How to Use Additive Synthesis for Horror Explosions

By Marcus Chen · April 16, 2026

How to Use Additive Synthesis for Horror Explosions

1) Introduction: why additive synthesis belongs in a “big, dirty” sound

Horror explosions are rarely about realism. They’re about physiology: the body’s startle reflex, the sense of pressure in the chest, the perception of unstable energy, and the uncanny feeling that something is “wrong” even when it’s loud and familiar. Traditional explosion design leans on recorded impacts, debris layers, and distortion. Additive synthesis seems like the opposite—clinical, controlled, harmonic. Yet that control is precisely why it excels at horror.

Additive synthesis lets you prescribe the explosion’s spectral trajectory in a way that sampled material and subtractive synths can’t easily match. You can dictate which partials bloom, which collapse, and how inharmonicity evolves across the first 300 ms—where the brain decides whether a sound is “a normal boom” or “a threatening event.” This article treats the question technically: how to build horror explosions from partials with measurable intent, then integrate them into professional sound design workflows.

2) Background: physics, perception, and engineering principles

Explosions in the real world: a simplified acoustic model

A physical explosion is a rapid energy release generating a shock front, followed by turbulent broadband noise, low-frequency pressure fluctuations, reflections, and debris transients. In the far field, the shock front becomes less discontinuous; what remains is a broadband pulse with strong low-frequency content and a time-varying spectral centroid.

Key features relevant to synthesis:

Fast onset: rise times on the order of 0.1–5 ms for sharp transients (distance and recording chain dependent).
Spectral nonstationarity: the spectrum changes rapidly over 50–500 ms.
Low-frequency dominance: significant energy below 120 Hz, often below 60 Hz in cinematic design.
Decorrelation across bands: different frequency regions evolve differently (a common failure of “single-envelope” approaches).

Why additive: deterministic control of partial trajectories

Additive synthesis reconstructs a waveform as the sum of sinusoids:

x(t) = Σ_k=1..N A_k(t) · sin(2π f_k(t) t + φ_k(t))

The power is in A_k(t) and f_k(t) being time-varying per partial. Horror explosions benefit from:

Controlled inharmonicity (detuning, stretched partial series): evokes structural failure and “impossible scale.”
Band-specific envelopes: low-end sustain while highs collapse quickly—mirrors cinematic expectations.
Phase strategy: manipulate crest factor and perceived punch without relying solely on clipping.

Perceptual anchors: what makes it “horror” rather than “action”

Psychoacoustically, horror leans on instability and ambiguity:

Unstable pitch cues: slow drift, micro-jitter, or partial “locking/unlocking.”
Roughness: modulation components in the ~15–80 Hz region create sensory dissonance (roughness perception peaks depending on frequency separation and center frequency).
Subharmonic implication: energy that suggests a fundamental below playback capability, created via harmonic spacing and saturation products.

Standards-wise, your monitoring and delivery context matters. If you’re designing for cinema, theatrical calibration (e.g., SMPTE reference alignment practices and common mix targets) implies different headroom and sub management than nearfield streaming deliverables. Even when not quoting a single number, the principle holds: design the spectrum for the reproduction chain, not an idealized waveform.

3) Detailed technical analysis: building blocks, targets, and measurable choices

A. Core architecture: three additive layers

A reliable horror explosion can be constructed from three additive components, each with its own partial strategy:

Pressure core (LF): 20–120 Hz “body,” long-ish decay, slight instability.
Blast shell (MF): 120 Hz–2 kHz energetic bloom and collapse, transient definition.
Shard field (HF): 2–12 kHz spiky, short-lived partial clusters for grit and fear cues.

B. Partial sets and frequency design

1) Pressure core

Use a quasi-harmonic series around a fundamental between 28–55 Hz for “cinematic mass.” If the playback system can’t reproduce 30 Hz, still consider building the harmonic scaffold; the ear can infer low fundamentals through upper harmonics.

Fundamental (f0): 35 Hz (example)
Partials: 1–12 harmonics (35–420 Hz), but weight heavily below 140 Hz
Amplitude law: A_k ∝ 1/k^p with p between 1.2 and 2.0; then override selectively (e.g., boost k=2–4 to push chest impact)
Detune strategy: ±3 to ±12 cents on select partials, time-varying (slow random walk over 200–800 ms)

Avoid making this layer a pure sine stack with identical envelopes. Instead, give lower partials slower decays. A practical template:

k=1–2: attack 5–15 ms, decay 700–1500 ms
k=3–6: attack 2–8 ms, decay 300–800 ms
k=7–12: attack 0.5–3 ms, decay 120–350 ms

2) Blast shell

This is where horror departs from a “clean” explosion. Build a dense but organized inharmonic set. Instead of strict integer multiples, use a stretched series:

f_k = f0 · k^α, with α ≈ 1.01–1.06

At α=1 you’re harmonic. Slightly above 1 produces a metallic, strained character reminiscent of tearing materials and structural resonance without sounding like a bell.

Base: f0 = 55–90 Hz
Count: 40–120 partials up to ~2 kHz
Amplitude: use a spectral tilt of −3 to −6 dB/octave; then create a formant-like hump around 250–600 Hz to add “throat”
Pitch drop: apply a global exponential glide of −1 to −4 semitones over 120–250 ms, but not uniformly—make higher partials drop faster to mimic dispersive decay

3) Shard field

High-frequency fear cues often come from brief, sparse, bright events that trigger attention systems. Additive makes this precise: generate 20–80 partials between 2–12 kHz, but don’t distribute them uniformly. Use clustered bands (e.g., 3.2–4.6 kHz and 7.5–10.5 kHz) because those regions tend to read as “bite” and “alarm” on many playback systems.

Envelopes: attack 0–1 ms, decay 20–120 ms
Micro-modulation: 20–60 Hz amplitude modulation at small depths (5–20%) on a subset of partials to induce roughness without turning into a synth tremolo
Random phase: prefer random φ to reduce tonal artifacts and prevent a single “ping” dominating

C. Crest factor, punch, and phase management

Engineers often focus on spectrum and ignore phase until something collapses in mono or loses punch after limiting. With additive explosions, phase is a design parameter.

If all partials align in phase at t=0, you can get an unnaturally high crest factor and a “clicky” transient. If all phases are random, the sound can become diffuse and lack impact.

A practical compromise:

LF partials (below 120 Hz): near-aligned phase (e.g., φ randomized within ±30°) to keep pressure coherent.
MF (120 Hz–2 kHz): moderate randomization (±90–180°) to avoid a single “tone.”
HF (2–12 kHz): full randomization for grit and width compatibility.

Measure crest factor before dynamics. Typical designed explosions might sit around 10–18 dB crest factor pre-limiting depending on layering. If you’re consistently above ~20 dB, you may be wasting headroom; if you’re below ~8–10 dB before any processing, it may sound flat unless the aesthetic demands it.

D. Band-dependent time constants: an “explosion envelope diagram”

Visualize envelopes as three stacked curves:

Amplitude
^
|   HF:  |\/\__ (0–120 ms, spiky clusters)
|   MF:  |\/----\____ (30–400 ms, main bloom)
|   LF:  |/---------\__________ (300–1500 ms, pressure tail)
+--------------------------------------------------> time
      0      50     200      800        1500 ms

The point is not the exact shape; it’s the asymmetry: highs die fast, lows persist, mids carry the “event.”

E. Practical data points: recommended starting values

Partial counts: LF 8–16, MF 60–120, HF 20–80 (per layer, not total tracks).
Explosion perceived size: correlate with slower LF decay (≥800 ms) and lower f0 (≤45 Hz), but guard against mud by ensuring MF clears by ~300–500 ms.
Roughness control: keep AM depth modest (≤20%) and restrict to partial subsets; too much reads as “effect.”
Spectral centroid trajectory: for horror, often falls quickly (first 100–200 ms) but with intermittent HF spikes (shards) to maintain threat.

4) Real-world implications and practical applications

Integration with typical post-production chains

Additive explosions rarely ship raw. They become the controllable “synthetic skeleton” underneath organic layers. Standard practices:

Transient management: a fast limiter or clipper can shape the initial 5–20 ms, but use it after you’ve achieved the right phase/crest factor; otherwise you’re solving a synthesis problem with dynamics.
Sub management: high-pass or low-shelf decisions must match deliverable specs (cinema vs nearfield). For smaller speakers, emphasize 60–120 Hz harmonics rather than boosting sub-30 content.
Midrange clarity: carve a notch or dynamic dip around 250–500 Hz if dialogue masking is a concern, but don’t remove it entirely—this band carries “impact realism.”
Spatialization: keep LF mono-compatible; widen HF shards with short decorrelation delays (e.g., 5–20 ms) or mid/side processing. Avoid widening the pressure core.

Why additive helps editorial flexibility

In post, directors ask for “bigger,” “scarier,” “less like a gun,” “more supernatural.” Additive parameters map to those notes:

“Bigger” → lower f0, longer LF decay, slightly increased partial density below 200 Hz.
“More supernatural” → increase α (stretched partials), add subtle frequency jitter, introduce non-integer clusters.
“Less mechanical” → reduce periodic AM, randomize shard timing, vary decay per partial.
“More painful” → add HF clusters around 3–5 kHz with short spikes and controlled roughness.

5) Case studies: professional-style examples

Case study A: “Possessed shockwave” (supernatural blast)

Goal: an explosion that feels like pressure plus an unstable “entity,” not combustion.

LF core: f0=38 Hz; 10 harmonics; p=1.5; k=2–4 boosted +3 dB; decay 1200 ms for k=1–2.
MF shell: f0=76 Hz; α=1.035; 90 partials to 2 kHz; global pitch drop −3 semitones over 180 ms; high partials drop 1.4× faster than low partials.
HF shards: 45 partials, clustered at 3.6–4.4 kHz and 8.2–10.2 kHz; random bursts over 0–90 ms; AM at 33 Hz depth 12% on 30% of partials.

Mix note: Keep LF mono, widen shards; add a short convolution of a concrete stairwell at −18 dB send to produce claustrophobic reflections without washing the transient.

Case study B: “Flesh-and-metal rupture” (body horror explosion)

Goal: impact plus organic tearing. Additive provides the controlled “rupture tone,” while foley handles viscera.

LF: f0=46 Hz; 12 partials; introduce slow FM on k=1–3 at 0.2–0.6 Hz (very subtle) to create unease.
MF: two additive banks:
- Bank 1 (structural): f0=92 Hz, α=1.02, 70 partials, decay 250 ms average.
- Bank 2 (tearing): inharmonic cluster set: choose 18 frequencies between 180–900 Hz spaced by 1.08–1.18× ratios, with independent short decays (40–160 ms).
HF: sparse shards 6–9 kHz, but fewer than normal (10–20 partials) to avoid “glass” and keep it wet/organic.

Editorial trick: sidechain a dynamic EQ band at 2.5–4 kHz keyed from dialogue to prevent the rupture from masking consonants, while leaving the LF and upper air intact.

Case study C: “Distant dread detonation” (scale without brightness)

Goal: the audience feels a huge event far away—mostly LF and low-mid movement, minimal HF.

LF: f0=30–34 Hz; harmonics to 240 Hz; long decay up to 2 s but with a gentle low-pass over time (simulate air absorption perceptually).
MF: reduced density (30–50 partials), α close to 1.01 to keep it less metallic.
HF: nearly none; instead use a faint noisy layer (could still be additive as dense random partials above 4 kHz) but down −25 dB relative to MF.

Mix note: the “distance” is largely in early reflections and reduced HF. Use a longer pre-delay and emphasize low-frequency room modes subtly; keep transient softened (attack 2–5 ms rather than 0 ms).

6) Common misconceptions (and what actually works)

“Additive is too clean for explosions.”
Clean is a choice, not a property. Inharmonic partial sets, time-varying detune, roughness modulation, and band-specific envelopes can be as feral as any distortion chain—while remaining editable.
“More partials automatically means more realism.”
Density without structure becomes noise. Realistic or convincing events have organized spectral motion: evolving centroid, decays that differ per band, and transient detail. Aim for purposeful distributions.
“Random phase is always best.”
Random phase can erase punch, especially in the LF. Coherence in the pressure core often improves impact and translation.
“Just add sub-20 Hz and it will feel huge.”
Many systems won’t reproduce it, and even in theaters, excessive infrasonic content can waste headroom and cause limiter pumping. Size is more reliably conveyed by 30–120 Hz energy plus harmonic implication and controlled decay.
“A single envelope is fine if the EQ is right.”
EQ is static; explosions are nonstationary. Separate envelopes by band/partial group, or you’ll fight the sound later with dynamic EQ and multiband compression.

7) Future trends and emerging developments

Analysis-resynthesis pipelines. Modern tools can extract partial trajectories from recordings (sinusoidal modeling) and let you exaggerate them: stretch inharmonicity, re-time decays, or re-map spectral envelopes. Expect more hybrid workflows where “additive” is the editable representation of recorded chaos.
Perceptual optimization. Instead of hand-tuning partial amplitudes, emerging systems use perceptual cost functions (roughness, loudness models, spectral salience) to automatically propose parameter sets that maximize “threat” while meeting loudness/headroom constraints.
Object-based and immersive mixing constraints. In Atmos and similar formats, keeping LF energy stable while moving upper components is increasingly important. Additive layers naturally separate into mono-stable LF objects and spatialized HF objects.
Realtime procedural sound for games. Additive synthesis scales well to parameter-driven events: distance, occlusion, damage state, supernatural intensity. Expect more horror titles to use additive cores with procedural modulation rather than relying exclusively on one-shot assets.

8) Key takeaways for practicing engineers

Design explosions as evolving spectra, not static EQ curves: band-dependent envelopes are non-negotiable.
Use three additive layers (LF pressure, MF blast, HF shards) with different partial rules and time constants.
Exploit controlled inharmonicity (stretched partial series with α≈1.01–1.06) to push a sound into horror territory.
Manage phase intentionally: coherent LF for punch, randomized HF for grit; measure crest factor before dynamics.
Build roughness surgically with modest AM (15–80 Hz region) on selected partials, not across the entire sound.
Optimize for the delivery system: “bigger” is usually 30–120 Hz and decay behavior, not infrasonic bloat.
Think editorially: additive parameters map cleanly to direction notes (bigger, stranger, more painful, more distant).

Additive synthesis won’t replace recorded debris, distortion, and room interaction—but it can become the most controllable, most repeatable part of an explosion: the spectral engine that makes the audience feel pressure, instability, and dread on command.