A sensor is a device that produces a signal in response to a physical phenomenon. A microphone produces a voltage in response to air pressure. A photodiode produces a current in response to light. A thermocouple produces a voltage in response to a temperature difference. After conditioning and digitization, the signal coming out of the sensor is our data.
The signal chain typically runs sensor → analog conditioning (amplification, filtering) → analog-to-digital converter → digital storage. Each stage can introduce its own noise and distortion, which is why the choice of sensor is only the first of several engineering decisions.
The phrase garbage in, garbage out applies first and most forcefully to sensors. A noisy, miscalibrated, or wrongly-placed sensor produces data that no amount of clever modelling can rescue.
Common signal sources in engineering applications include the IMU for measuring motion (which combines an Accelerometer, a Gyroscope, and a Magnetometer — three transducers in the strict sense) and biopotential electrode arrays like EEG for brain activity and ECG for heart activity. Electrodes are technically conductors picking up biological voltages rather than transducers converting one physical quantity to another, but the downstream signal-processing pipeline treats them the same way. For data not coming from sensors at all, the alternative is Web scraping.