Sensors, Transducers and Measurement Systems

Introduction and context

Mechatronic systems depend on information about the physical world. A controller cannot regulate a motor, position an actuator, or protect a machine unless it receives measurements of relevant variables. Sensors therefore form the interface between mechanical processes and electronic computation. Their signals influence safety, product quality, energy use, productivity, and maintenance decisions.

A sensor detects a physical quantity, such as displacement, temperature, pressure, or force. A transducer converts that quantity into a usable signal, commonly an electrical voltage, current, resistance, frequency, or digital data word. A measurement system includes the sensor, signal conditioning, analogue-to-digital conversion, computation, communication, display, and the procedures used to establish confidence in the result.

The central engineering task is not simply to choose the most sensitive sensor. It is to select and integrate a device whose range, accuracy, dynamics, environmental compatibility, cost, and reliability match the application.

Learning objectives

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

1. Distinguish sensors, transducers, measurement systems, and instrumentation.
2. Select suitable sensors for position, speed, force, pressure, temperature, flow, and vision.
3. Explain sensitivity, resolution, accuracy, precision, hysteresis, linearity, repeatability, drift, and uncertainty.
4. Model static and dynamic sensor behaviour.
5. Design basic signal-conditioning and sampling arrangements.
6. Evaluate noise, calibration, loading, and environmental effects.
7. Interpret measurement data and justify engineering choices.

1. Measurement-system architecture

A typical measurement chain is:

Figure: A measurement system converts a physical measurand into information used for control or decision-making.

The measurand is the quantity intended to be measured. The sensor responds to it, but may also respond to interference variables such as ambient temperature, vibration, electromagnetic fields, or supply voltage. Signal conditioning may provide excitation, bridge completion, amplification, isolation, linearisation, filtering, and protection. The converter maps the conditioned analogue signal to a digital representation.

A measurement system is open loop when its output is merely observed. It is part of a closed-loop control system when the measured value is compared with a reference and used to compute corrective action. In closed-loop systems, delay, noise, quantisation, and incorrect calibration can reduce stability as well as measurement quality.

2. Sensor classifications and selection

Sensors may be classified in several useful ways:

• Active sensors generate energy, as in thermocouples and photovoltaic devices. Passive sensors require excitation, as in strain gauges, resistance temperature detectors, and potentiometers.
• Analogue sensors produce continuously varying signals; digital sensors produce discrete states, counts, or communicated data.
• Contact sensors physically touch the object; non-contact sensors use magnetic, optical, ultrasonic, capacitive, or radar principles.
• Absolute sensors report a value relative to a fixed reference, whereas incremental sensors report change or count transitions.

Selection begins with a specification: variable, range, permissible error, bandwidth, resolution, interface, environmental conditions, mechanical mounting, safety requirements, life, and cost. A sensor with a 0–10 V output is not automatically suitable for a controller whose input is 0–3.3 V; compatibility requires attention to scaling, grounding, isolation, and protection.

3. Static characteristics and measurement quality

For an ideal linear sensor, output can be represented by:

where x is the measurand, y is output, y₀ is the zero offset, and S is sensitivity, with units of output per unit input. For a nonlinear sensor, sensitivity is local:

Sensitivity must not be confused with resolution. Sensitivity describes output change per input change; resolution is the smallest input change that can be distinguished. A highly sensitive sensor may still have poor resolution if it is noisy.

Range is the interval over which specified performance is maintained. Span is the difference between upper and lower range limits. Accuracy describes closeness to an accepted reference value, often expressed as ±% of full scale or ±% of reading. Precision describes the concentration of repeated results and does not guarantee accuracy. Repeatability is agreement under the same conditions; reproducibility concerns changed conditions such as operator, instrument, or laboratory.

Linearity error is deviation from a specified straight-line relationship. Hysteresis is the difference in output for the same input when approached from increasing and decreasing directions. Dead band is an input interval producing no detectable output change. Drift is a time-dependent change in output for constant input; it may be caused by ageing, temperature, or contamination.

For independent standard uncertainties, a combined standard uncertainty is commonly estimated by the root-sum-square relationship:

Here u_i are standard uncertainties, expressed in the same units, and independence is assumed. If inputs are correlated, covariance terms are required. An expanded uncertainty is often reported as U = ku_c, where k is a chosen coverage factor; k ≈ 2 is frequently associated with approximately 95% coverage only when the statistical assumptions are justified.

For a digital converter with N bits and an input span V_FS, the ideal quantisation step is:

This is a resolution interval, not a complete accuracy specification. Reference error, gain error, noise, nonlinearity, and missing codes may be larger than q.

Worked example: uncertainty and resolution

A displacement system has independent standard uncertainties of 0.20 mm from calibration, 0.12 mm from repeatability, and 0.15 mm from temperature effects. Thus,

u_c = √(0.20² + 0.12² + 0.15²) = √0.0784 ≈ 0.28 mm.

Using k = 2, the expanded uncertainty is approximately ±0.56 mm. A 12-bit ADC measuring 0–10 V has q = 10/4096 ≈ 2.44 mV. If the sensor sensitivity is 20 mV/mm, one ADC count corresponds ideally to about 0.122 mm. The ADC resolution is therefore finer than the expanded measurement uncertainty and does not by itself establish system accuracy.

4. Dynamic response, noise, and sampling

Static specifications are insufficient when a variable changes rapidly. A first-order sensor is often approximated by:

where τ is the time constant, K is static gain, x(t) is input, and y(t) is output. For a step input from 0 to X₀, with zero initial output:

At t = τ, output reaches 63.2% of its final value; after approximately 4τ it reaches about 98.2%. The model assumes linearity, constant parameters, and negligible higher-order dynamics.

A second-order system may exhibit overshoot and resonance. In practical measurement systems, bandwidth should be high enough to preserve the required signal but not so high that unnecessary noise is admitted. A low-pass filter reduces high-frequency noise but introduces attenuation and phase delay. In control applications, this delay can be more harmful than modest random noise.

Sampling frequency must satisfy the signal-content requirement. The Nyquist condition for a band-limited signal is:

where f_s is sampling frequency and f_max is the highest frequency present. In practice, sampling above the theoretical minimum and using an anti-aliasing filter are necessary. Aliasing is irreversible after sampling: a high-frequency component appears as a false lower-frequency component.

Noise may be thermal, shot, electromagnetic, quantisation, mechanical, or caused by power supplies and grounding. Good practice includes twisted or shielded cables, differential measurement, single-point grounding where appropriate, stable excitation, physical separation from power wiring, filtering, and verification with an oscilloscope or spectrum analyser. Filtering should never conceal a safety-critical fault without an independent diagnostic strategy.

5. Sensor technologies for common measurands

Position and displacement

Potentiometers are inexpensive and simple, but contact wear, friction, limited life, and loading affect performance. Linear variable differential transformers provide contactless displacement measurement with good resolution and robustness, but require excitation and signal conditioning. Optical encoders use coded discs and photodetectors. Incremental encoders provide pulses; quadrature channels A and B allow direction detection and commonly provide four transition counts per cycle. Absolute encoders provide a coded position after power interruption, although their interface and cost are greater.

Inductive and capacitive proximity sensors detect changes in electromagnetic or electric fields. Inductive sensors are particularly effective for metallic targets and are tolerant of dirt; capacitive sensors can detect non-metallic materials but are more sensitive to humidity and target dielectric properties. Ultrasonic sensors measure echo time and are useful where optical visibility is poor, but temperature and surface angle affect propagation.

Speed

Rotational speed can be obtained by differentiating position, counting encoder pulses, measuring pulse period, or using a tachogenerator. Pulse counting over a gate interval gives average frequency. If an encoder produces P pulses per revolution and C pulses are counted in time T, speed in revolutions per minute is:

The formula assumes each counted event corresponds to one pulse and that the speed is sufficiently constant during the gate interval. Period measurement gives better low-speed resolution but is more sensitive to timing jitter.

Force and torque

Strain gauges change resistance when strained. A Wheatstone bridge converts small resistance changes into a differential voltage. Load cells normally combine elastic elements, strain gauges, temperature compensation, and calibration. Piezoelectric force sensors have excellent dynamic response but cannot maintain a static force measurement indefinitely because charge leakage causes drift. Torque sensors may measure shaft strain directly or infer torque from motor current; the latter is convenient but depends on motor and friction models.

Pressure

Piezoresistive silicon, thin-film strain-gauge, capacitive, and piezoelectric pressure sensors are common. Pressure must be specified as absolute, gauge, or differential. A gauge sensor measures relative to local atmospheric pressure; an absolute sensor references a sealed vacuum. Installation requires attention to pressure spikes, pulsation, fluid compatibility, temperature, and blocked impulse lines.

Temperature

Resistance temperature detectors, especially platinum RTDs, offer good stability and accuracy. Thermistors provide high sensitivity over a limited range and are widely used in compact systems. Thermocouples operate over broad temperature ranges and are rugged, but their small output and dependence on reference-junction compensation require careful wiring. Infrared sensors measure emitted radiation without contact, but emissivity, reflections, distance, and field of view influence the result.

Flow

Differential-pressure flowmeters infer flow from pressure loss across a restriction; they require density information and often have a nonlinear relationship. Electromagnetic flowmeters work with conductive liquids and have no obstruction, but cannot measure insulating fluids. Vortex meters detect shedding frequency and suit many clean fluids. Coriolis meters directly measure mass flow with high performance but are relatively expensive. Ultrasonic flowmeters are non-invasive in some installations, yet depend on acoustic coupling and flow profile.

Machine vision

A vision system normally includes illumination, lens, image sensor, processing hardware, and algorithms. The camera converts light into pixel values; calibration maps pixel coordinates to physical coordinates. Common tasks include presence detection, dimensional inspection, alignment, barcode reading, and surface-defect classification. Resolution is determined not only by pixel count but also by optics, field of view, contrast, focus, motion blur, and lighting. A controlled lighting arrangement often improves reliability more than a higher-resolution camera.

6. Calibration and traceability

Calibration compares an instrument with a reference of known or stated uncertainty. It does not automatically repair an instrument. A calibration procedure should define the measurand, reference standard, environmental conditions, test points, increasing and decreasing sequences where hysteresis matters, acceptance criteria, and treatment of uncertainty.

A linear calibration can be estimated from paired reference and instrument readings. For a simple two-point correction, if output readings are y₁ and y₂ at reference inputs x₁ and x₂, the corrected input estimate for an observed output y is:

This interpolation assumes stable, approximately linear behaviour between the two points. Multi-point calibration may use least-squares fitting, lookup tables, or polynomial compensation, but increasing mathematical complexity does not remove poor reference data or environmental variation.

Traceability means that a measurement result can be related to a recognized reference through an unbroken documented chain of calibrations, each contributing to uncertainty. Laboratories and industries may follow quality systems such as ISO/IEC 17025, while functional safety applications may require additional sector-specific standards.

7. Worked application: conveyor inspection and sorting

Consider an illustrative packaging line that must detect package presence, measure spacing, reject overweight packages, and identify an incorrect label.

Background. Packages move at up to 1.2 m/s. The control system must reject a package within 50 mm of its target location. Package mass ranges from 0.2 to 5 kg.

Problem. A poor sensor choice could cause missed detections, inaccurate weighing, or false rejection under changing lighting and vibration.

Analysis. A photoelectric sensor provides fast presence detection, but transparent packaging may require a retro-reflective or through-beam arrangement. An encoder attached to the conveyor provides travel feedback. If the encoder produces 2,000 counts per revolution and the measuring wheel circumference is 0.5 m, the distance per count is 0.5/2000 = 0.25 mm. This is substantially finer than the 50 mm positioning requirement, although wheel slip must be checked.

A load cell beneath a mechanically isolated section measures mass. The weighing interval must be long enough for vibration to settle; digital averaging may improve repeatability but creates delay. A camera with diffuse, controlled illumination inspects the label. Triggering the camera from the encoder rather than from a fixed timer compensates for moderate speed variation.

Approach and solution. The system uses a through-beam photoelectric sensor for entry, encoder counts for position, a calibrated load cell with overload protection, and a camera whose region of interest is fixed in software. Sensor signals are shielded and separated from motor cables. The controller records raw values, corrected values, alarm states, and calibration status. A plausibility check rejects impossible combinations, such as a mass reading when no package is present.

Outcome. The integrated design improves repeatability because each sensor is assigned a measurand it can observe reliably. The example also demonstrates that sensor selection is a system problem: mechanical layout, lighting, timing, calibration, and software diagnostics are as important as the sensing principle.

Learning points. Use an encoder for motion-referenced timing, isolate force sensors from vibration, control illumination in vision systems, and validate the complete measurement chain under worst-case operating conditions.

8. Limitations, ethics, and contemporary developments

Every sensor has failure modes. A sensor may saturate, become disconnected, drift, be incorrectly installed, or report a plausible but false value. Redundant sensors, range checks, watchdogs, diagnostic excitation, and fail-safe states are appropriate where failure can injure people or damage equipment. Safety-related measurements require formal hazard analysis rather than reliance on nominal accuracy alone.

Measurement data can also affect workers, customers, and automated decisions. Vision systems may collect identifiable images; data should be limited to a legitimate purpose, secured, retained only as necessary, and governed according to applicable privacy and workplace rules. Calibration records and software changes should be auditable when measurements support compliance or product release.

Modern systems increasingly combine smart sensors, industrial Ethernet, edge computing, and machine-learning algorithms. AI can classify defects or estimate variables from multiple sensors, but it does not eliminate the need for traceable calibration, representative validation data, uncertainty analysis, and human responsibility. A learned model may fail when lighting, materials, or operating conditions differ from training data. Physics-based plausibility checks and confidence monitoring remain valuable.

Best practice is to: define the measurand precisely; select range before nominal accuracy; analyse dynamics and sampling; design grounding and mechanical mounting; calibrate the assembled chain; document uncertainty; test environmental extremes; provide diagnostics; and review the measurement against the decision it supports.

Chapter Summary

Sensors convert physical phenomena into information used by mechatronic systems. A complete measurement system includes sensing, excitation, conditioning, conversion, computation, and application. Sensor choice must consider measurand, range, sensitivity, resolution, accuracy, precision, hysteresis, drift, environmental effects, bandwidth, interface, safety, and cost. First-order dynamic models clarify time constant and response delay, while sampling theory explains aliasing and the need for anti-alias filtering. Position, speed, force, pressure, temperature, flow, and vision require different sensing principles and installation practices. Calibration establishes the relationship to a reference; uncertainty quantifies confidence but does not guarantee fitness for purpose. Reliable engineering integrates physical design, electronics, software, diagnostics, traceability, and ethical governance.

Key Terms

Sensor: Device that responds to a physical quantity.

Transducer: Device that converts one form of energy or quantity into another usable form.

Measurand: Quantity intended to be measured.

Sensitivity: Change in output per unit change in input.

Resolution: Smallest distinguishable change in the measurand.

Accuracy: Closeness of a result to an accepted reference value.

Precision: Degree of agreement among repeated measurements.

Hysteresis: Different output for the same input depending on input direction.

Drift: Time-dependent change in output for constant input.

Linearity: Degree to which input-output behaviour follows a specified straight line.

Calibration: Comparison with a reference under defined conditions.

Traceability: Documented connection of a result to a recognized reference.

Uncertainty: Quantified doubt associated with a measurement result.

Time constant: Parameter describing the speed of a first-order response.

Aliasing: False low-frequency representation caused by insufficient sampling.

Strain gauge: Resistive element whose resistance changes with strain.

Encoder: Sensor producing position-related pulses or coded data.

Load cell: Force transducer, commonly based on strain gauges.

Measuring range: Input interval over which specified performance applies.

Signal conditioning: Filtering, amplification, isolation, excitation, and related processing.

Review Questions

Short Answer Questions

1. Distinguish a sensor from a transducer.
2. Define sensitivity and resolution.
3. What is hysteresis, and why can it matter in positioning?
4. State two reasons for using signal conditioning.
5. Why is calibration different from repair?

Long Answer Questions

1. Explain the architecture of a complete measurement system and the function of each stage.
2. Compare incremental and absolute encoders for a servo-positioning application.
3. Discuss accuracy, precision, repeatability, reproducibility, and uncertainty.
4. Explain first-order dynamic response and the practical meaning of the time constant.
5. Compare RTDs, thermistors, thermocouples, and infrared temperature sensors.

Analytical / Application Questions

1. An encoder produces 1,500 pulses per revolution and is connected to a wheel of circumference 0.6 m. Derive the distance represented by one pulse and identify practical sources of error.
2. Design a sensor set for a conveyor that detects packages, measures mass, determines travel position, and verifies labels. Justify each technology and identify likely failure modes.
3. A pressure signal contains useful components below 40 Hz and interference near 300 Hz. Propose a sampling and filtering strategy, explaining aliasing, delay, and validation requirements.

Multiple Choice Questions with Answers

Question 1

Which quantity is the change in sensor output per unit change in input?

A. Sensitivity

B. Hysteresis

C. Resolution

D. Drift

Correct Answer: A. Sensitivity

Question 2

A 12-bit ADC measures a 0–10 V span. What is its ideal quantisation step?

A. 0.244 mV

B. 2.44 mV

C. 24.4 mV

D. 244 mV

Correct Answer: B. 2.44 mV

Question 3

At one time constant, an ideal first-order step response reaches approximately what percentage of its final value?

A. 37%

B. 50%

C. 63.2%

D. 99%

Correct Answer: C. 63.2%

Question 4

Which sensor is especially suitable for detecting metallic targets without contact?

A. Inductive proximity sensor

B. Thermocouple

C. Coriolis meter

D. Piezoelectric accelerometer

Correct Answer: A. Inductive proximity sensor

Question 5

An encoder has 2,000 pulses per revolution and a wheel circumference of 0.5 m. What distance does one pulse represent?

A. 0.025 mm

B. 0.25 mm

C. 2.5 mm

D. 25 mm

Correct Answer: B. 0.25 mm

Question 6

What is the principal danger of sampling a signal below twice its highest frequency component?

A. Hysteresis

B. Aliasing

C. Thermal drift

D. Saturation

Correct Answer: B. Aliasing

Question 7

Which device is commonly used to measure force through elastic deformation and resistance change?

A. Strain-gauge load cell

B. Ultrasonic range finder

C. Thermistor

D. Vortex meter

Correct Answer: A. Strain-gauge load cell

Question 8

Which statement about accuracy and precision is correct?

A. They are identical concepts.

B. Precision guarantees closeness to the reference.

C. Accuracy concerns closeness to a reference, while precision concerns repeatability.

D. Accuracy applies only to digital sensors.

Correct Answer: C. Accuracy concerns closeness to a reference, while precision concerns repeatability.

Question 9

For independent standard uncertainties, which relationship is normally used for combined standard uncertainty?

A. Arithmetic sum only

B. Root-sum-square

C. Largest value squared

D. Difference between maximum and minimum

Correct Answer: B. Root-sum-square

Question 10

In a machine-vision inspection system, which intervention often improves reliability more directly than simply increasing pixel count?

A. Uncontrolled ambient lighting

B. Controlled illumination

C. Removing calibration

D. Increasing cable length

Correct Answer: B. Controlled illumination