FM / PM Operator Texture — User Guide
A six-operator phase-modulation texture generator with three routing algorithms, operator-level envelopes, genuine one-sample feedback in Operator 6, optional melodic demonstration, and a separate global amplitude stage.
What this does
FM / PM Operator Texture constructs a network of six sinusoidal operators. Depending on the selected preset or advanced routing choice, operators are either heard directly as parallel carriers or used as phase modulators inside one or two operator stacks.
The synthesis has two time scales. Each operator has its own local envelope, which controls either audible carrier amplitude or modulation index. After the note or melody has been rendered, a separate global envelope can reshape the complete output.
operator envelopes → six-operator PM network → note edge protection → optional melody concatenation → Master volume → global amplitude envelope → final edge protection → optional down-only peak protection
The output is always mono and is synthesized directly at 44.1 kHz.
Why the guide says FM / PM
The digital operator engine is implemented as phase modulation:
carrier = A(t) × sin(2π fcarrier t + modulator(t))
The modulator signal is added directly to the downstream operator's phase in radians. With a sinusoidal modulator of fixed amplitude, this produces the familiar sideband structure associated with classic digital FM synthesis; here the operator amplitudes can also evolve through time, so the implementation is more precisely described as an operator-PM engine.
For a carrier operator, Level is audible output amplitude. For an operator used as a modulator, Level becomes the peak phase-deviation scale — effectively the modulation index in radians after its envelope and Brightness scaling are applied.
Operator 6 feedback
Operator 6 can feed its previous output sample back into its own phase:
y[n] = A[n] × sin(phase[n] + feedback × y[n-1])
This is genuine one-sample recursive feedback. It is not the earlier analytic approximation sin(phi + fb*sin(phi)). In Parallel mode Op6 is an audible self-feedback carrier; in Series and Dual modes it is the topmost feedback modulator.
Routing algorithms
| Algorithm | Signal path | Carrier operators |
|---|---|---|
| Parallel additive bank | Op1 + Op2 + Op3 + Op4 + Op5 + Op6(feedback) | All active operators are audible carriers. Their sum is divided by √(number of active operators). |
| Series 6→5→4→3→2→1 | Op6(feedback) → Op5 → Op4 → Op3 → Op2 → Op1 → OUT | Op1 only. Ops2–6 modulate the phase of the next operator downstream. |
| Dual 6→5→4 + 3→2→1 | Op6(feedback) → Op5 → Op4Op3 → Op2 → Op1then Op4 + Op1 | Op4 and Op1. Their sum is divided by √(number of active carrier outputs). |
Setting an operator Level to zero disables that operator; it does not reroute or bypass the preceding operator. For example, a zero intermediate modulator in a stack outputs zero modulation rather than automatically connecting its upstream source around it.
Brightness and effective operator levels
Brightness is not a filter cutoff. It changes the effective operator levels before synthesis:
| Algorithm | Brightness behavior |
|---|---|
| Parallel | Op1 is unchanged; Op2 is multiplied by √Brightness; Ops3–6 are multiplied by Brightness. Because all six are carriers here, this acts as a spectral-gain tilt toward the additional operators. |
| Series | Carrier Op1 is unchanged; modulator indices Ops2–6 are multiplied by Brightness. |
| Dual | Carriers Op1 and Op4 are unchanged; modulator indices Ops2, Op3, Op5 and Op6 are multiplied by Brightness. |
Allowed Brightness is greater than 0 and no more than 4.
Presets
Each preset defines the routing algorithm, six frequency ratios, six operator levels and local envelopes, Op6 feedback, and the operator decay-time references. The main-form controls Base frequency, Brightness, Decay scale, Global envelope, Duration, Master volume, Peak protection, Draw, and Play remain user controls.
| Preset | Algorithm | Ratios 1–6 | Levels 1–6 | Op6 FB | Local character |
|---|---|---|---|---|---|
| Bright Dual-Stack Keys | Dual | 1, 1, 14, 1, 1, 14 | 0.80, 1.50, 2.80, 0.55, 1.00, 2.20 | 0.30 | Decay/decay/snap + decay/decay/snap; snap 0.10 s, tone 0.85 s. |
| Feedback Bass Stack | Series | 1, 1, 2, 1, 0.5, 1 | 0.90, 1.40, 2.20, 0.90, 0.70, 0.80 | 2.20 | Mixed decay/snap/sustain; snap 0.055 s, tone 0.55 s. |
| Metallic Dual Stack | Dual | 1, 3.5, 10, 1.01, 2.75, 13 | 0.70, 1.80, 1.30, 0.55, 1.50, 1.00 | 0.45 | Two detuned/integer-related stacks with decay/snap envelopes; snap 0.18 s, tone 1.40 s. |
| Parallel Organ Spectrum | Parallel | 0.5, 1, 2, 3, 4, 8 | 0.75, 1.00, 0.68, 0.48, 0.30, 0.18 | 0 | Six near-sustain carriers. |
| Detuned Parallel Pad | Parallel | 1, 1.008, 2, 2.013, 3, 0.5 | 0.60, 0.55, 0.30, 0.28, 0.15, 0.18 | 0 | Six slow-attack carriers; tone reference 1.50 s. |
| Simple Series Test | Series | 1, 1, 2, 1, 1, 1 | 0.90, 1.60, 2.20, 0, 0, 0 | 0 | Three active operators for a compact 3→2→1 PM chain. |
Operator envelopes
Every operator can use one of four local level shapes. Decay scale multiplies the decay or slow-attack time references while retaining short-note safety.
| Envelope | Implementation |
|---|---|
| snap | Very short attack (up to 2 ms), followed by exponential decay using Snap decay time × Decay scale. |
| decay | Short attack (up to 10 ms), followed by exponential decay using Tone decay time × Decay scale. |
| sus | Short attack to 1, then a small 10% exponential settling component approaches a 0.9 sustain level. |
| slow | Linear attack lasting up to 0.5 × Decay scale seconds, capped at 60% of the note, then held at full level. |
Panel B of the visualization draws these effective envelopes after Brightness has been applied, so modulator curves represent the actual PM-index trajectories used for the representative note.
Global amplitude envelopes
The selected global envelope is applied after note or melody synthesis and after Master volume. It affects the complete final Sound, not the individual operator envelopes.
| Envelope | What it does |
|---|---|
| No Envelope | No additional musical amplitude shape. |
| Percussive | Exponential decay exp(-5x / DecayScale). |
| Slow Fade | Slower exponential decay exp(-0.30x / DecayScale). |
| Gate | Intentional hard 50%-duty periodic gate. Period = max(0.04, 0.10 / Brightness) s. |
| Reverse | A linear crescendo from 0 to full level across the Sound. It does not reverse samples or time. |
| Tremolo | Sinusoidal amplitude modulation. Rate = 5 + 5×Brightness Hz; depth also increases with Brightness and is limited to 0.95. |
| Swell | Linear attack to full level; attack reference = 0.30×DecayScale s, clipped safely to the Sound duration. |
| ADSR | Short-duration-safe attack, decay, sustain and release. Sustain increases moderately with Brightness; decay/release scale with Decay scale. |
| Stutter | Intentional hard on/off steps with step rate 10 + 10×Brightness. |
| Random Bursts | A Poisson field of short Hann windows. Target event rate = 2 + 3×Brightness bursts/s; each burst lasts 35–110 ms. Overlapping windows are combined by their maximum, so the gate never exceeds 1. |
Random Bursts in v0.4.1
The Poisson process now runs for the complete requested duration. The former silent 300-event truncation has been removed. A high 5000-event runaway guard remains; if it is ever reached, the script exits explicitly instead of silently omitting the rest of the stochastic envelope.
Random seed affects Random Bursts. Zero uses the current unpredictable random state; a positive seed reproduces the burst times and durations for the same settings. After seeded generation, Praat's unpredictable random initialization is restored.
Melody demo
With Melody demo enabled, the selected PM patch is rendered as an eight-note demonstration rather than one sustained note. The frequency ratios relative to Base frequency are:
1, 1.259921, 1.498307, 2, 1.498307, 1.259921, 1, 0.749154
This corresponds to a root–major-third–fifth–octave–fifth–third–root–lower-fifth contour in 12-tone equal temperament.
The base duration pattern is 0.4, 0.4, 0.4, 0.6, 0.4, 0.4, 0.4, 0.8 s, multiplied by Duration_s / 2. Therefore Duration_s is a scale control in demo mode, not the final phrase duration:
demo duration = 1.9 × Duration_s
The notes are concatenated without overlap. Each note has its own very short local edge fade (up to 5 ms). Master volume and the selected global envelope are then applied once to the complete concatenated phrase.
Advanced operator controls
Enable Edit operator details to open three compact pages after the preset has been loaded. The pages therefore fine-tune the selected preset rather than starting from unrelated generic values.
| Page | Controls |
|---|---|
| 1/3 | Algorithm; Op1 and Op2 frequency ratios, Levels, and envelopes. |
| 2/3 | Op3 and Op4 frequency ratios, Levels, and envelopes. |
| 3/3 | Op5 and Op6 frequency ratios, Levels, envelopes; Op6 feedback; Snap decay time; Tone decay time; Random seed. |
Frequency ratio means operator frequency = ratio × note fundamental. Operator levels must be 0–10; Op6 feedback must be 0–4.
Output, bandwidth QC, and level
| Property | Behavior |
|---|---|
| Channels | Mono. |
| Sample rate | Fixed 44,100 Hz. |
| Single-note duration | Exactly Duration_s. |
| Demo duration | 1.9 × Duration_s. |
| Single-note name | fm_pm_<preset>. |
| Demo name | fm_pm_<preset>_demo. |
Master volume and peak protection
Master volume multiplies the synthesized Sound before the global envelope. There is no note-level peak normalization and no automatic final target normalization.
If Peak protection is enabled, the final peak is measured after the global envelope and short outer edge fade. Only when the peak exceeds 0.95 does the script apply Scale peak: 0.95. This is therefore a down-only safety ceiling: quieter results are left unchanged.
Practical bandwidth checks
The script first rejects settings whose highest operator fundamental (maximum ratio × maximum note fundamental) exceeds 0.45 × 44.1 kHz. It also computes a rough PM spectral-top estimate from the largest effective modulation index. If that estimate exceeds the same practical headroom, the script reports a QC warning rather than changing the synthesis.
This warning is diagnostic: the engine does not oversample or automatically remove upper PM sidebands.
Visualization and QC
| Panel | What it actually shows |
|---|---|
| A — Operator Routing | The selected Parallel, Series, or Dual topology. Carriers are marked C, modulators M, and Op6 feedback is annotated when non-zero. |
| B — Effective Operator Levels | The six local operator-envelope trajectories for a representative note after Brightness scaling. In modulator roles these levels are PM-index trajectories in radians. |
| C — Model → Measurement | Measured output spectrogram with nominal carrier-frequency guides. Parallel mode shows all carrier fundamentals; Series shows Op1; Dual shows Op1 and Op4. In demo mode the guides follow the eight note regions. |
| D — Measured Output Spectrum | The actual output spectrum. PM sidebands are measured from the rendered Sound rather than inferred from the operator ratios. |
The QC strip reports routing, active operator count, Op6 feedback, all six ratios, operator-frequency headroom, the practical PM spectral-top estimate, Master volume, pre-protection peak, final peak, RMS, and whether peak protection was applied.
Further Reading
- Chowning, J. M. (1973). “The Synthesis of Complex Audio Spectra by Means of Frequency Modulation.” Journal of the Audio Engineering Society, 21(7), 526–534. The classic digital FM synthesis paper; directly relevant to the sideband behavior produced by sinusoidal phase/frequency modulation.