Instrumentation

A complete measurement system: sensor, signal conditioning, instrumentation amplifier, filtering, ADC, processing, calibration and communication. Learn CMRR, anti-aliasing, resolution vs ENOB, uncertainty budgets and industrial DAQ architecture.

Instrumentation engineering designs the full chain that turns a physical quantity into a trustworthy number in a control system or database. Each stage — excitation, amplification, filtering, conversion, scaling and transmission — adds gain and also adds error.

Good instrumentation is about budgeting: noise, offset, drift, linearity, bandwidth and latency, so the final measurement meets its specification with a documented uncertainty.

An analogy

A measurement chain is like a relay race: each runner (stage) must hand over the baton (signal) cleanly; one slow or clumsy runner ruins the team's time no matter how good the others are.

The measurement chain

Sensor → excitation/conditioning → amplifier → filter → ADC → digital processing (scaling, linearisation, calibration) → display, storage or communication (4–20 mA, Modbus, Ethernet). Each block must be specified for range, accuracy, bandwidth and noise.

Instrumentation amplifiers

An in-amp amplifies the difference between its inputs while rejecting the common-mode voltage both share — essential for bridges, thermocouples and shunts. High CMRR (100–130 dB), high input impedance and gain set by one resistor are its defining features.

Filtering and sampling

Before the ADC, an anti-aliasing low-pass filter removes content above f_s/2, which would otherwise fold into the band of interest. Sigma-delta ADCs oversample heavily so a simple RC is enough; SAR ADCs need a sharper filter or a higher sample rate. Digital filtering then sets the final bandwidth and rejects 50/60 Hz.

Resolution vs accuracy vs ENOB

A 24-bit ADC rarely delivers 24 useful bits: noise limits the ENOB, and offset, gain error, INL and reference drift limit accuracy. System calibration (two-point offset/gain, or multi-point for non-linearity) removes systematic errors; averaging reduces random noise by √N.

Uncertainty and industrial architecture

The GUM method combines independent standard uncertainties in quadrature and multiplies by a coverage factor (k = 2 for ~95 %). Industrial systems add isolation (galvanic barriers), intrinsic safety in hazardous areas, redundancy, timestamping and networks (PROFINET, EtherNet/IP, OPC UA) feeding SCADA and historians.

Formulas

• In-amp gain (INA128 style): G = 1 + 49.4 kΩ / R_G
• Common-mode error (output): V_err = V_cm · G / 10^(CMRR/20)
• ADC LSB: LSB = V_ref / 2ᴺ
• Uncertainty budget (RSS): u_c = √(u₁² + u₂² + u₃² + …)
• Anti-aliasing: attenuate signals above f_s/2 below 1 LSB

Worked example: Setting in-amp gain

Given: Desired G = 100; G = 1 + 49.4 kΩ/R_G

Find: R_G.

• R_G = 49.4k/(100 − 1) = 499 Ω (a standard E96 value)

Answer: 499 Ω

Worked example: Combined uncertainty

Given: Sensor ±0.5 %; Amplifier ±0.1 %; ADC ±0.05 % (independent, standard uncertainties)

Find: Combined uncertainty.

• u_c = √(0.5² + 0.1² + 0.05²) = √0.2625 = 0.51 %
• Expanded (k = 2): ±1.02 % at ~95 % confidence

Answer: 0.51 % (±1.0 % at 95 %) — dominated by the sensor

Common mistakes

• Wrong: Sampling without an anti-aliasing filter. — Right: Out-of-band noise folds into your band and can't be removed afterwards.
• Wrong: Quoting ADC bits as system accuracy. — Right: Build an error budget: noise, offset, gain, INL, reference, sensor and temperature drift.

Applications

• Test & measurement DAQ: Multi-channel data acquisition for vehicle testing, structural monitoring and research labs.
• Process automation: Field instruments feed PLC/DCS systems with calibrated, diagnosed measurements.
• Medical devices: ECG front-ends are extreme instrumentation: µV signals, huge common-mode interference, strict isolation.

Key takeaways

• Design the whole chain: sensor → in-amp → filter → ADC → processing → output.
• CMRR, anti-aliasing and ENOB often matter more than raw bits.
• Combine independent uncertainties by RSS; report with k = 2.
• Calibrate, isolate and synchronise for industrial-grade measurements.

Going further

• Auto-zero and chopper amplifiers: Chopper-stabilised amplifiers modulate the input to cancel offset and 1/f noise, achieving µV offsets and nV/°C drift — ideal for thermocouples and bridges.
• Isolation amplifiers: Capacitive, magnetic or optical isolation barriers let measurements cross thousands of volts safely, as with shunt measurements on HV DC links.
• Time synchronisation: Distributed DAQ systems align samples with IEEE 1588 PTP so phase and timing relationships between remote channels are preserved (sub-µs).