hsu C, Tu S, Yang C, et al. · IET Signal Processing · 2020 · Journal article DOI
  7. Processing of volcano infrasound using film sound audio post-production techniques to improve signal detection via array processing Williams R, Perttu A, Taisne B · Geoscience Letters · 2020 · Journal article DOI
  8. Sound effects without side effects—Suppression of artifacts in audio signal processing Case A · The Journal of the Acoustical Society of America · 2018 · Journal article DOI
  9. Analysis by synthesis spatial audio coding Elfitri I, Shi X, Kondoz A · IET Signal Processing · 2014 · Journal article DOI
  10. Digital removal of pulse‐width‐modulation‐induced distortion in class‐D audio amplifiers Aase S · IET Signal Processing · 2014 · Journal article DOI
  11. Applications and implications of digital audio databases for the field of ethnomusicology: A discussion of the CNRS — Musée de l’Homme sound archives Khoury S, Simonnot J · First Monday · 2014 · Journal article DOI