Modern mechatronic systems depend on reliable information about physical variables. A robot must know the position of its joints, a packaging machine must detect objects, a hydraulic press must measure pressure, and a thermal process must maintain temperature within safe limits. Sensors and transducers provide this information by converting a physical quantity into a usable signal.

Measurement is not simply the act of obtaining a number. It is a systematic process involving a measurand, a sensing element, signal conversion, conditioning, transmission, interpretation, calibration, and assessment of uncertainty. A technically suitable sensor may still produce poor control performance if it is incorrectly installed, inadequately shielded, badly calibrated, or connected to an incompatible interface.

This chapter develops the principles needed to select and apply industrial sensors for position, speed, force, pressure, flow, temperature, proximity, and machine vision applications.

Learning Objectives

1. Measurement Concepts and Terminology

The measurand is the quantity intended to be measured, such as shaft speed, temperature, or pressure. The sensor is the element that responds directly to the measurand. A transducer converts energy or information from one domain into another; in instrumentation it commonly converts a physical variable into an electrical signal. A transmitter usually incorporates sensing and conditioning electronics and produces a standardized output such as 4–20 mA.

physical variable → sensing element → transduction → signal conditioning → transmission → acquisition/control → decision or actuation

A sensor may be active or passive. An active sensor generates an electrical output without an external excitation, as in a thermocouple or piezoelectric element. A passive sensor requires excitation; examples include resistance temperature detectors, strain gauges, potentiometers, and inductive displacement sensors.

1.1 Measurement error

If the measured value is xₘ and the reference or accepted value is xᵣ, the measurement error is:

Percentage error is 100eᵣ percent, provided xᵣ is not zero. In practice, the reference value itself has uncertainty, so error should not be interpreted as perfectly known.

Accuracy describes closeness to an accepted value, whereas precision describes the agreement among repeated measurements. A sensor can be precise but inaccurate if it has a stable calibration bias.

Resolution is the smallest discernible change. Sensitivity is the change in output per unit input:

where Δy is output change and Δx is input change. For a linear sensor, y = a + bx, b is sensitivity and a is the zero offset.

Range is the interval between lower and upper measurable limits. Span is upper range value minus lower range value. Linearity indicates how closely the actual calibration curve follows a specified straight line. Hysteresis is the difference in output at the same input depending on whether the input is increasing or decreasing. Repeatability concerns agreement under identical conditions; reproducibility concerns agreement when conditions such as operator, instrument, or location change.

2. Static and Dynamic Characteristics

Static characteristics describe behaviour when the input is constant or changes slowly. Important characteristics include sensitivity, offset, linearity, hysteresis, dead band, drift, resolution, and repeatability.

Drift is a gradual change in output unrelated to a change in measurand. Zero drift changes the intercept of a calibration curve; sensitivity drift changes its slope. Drift may result from temperature, ageing, mechanical stress, contamination, or electronic component changes.

Dynamic behaviour matters when a sensor measures a changing variable. A common first-order model is:

where x is input, y is output, K is static gain, and τ is the time constant. For a step input x = X, with zero initial output, the response is:

After one time constant the response reaches approximately 63.2% of its final value; after about five time constants it is approximately 99.3% complete. This model assumes linear behaviour, constant parameters, and negligible transport delay. A fast control loop requires a sensor whose bandwidth is adequate for the process dynamics.

A second-order sensor may be described by its natural frequency and damping ratio. Such a model is useful for accelerometers and mechanical sensing systems, but resonance can produce overshoot or amplification. Sensor specifications should therefore be compared with the frequency content of the application, not only with the maximum static range.

3. Sensor Technologies for Industrial Variables

3.1 Position and displacement

Position sensors measure linear or angular location. Common technologies include potentiometers, linear variable differential transformers (LVDTs), encoders, resolvers, magnetic sensors, and laser displacement sensors.

A potentiometer provides a voltage related to wiper position. For an ideal unloaded linear potentiometer excited by Vₑₓ, with displacement x over total length L:

The equation assumes a uniform resistive track and negligible loading. Mechanical wear, contact noise, limited life, and loading error restrict its use in demanding applications.

An LVDT uses differential transformer action. It provides excellent repeatability and no sliding electrical contact, but requires excitation and demodulation electronics. It is appropriate for precision linear displacement and harsh environments when properly protected.

An encoder converts shaft motion into electrical codes. An incremental quadrature encoder produces two pulse trains, A and B, offset by 90 electrical degrees. Their phase relationship indicates direction. If an encoder produces N cycles per revolution and the counting system detects q counts per cycle, angular resolution is:

where Δθ is in radians. For example, a 1000-line encoder using four-edge counting gives 4000 counts per revolution, or 0.09° per count. This ideal result excludes eccentricity, interpolation error, missed pulses, and mechanical backlash.

An absolute encoder outputs a unique code for each position. It avoids the need to recover position by counting from a reference after power interruption. Gray code is often used because only one bit changes between adjacent positions, reducing transition ambiguity.

3.2 Speed

Speed may be measured using encoders, tachogenerators, Hall-effect sensors, magnetic pickups, or optical systems. If a pulse sensor produces n pulses during a gate interval Δt and the monitored object produces p pulses per revolution, rotational speed in revolutions per minute is:

The result assumes all pulses are detected and the speed is approximately constant during the gate interval. Increasing Δt improves resolution but delays the estimate. Period measurement, based on the time between pulses, is often better at low speed but more sensitive to timing jitter.

3.3 Force and torque

Strain gauges measure deformation. A bonded metal strain gauge has nominal relationship:

where GF is gauge factor, ΔR is resistance change, R is nominal resistance, and ε is mechanical strain. Strain is dimensionless and commonly expressed in microstrain. A Wheatstone bridge converts small resistance changes into a voltage and can compensate for temperature when matched gauges are used.

For a quarter-bridge with one active gauge and small strain, an approximate output is:

This approximation assumes a balanced bridge, small resistance changes, and ideal components. Instrumentation amplifiers with high common-mode rejection are normally required.

Load cells commonly use four strain gauges arranged so that tension and compression arms add their effects. Their output is often specified in mV/V, for example 2 mV/V at rated load. Excitation stability, mechanical alignment, overload protection, and cable shielding strongly affect accuracy.

where T is torque in N·m, F is tangential force in newtons, and r is perpendicular radius in metres. A torque transducer may use strain gauges, magnetoelastic effects, or optical telemetry.

3.4 Pressure

The SI unit is pascal, where 1 Pa = 1 N/m². Industrial instruments may report gauge pressure relative to atmospheric pressure, absolute pressure relative to vacuum, or differential pressure between two points.

Diaphragm, Bourdon-tube, piezoresistive, capacitive, and resonant sensors are widely used. A differential-pressure transmitter can measure flow, liquid level, or filter blockage. For a liquid of density ρ in a static column of height h:

where g is gravitational acceleration. This relationship assumes a stationary, approximately incompressible liquid and neglects acceleration and friction effects.

Pressure sensors must be selected for proof pressure, burst pressure, temperature range, chemical compatibility, response time, and connection type. Pulsation dampers may be needed in reciprocating machinery.

3.5 Flow

Flow may be measured as volumetric flow rate Q or mass flow rate ṁ. The relationship is:

Common meters include turbine, electromagnetic, vortex, Coriolis, ultrasonic, and differential-pressure meters. Electromagnetic meters require electrically conductive fluids and have no obstruction in the flow path. Coriolis meters can measure mass flow directly but are relatively expensive. Turbine meters provide pulses proportional to flow but are sensitive to viscosity, contamination, and flow disturbances.

For an idealized restriction meter, flow is related approximately to the square root of differential pressure:

Real installations require a meter coefficient, fluid properties, geometry, and adequate straight pipe length. A transmitter or PLC often performs square-root extraction; applying it twice is a common commissioning error.

3.6 Temperature

Resistance temperature detectors (RTDs) exploit the change in resistance of a metal. For a narrow temperature range, a linear approximation is:

where R₀ is resistance at reference temperature T₀, α is the temperature coefficient, and T is temperature. Platinum Pt100 sensors have nominal resistance 100 Ω at 0 °C. The approximation is not sufficient over the full industrial temperature range; standardized polynomial characteristics are then used.

Thermistors have large, nonlinear resistance changes and are useful for low-cost measurement over limited ranges. Thermocouples generate a small voltage from the Seebeck effect and can operate over wide temperature ranges. Their output requires cold-junction compensation because the measured voltage represents a temperature difference between the measuring and reference junctions.

Infrared sensors measure emitted radiation without contact. Emissivity, reflected radiation, viewing angle, surface condition, and atmospheric absorption must be considered. A non-contact sensor does not automatically provide a more accurate measurement.

3.7 Proximity and object detection

Inductive proximity sensors detect metallic targets through changes in an electromagnetic field. They are robust and insensitive to visible light but have short range and detect primarily metal.

Capacitive sensors detect changes in capacitance and can sense non-metallic materials, powders, liquids, and objects through some nonconductive barriers. Humidity, material dielectric constant, and adjustment affect performance.

Photoelectric sensors use a light beam. Through-beam arrangements generally provide the greatest range and reliability; retroreflective and diffuse arrangements simplify installation but may be affected by target colour, reflectivity, and ambient light.

Ultrasonic sensors use acoustic time of flight and are useful for distance or level measurement. Foam, turbulent surfaces, temperature gradients, and acoustic obstructions can reduce reliability.

A proximity sensor should be specified by sensing range, target material, switching frequency, response time, mounting constraints, output type, ingress protection, and immunity to interference.

3.8 Machine vision

A vision system typically contains illumination, optics, an image sensor, processing software, and an output interface. It can inspect shape, colour, dimensions, labels, orientation, and assembly completeness. Vision is powerful where multiple geometric features must be inspected, but it is not a universal replacement for a simple proximity switch.

Reliable vision depends on controlled lighting, stable camera mounting, appropriate lens selection, calibration, and a well-defined acceptance criterion. Pixel measurements must be converted to physical dimensions using a calibration target or known reference. Perspective distortion requires telecentric optics or geometric correction when dimensional accuracy is important.

Machine-learning vision can classify complex defects, but training data must represent expected variation and unacceptable defects. Performance should be validated using independent data, including false-positive and false-negative rates. Safety-related decisions require suitable validation and, where appropriate, a deterministic backup or certified safety architecture.

4. Signal Conditioning and Industrial Interfaces

Raw sensor signals are rarely suitable for direct control. Signal conditioning may include amplification, filtering, linearization, isolation, excitation, bridge completion, cold-junction compensation, and analog-to-digital conversion.

A voltage divider illustrates loading. For a sensor resistance Rₛ connected to a load Rᴸ across excitation Vₑₓ:

The ideal no-load assumption is valid only when Rᴸ is much larger than Rₛ. Otherwise, the measured voltage differs from the intended calibration.

where R is resistance in ohms, C is capacitance in farads, and f_c is the −3 dB cutoff frequency in hertz. The filter should attenuate unwanted noise without materially delaying the process signal. Excessive filtering can destabilize a control loop or hide a dangerous transient.

A differential input rejects voltage common to both input terminals. Shielding, twisted-pair wiring, single-point grounding, galvanic isolation, and physical separation from motor cables reduce electromagnetic interference. Shield termination depends on the system grounding design; indiscriminate grounding at both ends can create ground-loop currents.

Interface	Typical use	Main consideration
0–10 V	Short industrial analog links	Sensitive to voltage drop and interference
4–20 mA	Process transmitters	Robust over long cables; live-zero diagnostics
Pulse/frequency	Encoders and flow meters	Requires correct counting and timing
PNP/NPN discrete	Presence and limit detection	Must match PLC input convention
IO-Link	Smart point-to-point sensors	Parameterization and diagnostics
Fieldbus/Ethernet	Networked instruments	Configuration, timing, and cybersecurity

The 4–20 mA convention uses 4 mA as the lower live zero, allowing a broken wire or loss of power to be distinguished from a valid zero measurement. A linear transmitter mapping a variable x from xₘᵢₙ to xₘₐₓ gives:

The formula assumes a correctly ranged linear transmitter. For a 0–100 bar transmitter reading 60 bar, I = 4 + 16(0.60) = 13.6 mA.

5. Calibration and Uncertainty

Calibration compares an instrument with a reference of known traceability over specified points. A calibration procedure should define the measurand, reference standard, environmental conditions, ascending and descending points, acceptance limits, adjustment policy, and records.

A useful calibration sequence is: inspect and identify the instrument; allow thermal stabilization; apply zero; apply several known inputs across the range; record increasing and decreasing readings; calculate deviations; adjust only when authorized; repeat the test; label status and due date.

For independent standard uncertainties u₁, u₂, …, uₙ, the combined standard uncertainty is commonly estimated by the root-sum-square method:

This form assumes the contributions are independent and are expressed in compatible units. If a quantity y depends on correlated input variables, covariance terms may be required. An expanded uncertainty is often reported as:

where k is a coverage factor, often approximately 2 for a roughly normal distribution and high coverage, but its selection must reflect the uncertainty model and reporting convention.

For a difference y = a − b with independent uncertainties uₐ and uᵦ, uᵧ = √(uₐ² + uᵦ²). Thus, if a pressure reference contributes 0.20 bar and the transmitter contributes 0.30 bar, uᶜ = √(0.20² + 0.30²) = 0.361 bar. With k = 2, U ≈ 0.72 bar. This result is meaningful only if the stated uncertainty components and distributions are justified.

Calibration is not the same as verification. Calibration establishes the relationship between indication and reference; verification checks whether performance meets a specified requirement. Traceability means an unbroken documented chain of comparisons to recognized standards.

6. Selecting and Installing Sensors

Selection begins with the measurement requirement rather than with a preferred technology. Define range, required accuracy, resolution, response time, environment, process connection, power availability, output interface, maintenance access, safety requirements, and lifecycle cost.

Installation errors often dominate sensor errors. Examples include a thermocouple placed too close to a heat source, a pressure tap located in a pulsating region, a flow meter installed without adequate straight pipe, an encoder coupled with excessive shaft misalignment, or a vision camera exposed to uncontrolled illumination.

Redundancy may be justified for critical measurements, but two identical sensors do not remove common-cause failures. Diagnostic coverage, proof testing, voting logic, and fail-safe design should be considered in safety-related applications. Functional safety standards such as IEC 61508 and sector-specific standards may apply; compliance must be demonstrated by the responsible engineering organization rather than assumed from the presence of a certified component.

7. Worked Example: Encoder Speed and Measurement Quality

An incremental encoder produces 1024 pulses per revolution. A PLC counts 512 pulses during a 0.25 s gate interval. Determine the shaft speed.

The result assumes one selected channel and one pulse per encoder line. If the PLC counts both rising and falling edges of both quadrature channels, p must be replaced by the effective count per revolution. Confusing lines, pulses, and quadrature counts can produce a factor-of-two or factor-of-four error.

The gate interval also creates quantization uncertainty. One additional or missing count changes the result by:

This is only a counting-resolution estimate; mechanical error, timing jitter, missed pulses, and speed variation add further uncertainty.

8. Applied Case: Conveyor Inspection and Reject System

Background

A packaging line transports containers at variable speed. Each container must be detected, its cap presence checked, its fill height measured, and rejected if defective. The line includes a variable-frequency motor drive and a PLC-based control system.

Problem

The original system used a mechanical limit switch for presence detection and a manually inspected fill level. At higher speed, switch wear caused missed detections, and reject timing varied with conveyor speed. Electrical noise from the motor drive also caused occasional false PLC inputs.

Analysis

A non-contact photoelectric sensor was selected for container presence. A second sensor was required for cap inspection. Fill height could be measured using a laser displacement sensor, but container colour and surface reflectivity required controlled testing. An encoder mounted on the conveyor shaft was needed to relate object position to conveyor travel rather than to time alone.

The sensor outputs were 24 V discrete signals, an encoder pulse train, and a 4–20 mA level signal. The PLC input modules were checked for voltage compatibility, pulse frequency capacity, input filtering, and isolation. Shielded encoder cable was routed separately from motor power conductors.

Approach and solution

The presence sensor was mounted in a rigid bracket with a defined beam path. A background-suppression photoelectric sensor reduced sensitivity to the conveyor. The encoder provided distance pulses. If the conveyor encoder generated 500 pulses per metre and the reject station was 1.2 m downstream, the nominal reject displacement was 600 pulses. The PLC stored each detected container in a first-in, first-out queue and triggered the reject valve when the corresponding count was reached.

The level transmitter was calibrated at empty, nominal, and high reference points. A moving-average filter was applied only after testing its effect on response time. A fault state was generated for current below approximately 3.6 mA or above the configured overrange, subject to the transmitter manufacturer’s diagnostic specification.

Outcome

The system eliminated mechanical contact at the detection point and maintained reject timing when conveyor speed changed. Shielding and input isolation reduced false transitions. The most important improvement was architectural: reject timing was based on measured conveyor position, not on a fixed delay.

Learning points

9. Limitations, Ethics, and Contemporary Practice

Every sensor has a domain of validity. Specifications are usually obtained under defined conditions and may not apply after contamination, vibration, temperature excursions, or incorrect mounting. Data may appear precise while representing the wrong physical location or a biased process condition.

Measurement systems increasingly connect to industrial networks and cloud platforms. Smart sensors can report calibration status, operating hours, internal temperature, and diagnostic codes. These capabilities support predictive maintenance, but they introduce cybersecurity and data-governance requirements. Networked instruments should use authenticated access, controlled firmware updates, segmentation, backups, and least-privilege configuration.

AI-based inspection should be evaluated with representative data and documented performance. Engineers must consider bias caused by lighting, product variation, sampling decisions, and changing production conditions. False rejects create economic waste; missed defects may create safety or legal risks. Human oversight and clear responsibility remain necessary for consequential decisions.

Best practice is to maintain a sensor register containing tag, measurand, range, location, technology, signal type, calibration interval, uncertainty, failure response, and maintenance history. Acceptance testing should include normal operation, boundary conditions, fault injection where safe, and recovery after power interruption.

Chapter Summary

Industrial measurement converts physical variables into information used for monitoring, control, safety, and quality decisions. Sensors may be active or passive, analog or digital, contact or non-contact, absolute or incremental. Important performance characteristics include accuracy, precision, sensitivity, resolution, range, linearity, hysteresis, repeatability, drift, and dynamic response.

Position and speed are commonly measured with potentiometers, LVDTs, encoders, tachometers, and magnetic sensors. Strain gauges and load cells measure force and torque; pressure instruments use diaphragms, strain, capacitance, or fluid-column relationships; flow meters use velocity, differential pressure, electromagnetic, ultrasonic, turbine, or Coriolis principles. RTDs, thermocouples, thermistors, infrared sensors, proximity devices, and vision systems address temperature and detection requirements.

Signal conditioning includes excitation, amplification, filtering, isolation, linearization, and conversion. Industrial interfaces such as 4–20 mA, pulse signals, discrete PNP/NPN inputs, IO-Link, and fieldbus networks must match the controller and installation environment. Calibration establishes a traceable relationship with a reference, while uncertainty describes the doubt associated with the result. Good design integrates sensor physics, installation, diagnostics, safety, cybersecurity, and maintenance.

Key Terms

Signal conditioning: Processing that makes a sensor signal suitable for acquisition or control.

Review Questions

Short Answer Questions

Long Answer Questions

Analytical / Application Questions

Multiple Choice Questions with Answers

Question 1

Question 2

An encoder produces 1024 pulses per revolution. If 512 pulses are counted in 0.25 s, what is the speed?

Question 3

Correct Answer: B. It allows zero process value to be distinguished from a broken wire

Question 4

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

Question 5

For an ideal first-order sensor, the output reaches approximately what percentage of its final step value after one time constant?

Question 6

A load cell has a gauge factor of 2, strain of 500 microstrain, and a quarter-bridge arrangement. What is the approximate ratio Vₒ/Vₑₓ?

Question 7

Question 8

A pressure transmitter is ranged from 0 to 100 bar. What current corresponds ideally to 60 bar on a 4–20 mA output?

Question 9

Question 10

Which flow technology can directly measure mass flow and is commonly more expensive than simple turbine measurement?

Chapter 3 Sensors, Transducers and Industrial Measurement

Sensors, Transducers and Industrial Measurement

Introduction