Pulse width modulation (PWM) is a synthesis technique that changes the duty cycle of a pulse wave over time. The waveform’s period can remain the same while the proportion of each cycle spent in its high and low states changes, so the harmonic spectrum moves even though the oscillator’s intended fundamental frequency does not have to change.
PWM is closely associated with lush strings, animated pads, reedy leads, and moving bass tones because a single oscillator can acquire a strong chorus-like sense of motion. The effect begins with a simpler idea, however: pulse width is a static waveform parameter, while pulse width modulation means that parameter is being varied by an LFO, envelope, control voltage, modulation wheel, or another source.
Table of Contents
Pulse wave, square wave, pulse width, and duty cycle
A pulse wave is a periodic rectangular waveform with two states, usually described as high and low.
The important parameter is the proportion of each cycle spent in the high state.
That proportion is the duty cycle:
Duty cycle = Thigh / T × 100%
where:
- Thigh is the time spent in the high state;
- T is the total period of one cycle.
Because frequency is the reciprocal of period:
f = 1 / T
changing the internal division of the cycle does not necessarily change T.
For a 100 Hz oscillator, one cycle lasts:
T = 1 / 100 = 0.01 s = 10 ms
At 50% duty cycle, the waveform might spend:
- 5 ms high;
- 5 ms low.
At 25% duty cycle:
- 2.5 ms high;
- 7.5 ms low.
The period is still 10 ms, so the repetition frequency is still 100 Hz.
What changes is the waveform shape and therefore its harmonic content.
A square wave is one specific pulse wave
A perfect square wave has a 50% duty cycle.
Its high and low portions occupy equal amounts of time.
A 25% pulse wave is not square.
A 10% pulse wave is not square.
So:
every square wave is a pulse wave, but not every pulse wave is a square wave.
This distinction matters because many synthesizers label the oscillator waveform “square” even when a pulse-width control can continuously move it away from 50%.
Pulse width vs pulse width modulation
Pulse width or duty cycle describes the waveform’s current shape.
Pulse width modulation describes movement of that parameter.
If a pulse oscillator is manually set to 30% and left there, its spectrum differs from a square wave, but nothing is being modulated.
If an LFO moves the oscillator continuously from 30% to 70% and back, that is PWM.
The distinction is simple but important because static pulse width already changes timbre. PWM adds time-varying spectral movement.
How pulse width changes the harmonic spectrum
The sound of a periodic waveform depends heavily on which harmonics it contains and how strong they are.
A 50% square wave has a familiar property: in the ideal mathematical case, it contains the fundamental plus odd-numbered harmonics only.
That gives the classic series:
1, 3, 5, 7, 9, …
with harmonic amplitude decreasing as harmonic number increases.
Move away from 50% duty cycle and even harmonics appear.
At particular duty cycles, other systematic harmonic cancellations occur.
This is why turning a pulse-width knob can make the oscillator change from hollow and woody to nasal, thin, bright, or buzzy before any filter is touched.
Fourier series of an ideal pulse wave
For an ideal pulse wave that switches between +1 and -1, with duty cycle d between 0 and 1, the DC component is:
DC = 2d – 1
For harmonic number n ≥ 1, the peak amplitude magnitude is:
An = 4 / (nπ) × |sin(nπd)|
The phase of each harmonic depends on where the pulse is positioned within the cycle, but the magnitude formula is enough to show the major spectral pattern.
At d = 0.5:
sin(nπ/2)
is zero for every even n.
That is why an ideal square wave contains no even harmonics.
At d = 0.25, every fourth harmonic disappears because:
sin(nπ/4) = 0
when n is 4, 8, 12, and so on.
At d = 1/3, every third harmonic disappears.
At d = 0.1, every tenth harmonic disappears.
The harmonic cancellations move as the duty cycle moves.
Harmonic examples
For an ideal bipolar pulse wave:
| Harmonic | 50% duty | 33.3% duty | 25% duty | 10% duty |
|---|---|---|---|---|
| 1 | 1.273 | 1.103 | 0.900 | 0.393 |
| 2 | 0 | 0.551 | 0.637 | 0.374 |
| 3 | 0.424 | 0 | 0.300 | 0.343 |
| 4 | 0 | 0.276 | 0 | 0.303 |
| 5 | 0.255 | 0.221 | 0.180 | 0.255 |
| 6 | 0 | 0 | 0.212 | 0.202 |
| 7 | 0.182 | 0.158 | 0.129 | 0.147 |
| 8 | 0 | 0.138 | 0 | 0.094 |
| 9 | 0.141 | 0 | 0.100 | 0.044 |
| 10 | 0 | 0.110 | 0.127 | 0 |
The exact amplitudes matter less than the pattern.
At 50%, every even harmonic is gone.
At 25%, every fourth harmonic is gone.
At 10%, both odd and even harmonics are densely represented until the first systematic zero at the tenth harmonic.
That is the spectral reason narrow pulse waves sound very different from square waves.
What happens near 0% and 100%
As the duty cycle approaches either extreme, one state occupies almost the entire cycle.
At exactly 0% or 100%, an ideal pulse wave stops alternating and becomes a constant level.
In other words, it becomes DC rather than an alternating audio waveform.
In a normal AC-coupled audio path, the steady DC component is blocked, so the audible output settles toward silence.
Real synthesizers may clamp the pulse-width range before the mathematical extremes, distort the waveform near the limits, or implement their own behavior. That is why some instruments let PWM approach silence while others never quite reach it.
Why PWM sounds like a chorus
A slow triangle or sine LFO moving pulse width can make one oscillator sound surprisingly like two oscillators beating against each other.
That effect is one of PWM’s most useful characteristics.
Gordon Reid’s Synth Secrets analysis offers an intuitive way to understand it: a pulse wave is defined by two repeating transitions, one rising and one falling. If the oscillator period remains fixed while pulse width changes, one edge can remain tied to the oscillator’s regular cycle while the timing of the other edge moves backward and forward within that cycle.
The result can be viewed as a combination of:
- one periodic transition pattern that stays fixed;
- another transition pattern whose timing is being modulated.
That moving timing relationship produces an effect analogous to having one stable component and another whose apparent pitch is being modulated. The ear hears continuous beating and phase interaction, which can resemble the thickness of two slightly detuned oscillators.
This is why slow PWM can create a chorus-like sound without using a delay-based chorus effect.
It is not literally the same signal-processing method as chorus. A chorus effect duplicates or delays audio and modulates delay time. PWM reshapes the oscillator waveform itself.
The perceptual result can overlap even though the mechanism does not.
What PWM sounds like at different settings
PWM is not one sound.
The modulation source, rate, depth, and starting pulse width determine the result.
Slow LFO PWM
This is the classic patch.
A slow LFO moves pulse width back and forth while the note sustains.
Typical results include:
- wide pads;
- synthetic strings;
- animated basses;
- hollow-to-nasal movement;
- a soft beating or swirling quality.
The effect can remain smooth enough that the listener hears one stable note with a moving timbre rather than an obvious rhythmic modulation.
There is no universal “correct” LFO rate. The useful rate depends on register, patch context, modulation depth, and how much motion is wanted.
A slow setting around 1 Hz can be a sensible starting point for a bass or pad, but it is a workflow convention, not a law.
Envelope-controlled pulse width
An envelope produces one directional contour rather than a repeating cycle.
For example:
envelope -> pulse width
can make the waveform start narrow, move through square, and settle at another width.
This is useful when the timbral movement should belong to the attack of the note rather than continue indefinitely.
A short decay envelope can produce a brief nasal or bright transient.
A slower envelope can create a gradual spectral opening or closing without using the filter.
Because pulse width changes the oscillator spectrum before the filter, envelope-driven PWM and filter-envelope movement are not interchangeable.
Mod wheel PWM
A modulation wheel can control:
- pulse width directly;
- PWM depth;
- LFO depth routed to pulse width;
- a macro that combines PWM with other destinations.
The second approach is often more expressive.
Instead of the wheel manually setting a static duty cycle, it can fade in an LFO-driven PWM effect. At the bottom of the wheel, the oscillator remains stable. Push the wheel upward and the spectral movement becomes progressively deeper.
Audio-rate PWM
Pulse width can also be modulated by an audio-rate source.
At that point, the ear no longer hears a slow sweep.
The time-varying waveform produces modulation-related sidebands and a much denser spectrum. Depending on frequency relationships and depth, the result can become metallic, aggressive, buzzy, or highly inharmonic.
Audio-rate PWM therefore belongs to the same broad family of fast timbral modulation techniques as FM synthesis and phase modulation, but it is not the same mathematical operation.
FM changes oscillator phase or instantaneous frequency behavior.
PWM changes the timing relationship between the edges of a pulse wave.
The basic PWM controls on a synthesizer
PWM can appear as dedicated oscillator controls or as ordinary destinations in a modulation matrix.
The usual controls are straightforward.
Pulse width or shape
The base PW control sets the unmodulated duty cycle.
At or near the center, it commonly produces a square wave.
Move away from center and the pulse narrows in one direction or the other.
On some synths the control is shown as:
- Pulse Width;
- PW;
- Shape;
- Waveshape;
- Duty Cycle.
“Shape” may control more than pulse width, so the manual matters.
PWM source
The source might be:
- LFO;
- envelope;
- velocity;
- aftertouch;
- mod wheel;
- sequencer;
- random source;
- external CV;
- another audio-rate oscillator.
In a modulation matrix, pulse width is simply another destination.
PWM amount or depth
Depth determines how far the source moves the duty cycle away from its base value.
A shallow setting produces gentle animation.
A deep setting sweeps through a much larger spectral range.
Depth interacts with the base pulse width.
A 50% starting point leaves equal room in both directions for a bipolar modulator.
Starting near 10% and adding deep negative modulation may drive the pulse to an extreme where the alternating waveform disappears.
Practical PWM patches
Classic one-oscillator string pad
Start with:
- one pulse oscillator;
- pulse width near 50%;
- slow triangle LFO;
- moderate LFO-to-PW depth;
- low-pass filter;
- slow amplifier attack and release.
Listen first with the filter fairly open.
The oscillator itself should already sound wider and more animated than a static square wave.
Then shape the brightness with the filter.
This separation is useful because it reveals which motion comes from PWM and which comes from subtractive synthesis.
PWM bass
Use:
- one pulse oscillator;
- moderately narrow initial pulse;
- slow LFO;
- relatively restrained modulation depth;
- low-pass filter with enough cutoff to preserve upper harmonics.
Narrow pulses can sound lighter in fundamental weight and more aggressive in the upper spectrum. The filter can then rebalance the tone.
If the patch seems to disappear rhythmically, reduce PWM depth or move the base width away from the extreme.
Envelope PWM pluck
Patch:
short envelope -> pulse width
Use a quick attack and short decay.
Set the base pulse width so the transient starts from an asymmetric pulse and moves toward a more stable width.
This creates spectral motion during the attack without relying entirely on a filter envelope.
The result can add a sharp, reed-like or brassy articulation to a simple oscillator.
Performance-controlled PWM
Patch:
LFO -> pulse width
then route:
mod wheel -> LFO-to-PW amount
The LFO generates the cyclic movement.
The mod wheel controls how much of that movement reaches the oscillator.
This is a practical example of nested modulation: one modulation source controls the depth of another modulation route.
External modular PWM
In a modular synth:
LFO -> attenuator -> PWM CV input
The attenuator becomes the modulation-depth control.
For an envelope-controlled variation:
envelope -> attenuverter -> PWM CV input
The attenuverter can reverse the envelope direction.
The new Control Voltage article covers attenuation, inversion, offset, and voltage-range matching in more detail.
Analog PWM: the comparator method
A classic analog way to generate a variable-width pulse wave is to compare a continuously changing waveform against a threshold.
Imagine a sawtooth ramp rising through a voltage range.
A comparator asks a simple question continuously:
Is the ramp above or below the threshold voltage?
If it is above the threshold, output one state.
If it is below the threshold, output the other.
The result is a rectangular waveform.
Move the threshold and the ramp crosses it earlier or later in each cycle.
That changes the pulse width.
Modulate the threshold with an LFO or envelope and the pulse width changes continuously.
The master ramp can keep repeating at the same rate, so the pulse waveform changes duty cycle without requiring the oscillator period itself to change.
Doepfer’s current A-168-1 PWM Generator is a particularly clear hardware example. The module uses comparators to derive an adjustable rectangle signal from a triangle, sawtooth, sine, envelope, or other continuously changing input, and it provides a CV input for modulating the comparison threshold.
The Doepfer A-110-1 VCO uses the same practical idea at oscillator level: its rectangle output has manual pulse-width control plus voltage control for PWM.
Digital PWM and aliasing
An ideal pulse wave contains instantaneous vertical transitions.
Mathematically, those discontinuities imply harmonics extending without limit.
An analog circuit is never truly infinite in bandwidth, but a digital oscillator faces a specific problem: frequencies above the Nyquist frequency cannot be represented uniquely at a given sample rate.
At 44.1 kHz:
Nyquist frequency = 22.05 kHz
If a naive digital oscillator simply outputs +1 below a phase threshold and -1 above it, the discontinuities generate harmonics well above Nyquist.
Those components fold back into the audible spectrum as aliasing.
The problem gets more demanding with PWM because the falling edge moves as duty cycle changes.
A good digital oscillator therefore needs some form of band-limiting or anti-aliasing strategy.
BLEP, PolyBLEP, and related methods
One family of techniques corrects the discontinuities rather than accepting the naive vertical steps.
BLEP, short for band-limited step, models a band-limited version of a discontinuity.
PolyBLEP uses a short polynomial correction around the discontinuity, making it efficient enough for real-time oscillator use.
Research on virtual analog oscillator design also includes band-limited impulse trains, polynomial band-limited functions, oversampling, table-based approaches, and other anti-aliasing methods.
The exact algorithm depends on the synthesizer.
A producer does not need to know the implementation to use PWM, but the DSP explains why two digital pulse oscillators can sound different even when their front panels show the same 50% width and modulation depth.
Band-limited wavetables
Another approach is to use precomputed or dynamically selected band-limited waveform tables.
A synth can store pulse-like waveforms appropriate for different pitch ranges, then interpolate or switch between them while controlling duty cycle.
This connects conceptually with wavetable synthesis, although a virtual analog oscillator using tables internally is not automatically a wavetable synth in the musical sense.
Implementation technique and user-facing synthesis method are not always the same category.
PWM vs other synthesis and modulation techniques
PWM vs frequency modulation
PWM changes duty cycle.
FM changes frequency or phase behavior through modulation.
Slow FM applied to pitch sounds like vibrato.
Slow PWM usually sounds like timbral movement with a stable repeating period.
At audio rates, both can generate complex sideband structures, but the underlying operations are different.
PWM vs amplitude modulation
Amplitude modulation changes signal level over time.
Slow AM produces tremolo.
PWM leaves the nominal high and low waveform levels alone and changes how long the waveform remains in each state.
The oscillator’s spectral balance changes because the waveform shape changes.
PWM vs phase modulation
Phase modulation changes the instantaneous phase of an oscillator.
PWM changes the relative timing of the rising and falling edges of a pulse wave.
The concepts can interact mathematically, but the controls and signal-generation mechanisms are not interchangeable.
PWM vs phase distortion synthesis
Phase distortion synthesis, associated strongly with the Casio CZ architecture, warps the phase progression used to read a waveform.
It can create pulse-like or resonant spectra, but it does not require the comparator-style threshold mechanism of classic analog PWM.
Similar-looking waveforms do not imply identical synthesis methods.
PWM vs chorus
PWM can sound chorused.
That does not make PWM a chorus effect.
A conventional chorus creates time-varying delay relationships between signal paths.
PWM generates movement at the oscillator itself by changing pulse width.
PWM vs pulse-code modulation
Pulse width modulation and pulse-code modulation share the abbreviation “pulse” but describe unrelated processes.
Pulse-code modulation, or PCM, is a method of representing sampled audio numerically.
PWM is a waveform-shaping and modulation technique.
The Minimoog Model D and static pulse widths
The Minimoog Model D is a useful example of the difference between pulse width and pulse width modulation.
Its oscillator waveform selector includes:
- square;
- wide rectangle;
- narrow rectangle.
The current Moog manual explains how those shapes differ in harmonic content.
What the classic Model D architecture does not provide is a continuously variable pulse-width parameter that an LFO can sweep across the whole range as a normal PWM destination.
So the instrument provides multiple static pulse widths, not classic continuous PWM.
That distinction is easy to miss because a synth can contain pulse waveforms without having pulse width modulation.
Current hardware examples
Sequential Prophet-5
The current Sequential Prophet-5 exposes pulse width directly in the oscillator section.
Sequential’s manual describes the Pulse Width control as moving from square at the center toward narrow pulse shapes at either side, and explicitly recommends modulating pulse width with the LFO or Poly Mod to add movement and thickness, especially for pads and string-like sounds.
That makes it a clean current example of PWM inside a full polyphonic subtractive synthesizer.
Affiliate disclosure: MPG may earn a commission from qualifying purchases made through affiliate links, at no extra cost to you. See the affiliate disclosure.
Sequential Prophet-5 at Thomann is a current five-voice analog polysynth with per-oscillator pulse-width control and modulation routing suitable for classic PWM patches.
Doepfer A-110-1 Standard VCO
The Doepfer A-110-1 is a modular example with fewer abstractions.
Doepfer documents:
- simultaneous rectangle, sawtooth, triangle, and sine outputs;
- manual pulse-width control;
- voltage-controlled PWM;
- two pulse-width CV inputs, one with attenuation.
That makes the signal path easy to understand because the source, modulation voltage, depth, and pulse output are all visible as separate modular functions.
Doepfer A-110-1 Vintage Edition at Thomann is a current Eurorack VCO in the A-110-1 family.
Common PWM mistakes and myths
“A square wave and a pulse wave are the same thing”
A square wave is one type of pulse wave.
Its duty cycle is 50%.
Change the duty cycle and the waveform is still a pulse wave, but it is no longer square.
“Turning the pulse-width knob is PWM”
Not necessarily.
If the control is left in one position, that is a static pulse width.
PWM requires the width to change over time.
Manual movement technically changes the pulse width over time, but in synth programming the term usually refers to continuous modulation from an LFO, envelope, CV, or other modulation source.
“PWM changes the oscillator’s pitch”
Classic PWM does not require the oscillator period to change.
The intended fundamental repetition frequency can remain fixed while duty cycle and harmonic amplitudes move.
At extreme widths the fundamental can become weak, the spectrum can become very thin, and fast modulation can create sidebands that complicate pitch perception. So “pitch never changes perceptually” is too absolute.
The more precise statement is that pulse-width modulation can reshape the spectrum without changing the oscillator’s base period.
“PWM is just a chorus effect built into the oscillator”
No.
PWM can produce a chorus-like perception, but the mechanism is waveform reshaping rather than modulated delay.
“The Minimoog Model D has classic PWM”
It has several selectable pulse waveforms.
It does not provide the continuously variable LFO-driven pulse-width system associated with classic PWM.
“DCO means the PWM is digital”
Not necessarily.
A digitally controlled oscillator can use digital timing or control while generating or shaping the audio waveform in analog circuitry.
DCO describes how oscillator frequency is controlled, not a universal statement about the entire audio signal path.
“Digital PWM always aliases badly”
Only naive digital implementations do.
Well-designed digital oscillators use band-limiting, oversampling, BLEP-family correction, band-limited tables, or other methods to reduce aliasing.
“More PWM depth always sounds richer”
No.
Depth is useful only while the resulting motion serves the patch.
Push the duty cycle too close to 0% or 100% and the waveform may become extremely thin, click, sputter, or disappear depending on the implementation.
The useful range is musical, not maximal.
Related terms
- Control Voltage: the analog control medium often used to drive pulse width in modular and analog synths.
- LFO: the classic cyclic modulation source for slow PWM.
- Envelope: useful for one-shot pulse-width movement tied to a note.
- Modulation: the broader source-to-destination framework that includes PWM.
- Oscillator: the signal source whose pulse waveform is being reshaped.
- Subtractive Synthesis: the synthesis architecture most strongly associated with classic PWM.
- FM Synthesis: a different audio-rate modulation method that changes frequency or phase relationships.
- Wavetable Synthesis: can reproduce or extend pulse-like spectral movement through stored and interpolated waveform frames.
- Waveform: the time-domain shape whose duty cycle defines a pulse wave.
- Nyquist Frequency: the digital frequency limit that makes anti-aliasing important for pulse oscillators.
References and further reading
- Sound On Sound: Synthesizing Strings, PWM & String Sounds
- Reason Studios: Thor Demystified 3, Pulse Width Modulation
- Doepfer A-110-1 Standard VCO
- Doepfer A-168-1 PWM/Comparator
- Sequential Prophet-5 User’s Guide
- Moog Minimoog Model D Manual
- Aalto University: Filter-Based Oscillator Algorithms for Virtual Analog Synthesis