Sensors, Analogue-to-Digital Conversion, Sampling and Data Logging

The world is analogue; the computer is digital. Temperature, pressure, light, sound and position vary continuously. A sensor converts a physical quantity into an electrical signal, usually a voltage that varies continuously with the quantity. A computer cannot store a continuously varying voltage; it needs a number. An analogue-to-digital converter (ADC) measures the voltage at an instant and outputs a binary number. On the way back out, a digital-to-analogue converter turns a number into a voltage that drives an actuator — a motor, heater, valve or speaker. The full chain is: physical quantity → sensor → ADC → processor → DAC → actuator, and every control system, from a thermostat to a self-driving car, is built on it.

Two parameters, two kinds of error. Sampling rate is how many times per second the ADC takes a reading. Sample too slowly and a short peak between two samples is missed entirely; the rule is to sample at least twice as fast as the fastest change you need to capture. Resolution is the number of bits per sample, which sets how many distinct levels can be distinguished: 8 bits give 256 levels, 12 bits give 4 096. A range of 0–100 °C at 8 bits has a step of about 0.4 °C, so every reading is rounded to the nearest step — quantisation error. Raising either parameter improves accuracy but multiplies the data produced and the power consumed, which matters on a battery. The correct choice depends on the phenomenon: a room's temperature once a minute is enough; vibration in a machine bearing needs thousands of samples a second.

Data logging. A data logger combines sensors, a clock and memory, and records readings at set intervals without a person present — for hours, months or years. It is used where humans cannot stay (a river gauge, a weather station on a mountain), where readings must be continuous (a patient monitor), or where the phenomenon is slow (glacier movement). Advantages over manual recording: no gaps, no transcription errors, readings at night and in hazardous places, timestamps that are exact, and data that is already digital for analysis. Costs: equipment, calibration, and the risk that a faulty sensor produces plausible-looking wrong data for months before anyone notices — which is why loggers are checked against a manual reading periodically.

Sensor limitations to mention. Every sensor has a range outside which it saturates, an accuracy (how close to the true value), a precision (how repeatable), a response time (a thermometer takes seconds to follow a change), and drift over time that requires recalibration. Noise from electrical interference adds random error, which is reduced by averaging several samples — at the cost of slower response. An answer that names the sensor and its limitation, and then says what the system does about it, is what the mark scheme rewards.