A mechatronic product rarely operates as an isolated embedded controller. Sensors, actuators, drives, human–machine interfaces, safety devices, supervisory computers, and cloud gateways exchange information continuously. Communication therefore becomes part of the system’s control architecture, not merely an auxiliary feature.

The choice of communication technology affects sampling rates, control-loop delay, synchronization, fault response, wiring cost, electromagnetic compatibility, maintainability, cybersecurity, and interoperability. A protocol that is excellent for connecting a microcontroller to a nearby sensor may be unsuitable for a distributed industrial machine.

This chapter develops a practical framework for implementing and comparing UART, SPI, I²C, CAN, LIN, USB, Ethernet, Modbus, fieldbus systems, and wireless protocols. The focus is on selecting technologies according to bandwidth, latency, determinism, reliability, interoperability, and cost.

Learning Objectives

1. Communication Architecture in Mechatronics

UART, SPI, and I²C are normally peripheral-level interfaces. CAN, Ethernet, and USB define more extensive communication mechanisms. Modbus is principally an application protocol that may operate over serial links or Ethernet. A fieldbus is an industrial communication system generally combining a physical medium, data-link behavior, application conventions, diagnostics, and device interoperability rules.

A useful engineering distinction is between data rate and useful throughput. Data rate is the number of transmitted signaling bits per second. Useful throughput excludes framing, addressing, acknowledgements, padding, inter-frame gaps, and retransmissions.

For a message with N_p payload bits, N_o overhead bits, and signaling rate R_b bits/s, the ideal transmission time is:

Here, t_tx is in seconds. This equation assumes that all bits use the same signaling rate and excludes propagation delay, processing delay, queuing, and retransmission. If a protocol transmits N_p payload bits in a total frame of N_f bits, ideal payload efficiency is:

The terms represent queuing, transmission, propagation, processing, and response time. In a short machine, T_prop is often small, while processing and queueing can dominate.

2. Criteria for Protocol Selection

Bandwidth

Bandwidth is the capacity to transfer data. It must be considered together with message size and update frequency. A sensor requiring 12 bytes at 1 kHz generates 12,000 payload bytes/s, before protocol overhead. A link should provide sufficient margin for diagnostics, retries, synchronization, and future expansion.

Latency

Latency is the time between transmission request and useful reception. High nominal bandwidth does not guarantee low latency: a heavily loaded Ethernet network may delay a short message longer than a lightly loaded CAN bus.

Determinism

A deterministic system has a bounded and predictable response time. Scheduled time-triggered networks can be highly deterministic. Contention-based networks have variable access delay, although CAN’s priority arbitration provides a calculable upper bound when bus loading and message priorities are properly engineered.

Reliability

Reliability depends on differential signaling, shielding, error detection, acknowledgement, retransmission, redundancy, connector quality, and environmental design. A checksum detects many errors but does not automatically correct them. A protocol’s behavior during faults is as important as its normal data rate.

Interoperability

Interoperability is the ability of devices from different manufacturers to communicate with compatible meaning. A standard physical interface alone does not guarantee interoperability. For example, two Modbus devices may use the same transport but expose different register addresses and scaling conventions.

Cost and lifecycle

Cost includes transceivers, connectors, cabling, isolation, software development, diagnostic tools, certification, maintenance, and training. A low-cost interface can become expensive if it requires proprietary software or is difficult to diagnose.

3. Microcontroller Peripheral Interfaces

3.1 UART and asynchronous serial communication

UART transmits characters asynchronously, normally using separate transmit and receive lines. A frame commonly contains one start bit, 5–9 data bits, an optional parity bit, and one or more stop bits. Both endpoints must agree on baud rate, data length, parity, and stop-bit configuration.

With eight data bits, no parity, one start bit, and one stop bit, each character occupies 10 bit periods. The approximate payload rate is therefore:

For a 115,200 bit/s link, the ideal payload rate is 92,160 bit/s, or 11,520 bytes/s. This assumes continuous transmission and ignores software delays.

UART is simple and inexpensive, but ordinary UART has no inherent multi-drop arbitration, message addressing, or robust industrial fault handling. TTL/CMOS UART signals should not be confused with RS-232 or RS-485. RS-232 uses different voltage levels and point-to-point signaling; RS-485 commonly uses differential signaling and supports longer, noisier, multi-drop links when correctly terminated and biased.

Typical applications include bootloaders, GPS receivers, debug consoles, short device links, and Modbus RTU over RS-485.

3.2 SPI

Serial Peripheral Interface is a synchronous, usually master-controlled bus. Common signals are SCLK, MOSI, MISO, and chip-select lines. A master selects a slave and supplies the clock. Data can be shifted in both directions simultaneously.

SPI offers high throughput and low protocol overhead. It is especially suitable for displays, ADCs, DACs, memory devices, and inertial sensors located on the same circuit board. Its limitations include the need for a separate chip-select signal for many slaves, lack of a universal addressing scheme, limited cable length, and no intrinsic acknowledgement or error detection.

Clock phase and polarity are specified by the SPI mode. The firmware must also observe setup time, hold time, maximum clock frequency, and chip-select timing from the device data sheet. A robust design may add a checksum, framing command, timeout, and transaction-level validation because SPI itself does not guarantee message integrity.

3.3 I²C

I²C uses two open-drain lines: serial clock (SCL) and serial data (SDA). Pull-up resistors return the lines to the logic-high state. Devices share the bus and use addresses; arbitration and clock stretching are supported in the standard protocol.

Because the lines are capacitively loaded, rise time depends on pull-up resistance and total bus capacitance. A smaller resistance produces faster rising edges but increases low-level current. A larger resistance reduces current but can violate rise-time limits. Thus, pull-up selection must consider voltage, capacitance, device sink-current limits, speed mode, and electrical standards.

I²C is economical for multiple sensors and low-to-moderate speed board-level communication. It is less suitable for long cables or severe electromagnetic environments. Common failure modes include a slave holding SDA low, incorrect address selection, excessive capacitance, and inadequate recovery after reset. Bus-clear procedures, timeouts, and resettable peripherals improve robustness.

4. Controller Area Network

CAN was designed for distributed control systems such as vehicles and industrial machines. Nodes transmit frames onto a differential bus. A message identifier determines priority and meaning; it is not necessarily a node address.

CAN uses non-destructive bitwise arbitration. A dominant bit overrides a recessive bit, so the frame with the highest priority (numerically lowest identifier in classical CAN) continues while lower-priority transmitters withdraw and retry. This avoids corrupting the winning frame.

Classical CAN carries up to 8 data bytes per data frame. CAN FD permits larger payloads and can use a faster data phase, subject to transceiver, network, and timing support. CAN includes cyclic redundancy checking, acknowledgement, frame monitoring, error counters, and fault confinement. A node that repeatedly produces errors can be removed from bus participation.

A linear topology, appropriate termination at both physical ends, controlled stub lengths, suitable bit timing, and a common reference or galvanic isolation are essential. Star wiring and unterminated branches can produce reflections and intermittent faults.

CAN is strong where short messages require high reliability and priority-based access. Its limitations include limited payload per frame, configuration complexity, and the need for a higher-level convention such as CANopen, J1939, or a proprietary message dictionary.

5. LIN and Local Low-Cost Networks

Local Interconnect Network (LIN) is a low-cost, single-master, multi-slave serial network. The master schedules frame headers, and a designated slave supplies the response. This schedule gives predictable behavior for simple devices such as switches, mirrors, seats, and small actuators.

LIN uses a single communication wire plus ground and is less expensive than CAN hardware. It has lower speed and weaker fault tolerance, so it is normally used for non-critical subsystems rather than high-performance motion control or safety functions.

LIN illustrates an important design principle: the cheapest protocol is valuable when its performance and fault model match the application. Replacing every LIN connection with Ethernet may increase cost, software complexity, and maintenance without improving the product.

6. USB

Universal Serial Bus is a host-controlled, hierarchical serial interface. A host manages enumeration, assigns addresses, and schedules transactions. Devices identify themselves through descriptors and operate through defined endpoints and device classes.

USB supports standardized classes such as human-interface devices, mass storage, audio, communications, and vendor-specific devices. It offers convenient connectors and substantial bandwidth, but its host-centric architecture makes it less natural for peer-to-peer control networks. Cable length, connector retention, electromagnetic compatibility, power negotiation, and operating-system support must be considered.

USB is excellent for configuration, data logging, firmware updates, and connection to computers. It should not automatically be selected for hard real-time distributed control, where deterministic industrial Ethernet or CAN may be more appropriate.

7. Ethernet and Industrial Ethernet

Ethernet provides high bandwidth and a mature ecosystem of switches, controllers, diagnostic tools, and software. Modern switched Ethernet is normally full-duplex, reducing collision concerns present in historical shared-medium networks.

Standard Ethernet by itself does not guarantee bounded control latency. Queuing, switch behavior, operating-system scheduling, congestion, and protocol stacks can introduce variation. Industrial solutions address this using scheduled traffic, priority mechanisms, synchronization, traffic shaping, redundancy, or specialized hardware. Examples include EtherCAT, PROFINET, EtherNet/IP, and TSN-based systems, each with distinct implementation and certification requirements.

Ethernet’s strengths are bandwidth, interoperability at the general networking level, long-distance connectivity through standard infrastructure, and integration with supervisory systems. Its challenges include configuration, cybersecurity exposure, connector and cable requirements, and potentially nondeterministic behavior if the network is not engineered for the control task.

8. Modbus and Fieldbus Protocols

Modbus is an application-layer request–response protocol. Modbus RTU commonly operates over serial links such as RS-485, while Modbus TCP operates over Ethernet and TCP/IP. Devices expose coils, discrete inputs, input registers, and holding registers. A master or client requests data, and a slave or server responds.

Modbus is easy to understand and widely supported. However, register maps, byte order, signedness, scaling, and exception behavior must be documented explicitly. A value such as 250 may represent 250 °C, 25.0 °C, a raw ADC count, or an unsigned status code depending on the device specification. Modbus RTU uses timing gaps to delimit frames, so baud rate and silent interval requirements matter.

A fieldbus is broader than a simple register protocol. Industrial fieldbus families may define cyclic process data, acyclic parameter access, diagnostics, device profiles, synchronization, and conformance testing. Examples include PROFIBUS, CANopen, DeviceNet, FOUNDATION Fieldbus, and industrial Ethernet families.

Fieldbus selection should consider installed base, available drives and I/O, engineering tools, certification, vendor support, real-time requirements, and migration strategy. A technically superior network may be a poor business choice if it cannot integrate with existing plant assets.

9. Wireless Embedded and Industrial Networks

Wireless communication removes cables and supports mobile or difficult-to-wire devices, but introduces shared spectrum, propagation loss, interference, battery constraints, and security exposure. Technologies include Wi-Fi, Bluetooth Low Energy, Zigbee, Thread, WirelessHART, ISA100.11a, cellular IoT, and proprietary sub-GHz systems.

Wireless links should be selected according to range, data rate, power budget, topology, coexistence, roaming, latency, and security. A battery-powered sensor may favor low duty cycle and small packets, whereas a machine-vision system requires high throughput and generally uses wired Ethernet.

Wireless should not be treated as inherently unreliable. Reliability can be improved using channel planning, diversity, acknowledgements, retries, buffering, time synchronization, redundant paths, antenna design, and link-quality monitoring. Nevertheless, a safety function should not depend on an unverified wireless link merely because it is convenient.

10. Comparative Engineering Analysis

Technology	Typical strength	Main limitation	Determinism	Common use
UART	Very simple point-to-point link	No inherent arbitration	Moderate for one link	Debugging, modules
SPI	High board-level throughput	Many chip-selects; short distance	High locally	ADCs, displays, memory
I²C	Two-wire multi-device bus	Capacitance and noise limits	Moderate	Sensors, EEPROMs
CAN	Fault-aware priority bus	Limited payload and bandwidth	Good when engineered	Distributed control
LIN	Low hardware cost	Low speed, single master	Good scheduled behavior	Simple actuators
USB	Computer connectivity and classes	Host-centric and stack-heavy	Variable	Configuration, logging
Ethernet	High bandwidth and ecosystem	Configuration and latency variation	Depends on design	Plant and machine networks
Modbus	Broad interoperability	Weak semantic standardization	RTU depends on polling	PLC and instrumentation
Fieldbus	Industrial diagnostics and profiles	Vendor/ecosystem commitment	Often strong	Drives, I/O, automation
Wireless	Mobility and reduced cabling	Interference and security	Variable	Remote sensors, mobile systems

No single ranking is universally correct. A sensor inside a controller may use I²C, a fast converter may use SPI, distributed actuators may use CAN, and the machine may connect to a factory network through industrial Ethernet.

11. Worked Example: Selecting a Network for a Distributed Axis

Consider an illustrative machine with six motor drives. Each drive receives a 16-byte command and returns a 16-byte status message every 1 ms. Ignoring overhead, the payload per cycle is:

This is only the useful payload. Framing, identifiers, acknowledgements, inter-frame gaps, synchronization, diagnostics, and safety traffic require additional capacity. A network operating at 1 Mbit/s cannot carry this workload because its raw rate is already below the payload requirement. Classical CAN is therefore unsuitable for the stated traffic if all data must be delivered every millisecond in this form.

A higher-rate CAN FD or industrial Ethernet solution may be considered. The final decision must then examine worst-case scheduling, switch or bus behavior, drive support, synchronization accuracy, cable environment, safety certification, and cost. Reducing message size or using event-triggered status updates may also solve the problem more economically than simply selecting a faster network.

12. Implementation Practices

Reliable communication begins with a precise interface specification. It should define signal levels, connector pinout, baud rate, bit ordering, frame format, identifiers, units, scaling, timeout behavior, reset behavior, versioning, and fault responses.

A communication watchdog must distinguish a valid but stale message from a newly received message. Safety-related outputs should move to a defined safe state after loss of communication, but the safe state must be chosen through hazard analysis rather than assumed.

Testing should include nominal traffic, maximum bus load, malformed frames, unplugged nodes, short circuits where safe, power cycling, clock mismatch, EMI exposure, buffer overflow, packet duplication, packet reordering, and recovery after faults. Oscilloscopes, logic analyzers, protocol analyzers, bus monitors, and network captures are important engineering tools.

13. Illustrative Case Study: Packaging Machine Retrofit

Background

A packaging machine contains a central controller, four servo drives, distributed limit switches, a barcode reader, and a plant supervisory computer. The original machine used proprietary serial links and was difficult to maintain.

Problem

The retrofit must preserve the existing drives, improve diagnostics, support a 1 ms motion update, and expose production data to the factory network. Downtime and wiring changes must be minimized.

Analysis

The servo loop requires predictable cyclic traffic and accurate synchronization. Limit switches need little bandwidth but benefit from robust distributed wiring. The barcode reader produces bursts of data and is naturally connected to Ethernet or USB. The supervisory interface requires interoperability rather than hard real-time control.

Approach and solution

An industrial Ethernet network with a suitable real-time drive profile is selected for the servo drives. CAN is used for low-bandwidth auxiliary I/O where existing devices support it. The barcode reader uses Ethernet, while the supervisory system receives production data through a documented industrial Ethernet or Modbus TCP interface. A gateway separates motion traffic from supervisory traffic and prevents non-real-time requests from disturbing the motion schedule.

The design includes managed switches where appropriate, network segmentation, shielded and correctly grounded cabling, synchronized device clocks, diagnostic counters, and a defined safe response to lost drive communication. Existing proprietary serial devices are retained temporarily through a gateway, allowing staged migration.

Outcome

The architecture satisfies the motion requirement without forcing every device onto one protocol. Diagnostics become more accessible, the plant interface is standardized, and the migration risk is reduced. The principal engineering lesson is that a layered, heterogeneous network is often better than a single universal bus.

Learning points

14. Security, Safety, and Governance

Many legacy industrial protocols were designed for availability and simplicity rather than hostile environments. Modbus, CAN, and several fieldbus systems may lack strong authentication or encryption. An attacker with network access may inject commands, alter process values, or disrupt service.

Industrial networks should use segmentation, least privilege, controlled remote access, secure update procedures, asset inventories, logging, vulnerability management, and tested incident-response plans. Cryptography must be applied in a way that preserves timing and device-resource requirements. Security controls should not be added without considering deterministic behavior and recovery.

Safety and security are related but distinct. A checksum can detect accidental corruption but does not authenticate a sender. A secure protocol may still fail because of a broken cable or overloaded switch. Functional safety requires hazard analysis, independence where necessary, diagnostic coverage, validated response times, and compliance with applicable standards and regulations.

Engineers must also consider privacy when wireless or cloud-connected devices collect operator or production data. Documentation should state what data is collected, where it is stored, who may access it, and how long it is retained.

15. Contemporary Developments and Best Practice

Time-sensitive networking is extending Ethernet with time synchronization, scheduled traffic, and bounded performance mechanisms. OPC UA is increasingly used for information modeling and industrial interoperability, although it should not be assumed to replace every real-time field network. Edge gateways are combining protocol conversion, local analytics, and cybersecurity functions.

Artificial intelligence can assist network monitoring by identifying unusual traffic, timing drift, or error-counter patterns. It should support, not replace, deterministic protection mechanisms and engineering judgment. A learned anomaly detector cannot be the sole basis for a safety decision unless its behavior is rigorously validated for that purpose.

Chapter Summary

Embedded communication is a system-design decision involving physical signaling, framing, arbitration, application meaning, timing, reliability, safety, security, cost, and lifecycle support. UART, SPI, and I²C are economical local interfaces; CAN provides robust priority-based distributed control; LIN supports simple scheduled low-cost devices; USB is effective for host connectivity; Ethernet provides scalable bandwidth; Modbus offers broad but semantically limited interoperability; fieldbus systems add industrial profiles and diagnostics; and wireless technologies reduce cabling while introducing spectrum and security challenges.

Bandwidth must be calculated from payload and overhead, while latency must include queueing and processing. Determinism depends on access method and network engineering, not merely nominal bit rate. Successful designs commonly use several protocols connected through carefully defined gateways. Robust implementation requires explicit message definitions, error handling, diagnostics, fault states, security controls, and testing under worst-case conditions.

Key Terms

UART: Asynchronous serial interface using independently configured transmit and receive timing.

Modbus: Request–response industrial application protocol used over serial or TCP links.

Fieldbus: Industrial communication system supporting process data, diagnostics, and device profiles.

Baud rate: Number of signaling symbols transmitted per second; often equal to bit rate in binary signaling.

Latency: Time between initiating communication and receiving useful information.

Arbitration: Process by which competing transmitters determine which may use a shared medium.

Termination: Electrical impedance applied to a bus to reduce signal reflections.

Interoperability: Ability of independently produced systems to communicate with shared meaning.

Time-sensitive networking: Ethernet mechanisms supporting synchronized and bounded traffic.

Review Questions

Short Answer Questions

Long Answer Questions

Analytical / Application Questions

Multiple Choice Questions with Answers

Question 1

Which interface normally uses separate chip-select signals to select SPI slaves?

Question 2

Question 3

Question 4

Question 5

Question 6

A message contains 800 payload bits and 200 overhead bits. What is its ideal payload efficiency?

Question 7

Question 8

Why may ordinary Ethernet be unsuitable for hard real-time control without additional engineering?

Question 9

Question 10

Correct Answer: C. It is an application protocol that may use serial or TCP transport

Chapter 6 Embedded Communication and Industrial Networks

Embedded Communication and Industrial Networks

Introduction and Context