Data Acquisition Systems Basics: Complete Beginner's Guide

A data acquisition (DAQ) system is the chain of hardware and software that captures real-world signals and turns them into digital data a computer can store and analyse. A sensor produces a small electrical signal, signal conditioning cleans it up and scales it, an analog-to-digital converter (ADC) digitises it at precise time intervals, and software logs and displays the result.

DAQ matters because modern engineering runs on recorded evidence, not on watched meters. A vibration test on a bridge, a temperature profile of a furnace, or a week's worth of weather data all need thousands of samples captured automatically and accurately. The quality of that data is decided by a few settings — sample rate, resolution and input range — and getting them wrong quietly ruins the experiment.

Key Concepts

• Signal conditioning: amplification, filtering and isolation that prepare the raw signal for the ADC.
• Sampling: measuring the signal at regular intervals; the sample rate is expressed in samples per second (S/s).
• Nyquist criterion: the sample rate must be at least twice the highest frequency present (fs ≥ 2 × fmax), or higher frequencies fold back as false low-frequency "aliases".
• Aliasing: the distortion that appears when sampling too slowly; prevented by sampling fast enough plus an anti-aliasing low-pass filter before the ADC.
• Resolution: an n-bit ADC divides its input range into 2ⁿ steps; the quantisation step = range / 2ⁿ.
• Multiplexing: one ADC shared across many channels by switching between them; simultaneous-sampling DAQ gives each channel its own ADC.
• Input range: matching the ADC's voltage range to the signal size so that all the bits are actually used.

Formulas

• Nyquist minimum: fs ≥ 2 × fmax. In practice, use 5–10× fmax for a faithful waveform shape.
• Quantisation step: q = Vrange / 2ⁿ. A 12-bit ADC on a 0–10 V range: q = 10/4096 ≈ 2.44 mV.
• Quantisation noise (RMS) ≈ q/√12 — the theoretical noise floor of the ADC.
• Data rate = channels × sample rate × bytes per sample. Example: 8 channels × 10 kS/s × 2 bytes = 160 kB/s.

Worked Example

1. You must record a 100 Hz vibration signal with a 10 V peak-to-peak swing using a 12-bit DAQ.
2. Nyquist says fs ≥ 200 S/s; choose 1 kS/s (10×) with an anti-aliasing filter around 400 Hz.
3. Set the input range to ±5 V (10 V span): step = 10/4096 ≈ 2.44 mV, about 0.024% of full scale — far finer than the sensor's own accuracy.
4. A ±10 V range would double the step to 4.88 mV and waste half the codes, since the signal only swings ±5 V. Always match the range to the signal.

Common Misconceptions

• "Faster sampling is always better." Beyond about 10× the signal frequency you mostly gain bigger files, not better data; the right rate is set by Nyquist plus a sensible margin.
• "A 16-bit DAQ is automatically more accurate." Extra bits only help if noise and the sensor allow it; a noisy 16-bit system can be worse than a clean 12-bit one.
• "The DAQ replaces the sensor." The DAQ only digitises what the sensor and conditioning deliver; a bad sensor gives precisely recorded garbage.

Real-World Applications

• Vibration and modal testing of machines, vehicles and structures.
• Engine and vehicle test cells recording hundreds of channels.
• Weather and environmental monitoring stations.
• Industrial process logging and quality records.
• Biomedical signal recording such as ECG and EEG.

A DAQ system is a chain — sensor, conditioning, ADC, software — and data quality is set by its weakest link. Sample faster than twice the highest frequency, match the range to the signal, and filter before you sample.