---
title: Function generator
lang: en
type: tuto
tags:
  - eee
  - tuto
prerequis:
  - instruments-de-mesure-en
aa: []
phases:
  - preuve-de-concept
  - integration-et-tests
draft: false
source_fr: embarque/mesure/generateur-de-signaux.md
source_sha256: bc54ad4bc457799f3cfbfa968154147e7b9694b84cc072db649ff1aa25607e41
---

**The function generator** (in French labs usually called the **GBF**, for *générateur basse fréquence*, low-frequency generator) is the *reverse* instrument of measurement: instead of reading what the circuit does, it **injects a calibrated signal** into it (sine, square, triangle) whose shape, frequency and amplitude you choose. It lets you test a circuit with a perfectly known input, without waiting for the real signal — sensor, microcontroller — to be available. This is a tool tutorial of the [[instruments-de-mesure-en|measuring instruments]] hub.

![Front panel of a function generator in functional blocks: the display summing up shape, frequency, amplitude and offset; the waveform selection block (sine, square, triangle); the output block with its BNC connector, the Output key and the reminder of the High-Z or 50 Ω load setting; and the three knobs for frequency, amplitude and offset.|640](/ressources/img/generateur-de-signaux/face-avant.svg)

## What is it for?

Testing a circuit requires an input. As long as the sensor is not wired or the program is not written, that input does not exist. The GBF creates it:

- **stimulate** a sub-assembly with a controlled signal: you know exactly what goes in, you watch what comes out on the [[oscilloscope-en|oscilloscope]] — the response is compared with an expected one;
- **temporarily replace** a missing element: simulating the output of an analog sensor to test the conditioning stage or an [[adc-en|ADC]] input before the real sensor arrives;
- **characterise**: sweeping several frequencies to plot a filter's response, raising the amplitude to find the threshold of an input.

![Principle of the function generator: the GBF injects a controlled stimulus into the circuit under test, the oscilloscope watches the response, and the three devices share a common ground.](/ressources/img/generateur-de-signaux/injection.svg)

The GBF and the oscilloscope work as a **pair**: one injects, the other watches. It is the bench version of the simulation / measurement duo, except that here everything is real.

## Essential settings

Four parameters define the signal:

- **the waveform** — sine (frequency response, "analog" signals), square (logic signals, clocks, [[pwm-en|PWM]] testing), triangle (ramps, threshold tests);
- **the frequency** — that of the phenomenon to reproduce;
- **the amplitude** — watch the displayed unit: peak-to-peak (Vpp) most of the time;
- **the offset** — the DC shift of the signal. It is what turns a signal centred on zero into a purely positive one: for a 3.3 V logic input, set a 3.3 Vpp square **with a +1.65 V offset** — never a negative voltage on a microcontroller pin → [[niveaux-de-tension-en|logic levels]].

> [!warning] The classic trap: High-Z or 50 Ω
> The generator computes the displayed amplitude assuming a **50 Ω** load on its output. Connected to a high-impedance input (an ordinary electronic circuit), the real signal is **twice the displayed value**. Before any injection: set the output to **High-Z** in the menus (or mentally divide by two), and **check the real amplitude on the oscilloscope**.

## Connect and inject

1. **Set before connecting**: shape, frequency, amplitude, offset — output disabled.
2. **Check the signal on the oscilloscope** alone, before injecting it into the circuit: real amplitude, real offset.
3. **Connect the grounds**: the shield of the BNC cable carries the generator's ground, to be joined to the ground of the circuit.
4. **Inject and watch** the response — keeping amplitude and offset within the range the circuit tolerates.

## Pitfalls

**Trusting the displayed amplitude.** The High-Z / 50 Ω setting (see the warning) is the first cause of "I send 3.3 V, it receives 6.6". Systematic check on the oscilloscope.

**Injecting negative voltage into a logic input.** A sine or a square without offset goes below 0 V on every period: a microcontroller input does not survive that for long. The offset is not a decorative setting.

**Exceeding what the circuit tolerates.** The amplitude is chosen from the circuit under test (its datasheet, its [[niveaux-de-tension-en|levels]]), not from what the generator can produce.

**Forgetting the common ground.** Without a shared reference, the injected signal means nothing to the circuit — and neither does the measurement at the output.

## Where it fits in the project

- **[[preuve-de-concept-en|Proof of concept]] phase** — testing one stage of the chain without waiting for the others: the processing chain is validated with a synthetic signal before the real sensor is connected.
- **[[integration-et-tests-en|Integration and testing]] phase** — producing **repeatable** test cases: the same ramp, the same square, on every run, which a real sensor never guarantees.

## See also

- [[instruments-de-mesure-en|Measuring instruments]] — the hub: common method and choice of instrument
- [[oscilloscope-en|Oscilloscope]] — the natural teammate: inject on one side, watch on the other
- [[adc-en|ADC]] — the analog input most often tested with a GBF
- [[niveaux-de-tension-en|Logic levels]] — the range to respect before injecting anything