Signal Generator

Shipping

signal_generator

Periodic waveform source: sine, square, triangle, sawtooth, pulse, DC

Signature

Outputs

signalSignal

Parameters

Key	Type	Default	Notes
`waveform`	enum	sine	one of: sine, square, triangle, sawtooth, pulse, dc
`frequency`	float	1.0Hz
`amplitude`	float	1.0
`offset`	float	0.0
`phase`	float	0.0°
`duty_cycle`	float	50%
`symmetry`	float	50%
`sample_rate`	float	1000.0Hz
`duration`	float	1.0s

Description

Signal Generator is a source node that synthesizes a periodic waveform sampled uniformly in time. It produces one output signal on port signal with no inputs. Six shapes are available via the waveform parameter: sine, square, triangle, sawtooth, pulse, and dc. The generated waveform is $y [n] = A f (θ_{n}) + b$ , where $A$ is amplitude, $b$ is offset, and $f$ is the selected shape evaluated at the running phase $θ_{n} = 2 π F n / f_{s} + ϕ_{0}$ with frequency $F$ , sample rate $f_{s}$ , and initial phase $ϕ_{0}$ .

The time grid is fully specified by sample_rate and duration: the output holds $N = ⌊ f_{s} \cdot T ⌋$ samples at spacing $1/ f_{s}$ . phase is entered in degrees and converted to radians internally; duty_cycle and symmetry are entered as percentages (0-100) and converted to 0-1 fractions internally. duty_cycle sets the high fraction of a period for square/pulse; symmetry sets the rising-edge fraction for triangle. For dc, frequency and phase are ignored and the output is the constant $b$ (plus $A$ where applicable).

The node is stateless: it holds no accumulator or persistent phase between runs, so each evaluation reproduces the same buffer for identical parameters. The output carries no measurement uncertainty (the waveform is a deterministic mathematical construct, $σ = 0$ ); amplitude and offset are dimensionless magnitudes, while frequency, sample_rate (Hz) and duration (s) define the physical time axis. To inject stochastic error, sum this signal with noise_generator.

Mathematics

y [n] = A f (2 π \frac{F}{f _{s}} n + ϕ_{0}) + b, n = 0, 1, \dots, N - 1, N = ⌊ f_{s} T ⌋

f_{square} (θ) = {+ 1 - 1 (\frac{θ}{2 π} mod 1) < d otherwise, d = \frac{duty_cycle}{100}

f_{tri} (θ) = ⎩ ⎨ ⎧ \frac{2 τ}{s} - 1 1 - \frac{2 ( τ - s )}{1 - s} τ < s τ \geq s, τ = \frac{θ}{2 π} mod 1, s = \frac{symmetry}{100}

Examples

A 1 kHz sine sampled at audio rate

Set waveform = sine, frequency = 1000, amplitude = 1.0, sample_rate = 48000, duration = 0.5. This yields $N = 24000$ samples of $y [n] = sin (2 π \cdot 1000 n /48000)$ , i.e. a clean tone with 48 samples per cycle. Feed signal directly into fft to recover a single spectral line at 1 kHz, or into audio_output to hear it.

Asymmetric pulse train for a PWM stimulus

For a 25%-duty control pulse: waveform = pulse, frequency = 50, duty_cycle = 25, amplitude = 1.0, offset = 0. Each 20 ms period is high for the first 5 ms and low for the remaining 15 ms.

period = 1 / 50 Hz = 20 ms
high   = 0.25 * 20 ms = 5 ms
low    = 15 ms

Applications

Generating reference tones and sweeps to characterize filters, amplifiers, and audio chains (drive an `iir_filter` or `fft` and measure the response).
Producing deterministic test stimuli (square, pulse, sawtooth) for control-loop, PWM, and digital-logic simulation via `schmitt_trigger` or `comparator`.
Building synthetic ground-truth signals for validating detection and analysis algorithms such as `peak_detection`, `envelope`, or `derivative`.
Teaching and demonstrating sampling, aliasing, and Fourier concepts by pairing controlled `frequency`/`sample_rate` ratios with spectral nodes.

Neat

Choosing `frequency` close to or above `sample_rate / 2` deliberately produces aliasing, making the node a compact demonstrator of the Nyquist limit: a 900 Hz sine at $f_s = 1000$ Hz folds back to an apparent 100 Hz tone.

`triangle` and `sawtooth` are the same family: a triangle with `symmetry = 100` (all rising edge) is a rising ramp, and `symmetry = 0` is a falling ramp, so the sawtooth is the degenerate limit of the symmetry control.

Known issues

The sample count is $N = \lfloor f_s \cdot \text{duration}\rfloor$; when $f_s \cdot T$ is non-integer the final fractional sample is truncated, so the emitted duration can be slightly shorter than requested.

Non-band-limited shapes (`square`, `sawtooth`, `pulse`, sharp `triangle`) contain harmonics extending past Nyquist; at modest `sample_rate` these fold back as aliasing artifacts on the ideal edges.