Measurement, Instrumentation, and Data Acquisition

Introduction and Context

Measurement is the process of obtaining information about a physical quantity by comparison with an agreed reference. Instrumentation is the organized use of sensors, signal-conditioning circuits, data-acquisition hardware, software, and procedures to make that measurement useful. In mechatronics, measurement connects the physical plant to computation and control: a robot estimates position from encoders, a motor drive uses current and speed feedback, and a production system detects dimensional error through vision or displacement sensors.

A measurement system is not merely a sensor. It is a chain whose performance depends on every stage:

Figure: A measurement system transforms a physical measurand into evidence for engineering action; validation and feedback can improve the complete chain.

Reliable data must be both accurate enough for its purpose and supported by evidence about how it was produced. A highly precise sensor can still give invalid results if it is poorly calibrated, incorrectly installed, undersampled, or affected by an unrecognized disturbance.

Learning Objectives

By the end of this chapter, a student should be able to:

1. Explain the roles and operating principles of common mechatronic sensors.
2. Select sensors and signal-conditioning circuits for a specified measurand and performance requirement.
3. Design a sampling and data-acquisition arrangement while avoiding aliasing and overload.
4. Distinguish accuracy, precision, resolution, sensitivity, hysteresis, drift, and uncertainty.
5. Calibrate a measurement channel and report results with appropriate units and uncertainty.
6. Apply traceability, validation, documentation, and ethical data-governance practices.

Measurement Concepts and Terminology

The measurand is the quantity intended to be measured, such as shaft speed, temperature, force, or displacement. The indication is the value shown or recorded by an instrument. The measurement result is the indication interpreted using a measurement model, corrections, and an uncertainty statement.

Important performance characteristics include:

A sensor converts one form of energy or physical state into another. A transducer may include the sensing element and conversion mechanism; a transmitter generally produces a standardized signal suitable for transmission. Sensors may be passive, requiring excitation, or active, generating an output from the measurand itself.

Static Characteristics

For a linear sensor model,

where x is the measurand in its physical unit, y is the output, a is the offset, and b is sensitivity with output-unit/input-unit. This model is applicable over the calibrated operating interval; it should not be extrapolated beyond that interval without evidence.

Sensitivity is

where ∆y and ∆x are corresponding output and input changes. For a nonlinear sensor, sensitivity may depend on operating point and is represented locally by the slope of the calibration curve.

The normalized full-scale error is often reported as

This convention is useful only when the reference, span, and sign convention are explicitly stated. Other conventions, such as percent of reading, are not interchangeable.

Sensor Technologies for Mechatronics

Resistive sensors include strain gauges, resistance temperature detectors, thermistors, and potentiometers. A strain gauge changes resistance with deformation and is commonly placed in a Wheatstone bridge. Bridges reject some common-mode effects and permit small resistance changes to be converted into differential voltage, but they require stable excitation and careful wiring.

Thermal sensors include thermocouples, RTDs, and thermistors. Thermocouples are rugged and have a wide range but produce small voltages and require reference-junction compensation. RTDs provide good stability and approximate linearity over a limited interval. Thermistors are sensitive and inexpensive but strongly nonlinear.

Inductive and magnetic sensors include linear variable differential transformers, variable-reluctance pickups, Hall-effect devices, and resolvers. They are useful in harsh environments, although magnetic interference, target geometry, and excitation frequency may affect performance.

Capacitive sensors detect changes in capacitance caused by displacement, proximity, level, or humidity. They can provide high resolution but are sensitive to stray capacitance, contamination, and grounding.

Optical sensors include photodiodes, reflective sensors, optical encoders, and machine-vision systems. Incremental encoders measure motion through pulses; quadrature channels provide direction. Absolute encoders provide a coded position without requiring a homing operation, but their data interface and latency must be considered.

Piezoelectric sensors are especially useful for dynamic force, pressure, and vibration because they have high stiffness and bandwidth. They are not suitable for indefinite static measurement without appropriate charge or voltage-conditioning methods.

MEMS inertial sensors measure acceleration and angular rate. Bias instability, scale-factor error, cross-axis sensitivity, temperature effects, and vibration rectification must be considered in navigation and control applications.

Sensor selection should begin with the requirement, not the available component. Specify measurand, range, bandwidth, environmental conditions, permissible loading, output interface, power, safety, cost, calibration availability, and required uncertainty.

Signal Conditioning

Signal conditioning makes a sensor output compatible with the DAQ and improves its immunity to interference. Typical functions are amplification, attenuation, buffering, bridge excitation, isolation, linearization, demodulation, filtering, cold-junction compensation, and excitation monitoring.

A differential amplifier responds primarily to the voltage difference between two input terminals. Its common-mode rejection ratio is

where A_d is differential gain and A_cm is common-mode gain. The equation assumes gains are expressed as voltage ratios. High CMRR is valuable when a small sensor signal is superimposed on a larger common-mode voltage, but wiring, resistor mismatch, input range, and frequency can reduce practical performance.

An instrumentation amplifier is often used for bridge and low-level signals because it provides high input impedance, adjustable gain, and good common-mode rejection. The gain must be chosen so the largest expected signal, including fault and transient conditions, remains within the amplifier and ADC input range.

Low-pass filtering attenuates high-frequency content. For a first-order RC low-pass filter,

where R is resistance in ohms, C is capacitance in farads, and f_c is the −3 dB cutoff frequency in hertz. This ideal relation assumes a sufficiently high load impedance and negligible source impedance; real circuits require consideration of loading and component tolerances. An analog anti-alias filter should precede the ADC.

Ground loops, electromagnetic interference, shield currents, thermoelectric voltages, inadequate decoupling, and cable motion can create errors larger than the sensor’s nominal resolution. Good practice includes twisted differential pairs, appropriate shielding, star or planned grounding, physical separation from power wiring, filtering matched to the signal bandwidth, and isolation where safety or ground potential requires it.

Sampling, ADCs, and DAQ Platforms

A DAQ system normally contains sensors, conditioning, multiplexers or simultaneous channels, sample-and-hold circuitry, an analog-to-digital converter (ADC), timing hardware, communication, storage, and analysis software. DAQ platforms may be embedded microcontrollers, programmable automation controllers, USB instruments, industrial Ethernet modules, or real-time research systems.

The sampling theorem states that a band-limited signal with highest frequency f_max can be reconstructed in principle when

where f_s is sampling frequency. This is a theoretical minimum for ideal conditions. Engineering systems normally sample substantially faster because real filters have finite transition bands, timing is imperfect, and the signal may contain disturbances above the intended band. Frequencies above half the sampling frequency can appear as lower, false frequencies: this is aliasing. Once aliased, digital processing cannot reliably recover the original frequency.

ADC resolution is described by the number of bits N. For an ideal unipolar ADC with input span V_FS, the nominal code width is approximately

where q is in volts per code. Quantization error under the usual rounding model is bounded approximately by ±q/2, but total accuracy also includes reference error, integral and differential nonlinearity, noise, gain error, offset, and timing effects. A 16-bit ADC is not automatically a 16-bit accurate measurement instrument.

Sampling frequency, channel count, conversion time, buffering, timestamp quality, data type, storage rate, and synchronization must be specified. For multi-channel systems, clarify whether channels are simultaneous or sequentially multiplexed. Rapidly changing systems may require simultaneous sampling or compensation for channel-to-channel time skew.

A robust acquisition workflow is:

1. Define the measurand and decision or control requirement.
2. Establish range, bandwidth, environment, and allowable uncertainty.
3. Select the sensor and interface.
4. Design excitation, conditioning, grounding, isolation, and anti-alias filtering.
5. Select ADC resolution, sampling rate, timing, and storage.
6. Calibrate the complete channel where practical.
7. Verify performance using independent reference data.
8. Record metadata, raw data, corrections, and processing steps.

Calibration, Traceability, and Validation

Calibration compares an instrument or measurement system with a reference under specified conditions and establishes a relation between indications and values. Calibration does not automatically repair an instrument, certify its fitness for every use, or guarantee accuracy between calibration points.

A two-point linear calibration can be written as

where x̂ is the estimated measurand, y is the observed output, and (x_0, y_0) and (x_1, y_1) are paired calibration points. This model assumes stable, approximately linear behavior between the points and consistent measurement conditions.

Traceability is an unbroken documented chain of calibrations, each contributing to measurement uncertainty, linking a result to a recognized reference. Traceability requires stated procedures, competent personnel, suitable equipment, records, and uncertainty evaluation. A manufacturer’s certificate may support traceability, but its scope and conditions must match the intended use.

Calibration data should include instrument identification, reference standards, environmental conditions, applied values, indications, corrections, uncertainty, date, operator, and acceptance criteria. Verification asks whether a system meets a stated requirement; validation asks whether the system is fit for its intended application. Both are essential in research because a technically functioning instrument can still produce data inadequate for the research question.

Uncertainty Analysis

Error is the difference between a measured value and a reference or true value, although the true value is generally not known exactly. Uncertainty is a quantified expression of doubt about a measurement result. It should not be confused with an error that can be corrected.

For independent input quantities x_i combined through a measurement model y = f(x_1, x_2, …), the standard uncertainty propagation approximation is

where u_c(y) is combined standard uncertainty, ∂f/∂x_i is the sensitivity coefficient, and u(x_i) is the standard uncertainty of input x_i. The approximation assumes small uncertainties, a sufficiently smooth model, and independent inputs. If inputs are correlated, covariance terms must be added:

where u(x_i, x_j) is covariance. In practice, uncertainty components may arise from reference calibration, repeatability, resolution, drift, temperature, positioning, and data processing.

For repeated observations x₁, x₂, …, x_n, the sample mean is

and the sample standard deviation is

The standard uncertainty of the mean, when observations are appropriately independent and representative, is s ÷ √n. Repeating measurements reduces random variation in the mean but does not remove systematic bias, drift, or an incorrectly specified model.

An expanded uncertainty is commonly reported as

where k is a coverage factor selected according to the desired coverage and statistical assumptions. A report should state the result as, for example, x = 12.40 V ± 0.08 V, with k and the coverage convention identified. Do not report more digits than the uncertainty supports.

Worked Example: Force Measurement

A load cell has sensitivity 2.00 mV/V, excitation 5.00 V, and amplifier gain 100. The measured output above zero is 0.500 V. Assuming zero offset has already been removed and the load cell response is linear, the load-cell bridge output is 0.500 ÷ 100 = 0.00500 V = 5.00 mV. The corresponding force is

The denominator is 10.0 mV, so the reading is 0.500 F_FS. If the cell full-scale rating is 1,000 N, the estimated force is 500 N. This result is valid only under the stated excitation, calibration, linearity, temperature, and zero-offset assumptions. It is not a complete uncertainty statement.

Applied Case: Instrumenting a Servo-Driven Linear Axis

Background. An illustrative manufacturing cell uses a ball-screw axis to place a tool within ±0.10 mm. The controller has an incremental encoder, while an independent displacement measurement is required for research validation.

Problem. Initial tests show occasional position errors and inconsistent results between repeated trials. The team must determine whether the cause is mechanical, control-related, or measurement-related.

Analysis. The expected motion bandwidth is below 20 Hz, but encoder edges and motor switching generate higher-frequency disturbances. A laser displacement sensor is selected with a 100 mm range and a specified uncertainty smaller than one-third of the allowable positioning error. The sensor is mounted to avoid observing frame vibration as if it were carriage motion. Its analog output is conditioned differentially and low-pass filtered. The cutoff is selected above the required motion band but below frequencies likely to alias.

The DAQ samples at 1 kHz, provides synchronized timestamps, and records the encoder count, commanded position, motor current, laser output, temperature, and system state. A static calibration is performed at several positions using a traceable gauge reference. Forward and reverse approaches are recorded to reveal hysteresis. A separate dynamic check compares the laser and encoder during controlled moves.

Approach and solution. The research team first preserves raw data and records calibration coefficients rather than overwriting the original sensor signal. They identify a loose sensor bracket from a frequency peak that appears in the laser signal but not in the encoder. After mechanically correcting the bracket, they discover a second issue: an incorrect encoder scale factor. The scale factor is corrected through documented software configuration, and the complete system is retested.

Outcome. Static errors meet the ±0.10 mm requirement, and the uncertainty budget identifies bracket deflection and reference positioning as dominant components. Dynamic tests show that the axis itself has a resonance near the upper part of the intended bandwidth, so the controller is retuned rather than hiding the feature with excessive filtering.

Learning points. Independent validation is essential; sensor installation can dominate sensor specifications; raw data and metadata preserve research credibility; filtering cannot correct mechanical error; and a measurement system should be designed around the decision it must support.

Ethics, Safety, and Data Governance

Measurement results influence safety decisions, product acceptance, scientific claims, and public resources. Researchers must not discard inconvenient observations without a documented rule, selectively report only successful trials, alter calibration records, or present processed data as raw data. Automated outlier rejection requires a stated criterion and retention of the original observation.

Electrical isolation, safe excitation, sensor mounting, pressure containment, laser classification, rotating machinery guards, and fail-safe behavior are engineering responsibilities. Measurement equipment connected to a live industrial system can create hazards through ground faults or unintended control commands.

Data files should include time base, units, channel names, calibration version, sensor serial numbers, sampling settings, environmental conditions, operator, and processing history. Access control and integrity checks are appropriate when data support regulated decisions. Relevant quality systems may include ISO/IEC 17025 principles for calibration and testing competence, while legal and sector-specific requirements must be checked for the application.

Contemporary Developments and Best Practices

Smart sensors increasingly perform local linearization, diagnostics, timestamping, and digital communication. Industrial networks can distribute synchronized measurements, but network latency, packet loss, clock synchronization, and cybersecurity must be characterized. Edge computing can reduce bandwidth and latency, yet it should not eliminate access to auditable raw or minimally processed data.

Machine learning can support anomaly detection, sensor fusion, and virtual sensing. It does not remove the need for calibration, physical plausibility checks, uncertainty evaluation, or representative validation data. A model trained on biased or drifting measurements may produce confident but invalid conclusions. AI-assisted processing should therefore preserve provenance, version models, expose decision criteria where possible, and maintain human responsibility for safety-critical judgments.

A practical best-practice checklist is:

• define the measurand, required range, bandwidth, and uncertainty before selecting hardware;
• calibrate the complete measurement path where possible;
• use anti-alias filtering and document sampling assumptions;
• separate random variation, bias, drift, and instrument resolution;
• preserve raw data, metadata, calibration records, and software versions;
• verify installation effects, environmental sensitivity, and loading;
• synchronize clocks and document latency in multi-channel systems;
• report units, significant digits, corrections, uncertainty, and coverage;
• validate against an independent reference or physical conservation law;
• review safety, cybersecurity, privacy, and ethical consequences.

Chapter Summary

Measurement systems in mechatronics transform physical quantities into evidence for control, research, and engineering decisions. Their reliability depends on the entire chain: measurand definition, sensor selection, mounting, excitation, signal conditioning, filtering, ADC conversion, timing, software, calibration, and interpretation. Static characteristics such as sensitivity, linearity, hysteresis, resolution, repeatability, and drift describe performance but must be related to the intended application.

Sampling must satisfy bandwidth requirements and prevent aliasing through appropriate analog filtering. ADC resolution describes code spacing, not total accuracy. Calibration establishes a relation to a reference, while traceability documents the connection to recognized standards. Uncertainty analysis combines relevant input contributions and distinguishes uncertainty from error. Validation demonstrates fitness for purpose using independent evidence.

Professional practice requires safe design, honest reporting, preserved raw data, complete metadata, controlled processing, and awareness of environmental, ethical, regulatory, and cybersecurity issues. Modern digital and AI-enabled instruments provide powerful capabilities, but they do not replace physical reasoning, calibration, or critical evaluation.

Key Terms

Measurand: Quantity intended to be measured.

Sensor: Device that responds to a physical quantity.

Transducer: Device or assembly that converts a physical quantity into a usable signal.

Sensitivity: Output change per unit input change.

Resolution: Smallest distinguishable change in indication.

Accuracy: Closeness of a result to a reference value.

Precision: Closeness among repeated results.

Hysteresis: Dependence of output on the direction of input change.

Drift: Time-dependent change under constant input.

Signal conditioning: Amplification, filtering, isolation, excitation, or conversion applied to a sensor signal.

Common-mode rejection ratio: Measure of rejection of equal signals applied to differential inputs.

Anti-alias filter: Analog filter used to limit frequency content before sampling.

Sampling frequency: Number of samples acquired per second.

Quantization: Representation of a continuous amplitude by discrete digital codes.

Calibration: Comparison with a reference to establish a measurement relationship.

Traceability: Documented chain linking a result to a recognized reference.

Uncertainty: Quantified doubt associated with a measurement result.

Validation: Demonstration that a system is fit for its intended purpose.

Data provenance: Record of data origin, transformations, and processing history.

Review Questions

Short Answer Questions

1. Distinguish accuracy from precision.
2. What is the purpose of an anti-alias filter?
3. Define measurement traceability.
4. Why is a sensor’s installation part of the measurement system?
5. What information should accompany a calibrated measurement result?

Long Answer Questions

1. Explain the complete signal chain from measurand to engineering decision, identifying likely sources of error at each stage.
2. Compare resistive, optical, magnetic, capacitive, and piezoelectric sensors for mechatronic applications.
3. Explain how sampling frequency, bandwidth, filtering, and ADC resolution interact in a DAQ design.
4. Describe a calibration procedure for a nonlinear displacement sensor, including reference selection and documentation.
5. Explain the difference between error, uncertainty, repeatability, reproducibility, and validation.

Analytical / Application Questions

1. A vibration signal contains useful components up to 400 Hz. Propose a sampling and anti-aliasing strategy, and justify why sampling exactly at 800 Hz would be unsuitable in practice.
2. Design a measurement channel for a 0–10 N force range with a required uncertainty below 0.05 N. Specify sensor, conditioning, ADC, calibration, and validation considerations.
3. A temperature channel gives stable but biased readings after a long warm-up. Analyze possible causes and propose tests that distinguish offset, drift, hysteresis, and environmental effects.

Multiple Choice Questions with Answers

Question 1

Which quantity is the intended physical quantity being measured?

A. Measurand

B. Resolution

C. Indication

D. Quantization

Correct Answer: A. Measurand

Question 2

A first-order RC low-pass filter has R = 1 kΩ and C = 1 µF. Its approximate cutoff frequency is:

A. 15.9 Hz

B. 159 Hz

C. 1.59 kHz

D. 15.9 kHz

Correct Answer: A. 15.9 Hz

Question 3

What is the primary purpose of an analog anti-alias filter?

A. Increase ADC bit depth

B. Limit frequency content before sampling

C. Remove all sensor offset

D. Synchronize independent clocks

Correct Answer: B. Limit frequency content before sampling

Question 4

For a signal band-limited to 400 Hz, which sampling frequency satisfies the theoretical Nyquist condition?

A. 400 Hz

B. 600 Hz

C. 800 Hz exactly

D. 1,000 Hz

Correct Answer: D. 1,000 Hz

Question 5

A 12-bit ideal unipolar ADC with a 4.096 V span has an approximate code width of:

A. 0.1 mV

B. 1 mV

C. 10 mV

D. 100 mV

Correct Answer: B. 1 mV

Question 6

Which characteristic describes closeness among repeated measurements?

A. Accuracy

B. Precision

C. Sensitivity

D. Span

Correct Answer: B. Precision

Question 7

Traceability requires:

A. An unbroken documented calibration chain to a recognized reference

B. A sensor with the highest available resolution

C. Repeated measurements without a reference

D. A manufacturer’s name printed on the instrument

Correct Answer: A. An unbroken documented calibration chain to a recognized reference

Question 8

In uncertainty propagation, covariance terms are especially needed when input quantities are:

A. Exactly zero

B. Correlated

C. Measured in volts

D. Recorded digitally

Correct Answer: B. Correlated

Question 9

Why can repeated measurements fail to remove a systematic bias?

A. Averaging increases the bias

B. Bias is not random variation that averages away

C. Repeated measurements reduce resolution

D. Sampling always causes bias

Correct Answer: B. Bias is not random variation that averages away

Question 10

Which practice best supports reproducible research measurements?

A. Overwrite raw data with corrected values

B. Report only successful trials

C. Preserve raw data, metadata, calibration records, and processing history

D. Change filtering settings between trials without recording them

Correct Answer: C. Preserve raw data, metadata, calibration records, and processing history