5. Signal Conditioning, Data Acquisition and Instrumentation

Introduction and Context

Mechatronic systems depend on a continuous exchange of information between the physical world and computational controllers. Temperature, pressure, displacement, force, acceleration, speed, current and position are physical variables; controllers and computers, however, process electrical signals represented as voltages, currents or digital numbers. The interface between these two domains is instrumentation. It determines whether a sensor signal is faithfully measured, whether a controller receives useful information, and whether an actuator is commanded safely and accurately.

A sensor rarely produces a signal that can be connected directly to a microcontroller or industrial computer. Its output may be very small, noisy, nonlinear, high-impedance, differential, or superimposed on a large common-mode voltage. Signal conditioning modifies the signal so that it is compatible with subsequent measurement hardware. Data acquisition then samples, converts, stores and communicates the conditioned signal. The design must account for amplification, isolation, filtering, sampling, quantisation, grounding, interference and timing.

These issues are especially important in closed-loop mechatronic systems. A poorly conditioned measurement can introduce delay, aliasing or bias into a feedback loop. The resulting controller may oscillate, respond slowly or make unsafe decisions even when its algorithm is mathematically correct. Thus, instrumentation is not an accessory to control engineering; it is part of the dynamic system.

Learning Objectives

After studying this chapter, the reader should be able to:

1. Explain the purpose and architecture of signal-conditioning and data-acquisition systems.

2. Select suitable amplification, isolation and filtering methods for sensor interfaces.

3. Analyse sampling, aliasing, quantisation and conversion errors in ADC and DAC systems.

4. Design grounding, shielding and cabling arrangements that reduce interference.

5. Compare single-ended, differential and instrumentation-amplifier interfaces.

6. Estimate resolution, conversion range, sampling rate and measurement uncertainty.

7. Evaluate practical and professional trade-offs in real-time data acquisition.

8. Diagnose common instrumentation failures and propose robust solutions.

From Physical Variable to Digital Measurement

A general measurement chain consists of a measurand, sensor, signal conditioner, multiplexer or sample-and-hold circuit, analogue-to-digital converter, processor and software. In the reverse direction, a digital controller may produce a digital command that is converted by a DAC, filtered, amplified and delivered to an actuator.

The sensor performs transduction: it converts a physical quantity into an electrical variable. A strain gauge changes resistance, a thermocouple generates a small thermoelectric voltage, an optical encoder produces pulses, and a piezoelectric accelerometer generates charge or voltage. The raw signal must then be adapted to the input range and electrical characteristics of the acquisition system.

Important performance measures include sensitivity, range, resolution, accuracy, precision, linearity, hysteresis, repeatability, bandwidth and response time. Accuracy describes closeness to the true value, whereas precision describes the consistency of repeated measurements. Resolution is the smallest discernible change, although practical resolution is often worse than the nominal converter resolution because of noise and drift.

The measurement chain should be considered as a system. Sensor noise may be amplified along with the desired signal; a filter may reduce noise but introduce phase delay; isolation may improve safety but add capacitance and bandwidth limitations; and a high-resolution ADC is of little value if the analogue front end is unstable or poorly grounded.

Signal Conditioning: Matching the Sensor to the Interface

Amplification and attenuation

Many sensors provide millivolt-level outputs. An amplifier increases the signal to use a substantial portion of the ADC input range, thereby improving effective resolution. If a sensor produces 0–50 mV and the ADC accepts 0–5 V, a gain of approximately 100 is required. The design must also allow for sensor tolerance, overload and transient signals; operating permanently at the extreme limits is undesirable.

Attenuation is required when the sensor output exceeds the converter range. A resistor divider is suitable for stable, low-impedance DC signals, but its loading, resistor tolerance, bandwidth and protection must be considered. A buffer may be required between the divider and ADC input.

Operational amplifiers are commonly used for voltage amplification, buffering, summing and filtering. Their input offset voltage, bias current, noise density, gain-bandwidth product, slew rate and output-drive capability affect measurement quality. A rail-to-rail amplifier may be needed when signals approach the supply rails. For precision systems, resistor matching and thermal stability are as important as nominal gain.

Differential and instrumentation amplification

A single-ended measurement is made relative to a common ground. A differential measurement responds to the voltage difference between two input terminals. Differential interfaces are generally preferable when the sensor and acquisition device are separated by long cables or powered from different locations.

An instrumentation amplifier provides high input impedance, accurate differential gain and high common-mode rejection ratio (CMRR). CMRR indicates how effectively a circuit rejects a voltage appearing equally on both input terminals. In decibels,

CMRR = 20 log10(Ad / Ac)

where Ad is differential gain and Ac is common-mode gain. High CMRR is valuable in bridge sensors, thermocouples and industrial environments, but it is reduced by resistor mismatch, frequency, cable imbalance and input overload. A large common-mode voltage can still saturate the amplifier even when the differential signal is small; therefore, input common-mode range must be checked.

A bridge circuit, such as a Wheatstone bridge containing strain gauges, often produces a differential output proportional to mechanical strain. Bridge excitation must be stable, and lead resistance, temperature effects and imbalance should be considered. Ratiometric measurement, in which the same excitation reference is used by the ADC, can reduce errors caused by excitation variation.

Isolation

Galvanic isolation prevents direct conductive current between two circuits. It is used to protect users, prevent ground loops, interface high-voltage equipment and improve electromagnetic compatibility. Isolation may be achieved using transformers, optocouplers, isolated amplifiers or isolated digital communication devices.

Isolation is not merely a safety component. It introduces propagation delay, parasitic capacitance, finite common-mode transient immunity and sometimes additional noise. An isolated analogue amplifier must be selected according to input range, bandwidth, isolation rating and accuracy. Digital isolators are often placed after an ADC, allowing the analogue front end to remain local to the sensor.

Safety isolation requires appropriate creepage, clearance, insulation and fault considerations. A component described as isolated is not automatically suitable for a particular mains or industrial application; the relevant standards and installation category must be examined.

Filtering

Filtering separates desired frequency components from unwanted components. A low-pass filter attenuates high-frequency noise, a high-pass filter removes slowly varying offsets, a band-pass filter selects a frequency region, and a notch filter suppresses a narrow interference frequency such as mains hum.

A first-order RC low-pass filter has a cutoff frequency

fc = 1/(2πRC)

At the cutoff frequency, the magnitude is approximately 0.707 of the passband value, corresponding to −3 dB. The roll-off of a first-order filter is approximately 20 dB per decade. Higher-order filters provide sharper attenuation but may produce greater phase shift, sensitivity to component tolerances and transient overshoot.

Filter selection should follow the measurement objective. A slowly varying temperature signal may tolerate heavy filtering, whereas an acceleration signal used for vibration control may require a wide bandwidth. Filtering should not remove information needed by the control algorithm. In feedback systems, phase delay can reduce stability margins.

Analogue anti-aliasing filters are placed before the ADC. Digital filters placed after sampling cannot recover frequency components that have already been aliased. A practical design often combines an analogue low-pass filter with oversampling and a digital filter.

Protection and excitation

Inputs should be protected against electrostatic discharge, wiring errors, overvoltage and transients. Series resistors, clamp diodes, transient suppressors, resettable fuses and current-limiting networks are common measures. Protection devices must not add excessive leakage, capacitance or nonlinear distortion.

Many sensors require excitation. Resistance temperature detectors need a current or voltage source; strain-gauge bridges require stable excitation; and active sensors may need regulated supply power. Excitation circuits should be designed with attention to self-heating, noise, compliance voltage and fault detection. A four-wire resistance measurement can reduce the effect of lead resistance compared with a two-wire arrangement.

Sampling, Quantisation and Conversion

Sampling theory and aliasing

An ADC represents a continuous-time signal by taking measurements at discrete intervals. If the sampling frequency is fs, the Nyquist frequency is fs/2. A band-limited signal must be sampled at a frequency greater than twice its highest significant frequency to avoid aliasing.

Aliasing occurs when a frequency component above fs/2 appears in the sampled data as a false lower frequency. For a sinusoid of frequency fin, the apparent frequency is related to the difference between fin and an integer multiple of fs. The result may look plausible and therefore be difficult to diagnose. Increasing the sample rate alone is not sufficient if interfering frequencies remain above the new Nyquist frequency; an appropriate anti-alias filter is essential.

The sampling rate must also account for control-loop timing, computational delay, multiplexing, communication latency and the bandwidth of the sensor. In real-time control, deterministic sampling is often more important than a very high average rate. Clock jitter creates additional error, particularly for high-frequency or rapidly changing signals.

Quantisation and ADC performance

An ideal N-bit ADC with input range VFS has a nominal step size

q = VFS / 2^N.

Quantisation error is commonly modelled as approximately ±q/2 for an ideal converter. The ideal signal-to-quantisation-noise ratio for a full-scale sinusoid is approximately 6.02N + 1.76 dB. Real converters perform worse because of thermal noise, nonlinearity, reference noise, aperture uncertainty and missing codes.

The effective number of bits (ENOB) expresses the resolution actually achieved in the presence of noise and distortion. A 16-bit converter may have an ENOB substantially below 16. Oversampling and averaging can improve effective noise performance when the noise is sufficiently uncorrelated, but they do not correct gain error, offset error or systematic interference.

ADC architectures include successive-approximation-register (SAR), sigma-delta, flash and pipeline converters. SAR ADCs offer a useful balance of speed, resolution and power for many control systems. Sigma-delta ADCs provide high resolution and integrated digital filtering, often at lower effective bandwidth. Flash converters are extremely fast but require many comparators and are generally used where very high sampling rates are necessary.

Reference voltage stability is fundamental. The ADC measures relative to its reference, so reference noise and drift directly affect the result. Calibration may be performed at zero and at one or more known input levels. For high-accuracy systems, temperature compensation and periodic recalibration may be required.

DACs and output reconstruction

A DAC converts a digital code into an analogue voltage or current. Its important characteristics include resolution, settling time, monotonicity, integral nonlinearity, differential nonlinearity, glitch energy and output drive. A DAC output often contains a staircase waveform and spectral images at multiples of the update frequency. A reconstruction low-pass filter smooths the output when a continuous analogue command is required.

PWM is an alternative to a dedicated DAC. The duty cycle represents the command, while a low-pass filter or actuator dynamics produce an average value. PWM resolution, carrier frequency, ripple, switching noise and update timing must be considered. For motors and power converters, PWM is often used directly because the power stage is inherently switching.

Data-Acquisition Architectures

A data-acquisition system may be centralised, distributed or networked. In a centralised arrangement, multiple sensors are wired to a common acquisition unit. A distributed system places local conditioning and conversion near the sensors and communicates digital data over a fieldbus or industrial Ethernet. Distributed acquisition reduces analogue cable length but increases requirements for synchronisation, network reliability and local power.

Multiplexing allows one ADC to serve several channels. A multiplexer introduces channel-switching time, crosstalk and settling requirements. After changing channels, the source impedance and sample-and-hold capacitor must settle before conversion. High-impedance sensors may therefore require a buffer or longer acquisition time.

Simultaneous-sampling ADCs are preferred when phase relationships between channels matter, as in vibration analysis, power measurement and multi-axis control. Sequential sampling can create time skew that is insignificant for slow variables but serious for high-frequency signals.

A real-time acquisition system must define a timing policy. Sampling, conversion, computation and actuation should occur within known deadlines. Buffering and interrupts can improve efficiency, but uncontrolled operating-system delays may cause jitter. Critical control loops commonly use timer-triggered acquisition and deterministic scheduling, while supervisory data logging can operate asynchronously.

Data integrity also requires range checking, timestamping, calibration information, diagnostic flags and detection of missing or implausible data. A value should not be accepted merely because it lies within the electrical range; rate-of-change limits, redundancy and sensor-status information may be needed.

Grounding, Shielding and Electromagnetic Compatibility

Many instrumentation problems are caused not by the sensor or ADC but by the physical installation. A ground loop occurs when two grounded points are connected by more than one conductive path and a voltage difference drives unwanted current through signal conductors. The resulting error may appear as mains-frequency hum or a slowly varying offset.

Signal ground, protective earth, chassis ground and power-return conductors have different functions. Protective earth is primarily a safety path, whereas signal reference establishes the voltage against which a measurement is interpreted. They may be connected at a deliberate point, but they should not be confused.

Single-point grounding can reduce circulating currents in low-frequency analogue systems. At higher frequencies, however, a low-impedance ground plane and controlled return paths are often more effective than long radial connections. The correct arrangement depends on frequency, physical dimensions, isolation and safety requirements.

Shielded twisted-pair cable reduces magnetic pickup and provides a defined signal-return geometry. Shield termination is application-dependent: terminating one end may reduce low-frequency ground-loop current, whereas terminating both ends can improve high-frequency shielding when the chassis and bonding arrangement are appropriate. The shield should not be assumed to carry the measurement return current.

Electromagnetic interference can be conducted through power and signal lines or radiated through space. Mitigation includes physical separation of power and signal cables, twisted pairs, filtering at entry points, ferrites, enclosure bonding, differential signalling, controlled cable routing and reduction of switching-loop area. Motor drives, relays and PWM converters deserve particular attention because their rapid voltage and current transitions generate broad-spectrum interference.

Measurement Uncertainty and Calibration

A credible measurement includes an estimate of uncertainty. Sources include sensor calibration, amplifier gain, ADC quantisation, reference drift, temperature, noise, wiring, repeatability and environmental influence. Random uncertainty may be reduced through repeated observations, while systematic error requires calibration, modelling or hardware correction.

Calibration compares the instrument with a traceable reference under specified conditions. A two-point calibration estimates offset and gain; multipoint calibration can characterise nonlinearity. Calibration data should be version-controlled and associated with the instrument, channel and date. Software correction is useful, but it does not compensate for an unstable or unsafe physical interface.

In industrial systems, diagnostic coverage is important. Open-circuit detection, out-of-range detection, redundant sensors, plausibility checks and fail-safe states can prevent a faulty measurement from producing dangerous actuator commands. The design should define what happens when a sensor disconnects, saturates or reports contradictory information.

Applied Illustrative Case: Vibration Monitoring of a High-Speed Spindle

Background

A manufacturing laboratory develops a mechatronic monitoring system for a high-speed spindle. An accelerometer is mounted near the spindle bearing, and the system must identify vibration changes associated with imbalance or bearing deterioration.

Situation and problem

The accelerometer produces a small differential voltage with useful content up to 8 kHz. The initial prototype connects it through a long unshielded cable to a 12-bit microcontroller ADC sampling at 10 kHz. The recorded waveform shows low-frequency components that are not present in a reference analyser. Measurements also change when a variable-frequency motor drive is switched on.

Analysis

The 10 kHz sampling rate gives a Nyquist frequency of 5 kHz, below the sensor bandwidth. Components between 5 and 8 kHz can therefore alias into the measured band. The long unshielded cable is susceptible to switching interference, and the single-ended ADC input provides poor rejection of voltage differences between the sensor and controller grounds. The 12-bit range is also used inefficiently because the accelerometer output occupies only a small fraction of full scale.

Approach and solution

The revised interface uses a differential instrumentation amplifier located near the sensor. Its gain is selected so that the expected acceleration range uses approximately 70–80 percent of the ADC input range, leaving overload margin. A shielded twisted-pair cable carries the differential signal, while the shield is connected to the chassis at the acquisition enclosure according to the system grounding plan.

An analogue low-pass filter is designed with a passband covering the required 8 kHz content and strong attenuation above the selected sampling limit. The ADC sampling rate is increased to 25 kS/s, and a digital band-pass filter is applied after conversion for diagnostic analysis. A stable reference, channel calibration and timestamped acquisition are added. The motor-drive cable is physically separated from the sensor cable, and the acquisition electronics are housed in a bonded enclosure.

Outcome

The false low-frequency components are substantially reduced, and the measured spectrum agrees with the reference analyser within the intended accuracy. The system can distinguish normal operating vibration from a developing imbalance condition. The increased sampling rate alone did not solve the problem; improvement resulted from coordinated changes to bandwidth, anti-alias filtering, differential measurement, shielding, grounding and calibration.

Learning points

The case demonstrates that instrumentation errors are often system-level faults. A high-resolution ADC cannot compensate for aliasing. Shielding is ineffective if return currents are uncontrolled, and differential amplification cannot correct an input that exceeds its common-mode range. Design decisions must be based on the complete signal path and the physical environment.

Common Design Errors and Professional Considerations

A frequent mistake is selecting an ADC solely by its number of bits. Effective resolution, reference quality, input drive requirements, sampling rate and noise performance are more informative. Another mistake is placing a digital filter after an inadequately sampled signal and assuming that aliasing can be removed. It cannot.

Other common errors include leaving amplifier inputs without bias-current paths, ignoring sensor loading, using an unsuitable op-amp near the supply rails, routing analogue and motor cables together, connecting shields indiscriminately, neglecting warm-up drift, and failing to test overload recovery. Designers may also overlook the phase delay introduced by filtering in a feedback loop.

Instrumentation for safety-related machinery requires documented requirements, verification, fault response and traceability. Measurement data may influence personnel safety, product quality or regulatory compliance; therefore, calibration records, access control, secure configuration and controlled software changes are professional responsibilities. Network-connected acquisition systems should protect data integrity and availability through authentication, segmentation, update management and diagnostic logging.

Digital Tools and Emerging Directions

Modern acquisition platforms increasingly combine local processing, high-speed converters, industrial Ethernet and edge analytics. Embedded systems can perform spectral estimation, feature extraction and anomaly detection close to the sensor, reducing communication bandwidth and response time. Machine-learning methods can assist predictive maintenance, but their performance depends on representative, well-calibrated data. A model cannot reliably compensate for systematic sensor faults or undocumented changes in the acquisition chain.

Time-sensitive networking and synchronised clocks support coordinated measurements across distributed devices. Smart sensors may communicate calibrated values and health information rather than raw voltages. Nevertheless, analogue protection, grounding, filtering and isolation remain necessary because digital communication begins only after a physical signal has been acquired.

Best-Practice Design Workflow

A disciplined design process begins by specifying the measurand, range, required accuracy, bandwidth, response time and environmental conditions. The engineer then selects the sensor and defines excitation, computes gain and input range, determines the sampling rate, and designs analogue anti-alias filtering. Grounding, shielding, isolation, protection and cable routing should be designed before hardware construction.

The prototype should be tested with known signals and realistic interference. Measurements should be checked for offset, gain, noise spectrum, linearity, channel crosstalk, timing jitter and overload behaviour. Calibration and uncertainty analysis should follow. Finally, the complete system should be tested under normal, fault and worst-case conditions rather than only with an ideal laboratory source.

Chapter Summary

Signal conditioning and data acquisition form the essential interface between physical processes and digital mechatronic controllers. Amplification, attenuation, buffering, differential measurement and isolation make sensor outputs compatible with acquisition hardware. Analogue filters reduce unwanted frequency components, while anti-aliasing filters are required before sampling. Sampling frequency, quantisation, reference stability, converter architecture and timing determine the quality of digital data.

Grounding, shielding, cable routing and electromagnetic compatibility are as important as circuit selection. Real-time acquisition must address channel settling, synchronisation, latency and data integrity. Calibration, uncertainty analysis and fault handling convert electrical measurements into trustworthy engineering information. Effective design is therefore a coordinated process involving sensors, analogue electronics, converters, software, physical installation, safety and professional documentation.

Key Terms

Signal conditioning: Modification of a sensor signal to make it suitable for measurement or control.

Transducer: A device that converts a physical quantity into an electrical signal, or vice versa.

Instrumentation amplifier: A precision differential amplifier with high input impedance and high common-mode rejection.

Common-mode rejection ratio: The ratio of differential gain to common-mode gain, usually expressed in decibels.

Galvanic isolation: Electrical separation that prevents direct conductive current between circuits.

Anti-aliasing filter: An analogue filter that limits input bandwidth before sampling.

Sampling: Measurement of a continuous-time signal at discrete time intervals.

Nyquist frequency: One-half of the sampling frequency.

Aliasing: Misrepresentation of high-frequency signal content as a lower-frequency component after sampling.

Quantisation: Mapping of a continuous amplitude into discrete digital levels.

Effective number of bits: The practical resolution of an ADC after noise and distortion are considered.

DAC: A digital-to-analogue converter that produces an analogue output from a digital code.

Multiplexing: Switching several input channels to a common acquisition converter.

Ground loop: An unwanted circulating current caused by multiple conductive grounding paths.

Shielding: Use of a conductive enclosure or cable layer to reduce electromagnetic coupling.

Calibration: Comparison and adjustment of an instrument against a reference.

Uncertainty: A quantified indication of doubt associated with a measurement result.

Settling time: Time required for a circuit or converter output to reach and remain within a specified error band.

Ratiometric measurement: A measurement in which the signal and reference depend on the same excitation quantity.

Review Questions

Short Answer Questions

1. Why is an analogue anti-aliasing filter required before an ADC?

2. Distinguish between accuracy, precision and resolution.

3. What is the function of an instrumentation amplifier?

4. Define a ground loop and state one method of reducing it.

5. What factors influence the practical performance of an ADC beyond its nominal bit count?

Long Answer Questions

1. Explain the complete signal path from a physical measurand to a digital controller, including the purpose of each stage.

2. Discuss the selection of gain, filter bandwidth, sampling rate and ADC resolution for a sensor interface.

3. Compare single-ended, differential and isolated measurement architectures.

4. Explain aliasing, quantisation and reference-voltage errors and describe methods of mitigation.

5. Discuss grounding, shielding, cable routing and electromagnetic interference in industrial data-acquisition systems.

Analytical / Application Questions

1. A 0–100 mV sensor is connected to a 0–5 V, 12-bit ADC. Determine a suitable nominal amplifier gain, allowing reasonable overload margin, and calculate the approximate ADC step referred to the sensor input.

2. A vibration signal contains useful components up to 6 kHz, but the acquisition system samples at 8 kS/s. Analyse the risk and propose a revised filtering and sampling strategy.

3. A differential sensor produces a 20 mV signal while both input terminals are at 12 V relative to the controller ground. Select and justify an interface approach, identifying common-mode and isolation issues that must be checked.

22. Multiple Choice Questions with Answers

Question 1

A sensor produces a 0–50 mV signal and an ADC accepts 0–5 V. Which nominal amplifier gain best uses the ADC range?

A. 10

B. 50

C. 100

D. 500

Correct Answer: C. 100

Question 2

What is the primary purpose of an analogue anti-aliasing filter?

A. To increase ADC resolution

B. To limit frequencies entering the ADC above the permitted bandwidth

C. To convert digital data into analogue voltage

D. To eliminate quantisation error after conversion

Correct Answer: B. To limit frequencies entering the ADC above the permitted bandwidth

Question 3

An ADC samples at 20 kS/s. Its Nyquist frequency is:

A. 5 kHz

B. 10 kHz

C. 20 kHz

D. 40 kHz

Correct Answer: B. 10 kHz

Question 4

Which interface is generally most suitable for a small sensor signal superimposed on a large common-mode voltage?

A. Unbuffered single-ended input

B. Instrumentation amplifier with adequate common-mode range

C. Simple resistor divider only

D. Digital filter connected before the sensor

Correct Answer: B. Instrumentation amplifier with adequate common-mode range

Question 5

A 12-bit ADC has a 0–4.096 V input range. Its ideal quantisation step is approximately:

A. 0.001 V

B. 0.01 V

C. 0.1 V

D. 1 V

Correct Answer: A. 0.001 V

Question 6

Why can a digital filter not correct aliasing that has already occurred?

A. Digital filters cannot attenuate low frequencies

B. Aliased components have already been represented as false lower-frequency data

C. Digital filters operate only on analogue signals

D. Sampling always removes phase information

Correct Answer: B. Aliased components have already been represented as false lower-frequency data

Question 7

Which condition is most likely to produce a ground-loop error?

A. A single isolated sensor supply

B. Multiple conductive ground paths between separated instruments

C. A properly terminated shield

D. A stable ADC reference

Correct Answer: B. Multiple conductive ground paths between separated instruments

Question 8

In a multiplexed ADC system, why may a buffer be required for a high-impedance sensor?

A. To increase cable length without attenuation

B. To provide sufficient drive current and allow the sample-and-hold capacitor to settle

C. To reduce the ADC number of bits

D. To create galvanic isolation automatically

Correct Answer: B. To provide sufficient drive current and allow the sample-and-hold capacitor to settle

Question 9

Which change directly improves rejection of voltage differences between a remote sensor and acquisition unit?

A. Using a differential input arrangement

B. Reducing the ADC resolution

C. Removing all filtering

D. Increasing the sensor impedance

Correct Answer: A. Using a differential input arrangement

Question 10

In the spindle vibration case, why was increasing the sample rate alone insufficient?

A. The sensor produced no electrical signal

B. Grounding, interference and differential-interface problems also affected the measurement

C. A higher sample rate always increases aliasing

D. The ADC could not perform quantisation

Correct Answer: B. Grounding, interference and differential-interface problems also affected the measurement