29 Real-Time ISA-95: Determinism and Technology Mapping

applications

iiot

real

time

29.1 Start With the Decision

A fast average link can still miss a control deadline when jitter spikes. Determinism, clock error, and deadline class decide which network belongs at each ISA-95 level.

29.2 Route Overview

This is part 2 of 3. Review Real-Time ISA-95: Timing Foundations for the preceding evidence.

29.3 Learning Objectives

Distinguish hard, firm, and soft real-time deadlines.
Map jitter, synchronization, and network technology to ISA-95 levels.

29.4 Chapter Roadmap

Determinism vs. Throughput
Hard vs Soft Real-Time Summary
Checkpoint: Deadline Classes
Jitter and Synchronization
Jitter Impact Calculator
Knowledge Check: Real-Time Requirements
Answer
Answer
Answer
Technology Mapping
Technology Mapping Summary
Checkpoint: Technology Placement
ISA-95 Level Selector Tool
Knowledge Check: Technology Mapping
Answer
Answer

29.5 Determinism vs. Throughput

With the level timing ladder in place, the next question is what kind of promise each deadline needs. Some deadlines are hard failures; others are quality-of-service targets.

Industrial systems distinguish between different real-time guarantees:

Look at Figure 29.1 for what changes when a deadline is a failure rather than a target.

Hard real-time and soft real-time timeline comparison showing bounded deterministic intervals for hard real-time and variable intervals for soft real-time. — Figure 29.1: Hard real-time versus soft real-time timing

The two timelines in Figure 29.1 carry the same task under different promises. On the upper track the intervals are even and a bound sits at the end of each one, because the guarantee is about the worst case rather than typical behaviour. On the lower track the intervals stretch and shrink around an average, and a late arrival is marked as acceptable. The panels underneath name the consequence. Miss the bounded deadline and the system has failed, so it must be sized for the slowest path through the code. Miss a soft target and the result is only degraded, so average load is enough to design for.

Hard real-time paths are engineered around bounded worst-case behavior; soft real-time paths are judged by usefulness under expected delay and jitter.

Hard vs Soft Real-Time Summary

Hard real-time: Response time is bounded and predictable. Missing a deadline is a system failure.
Soft real-time: Response time varies, and occasional delays degrade quality but do not stop the system.
WCET guarantee: Hard real-time systems must prove worst-case execution time stays within the deadline.
Statistical QoS: Soft real-time systems optimize average performance and tolerate some jitter.

29.5.1 Hard Real-Time (Deterministic)

Definition: Must respond within guaranteed time; missing deadline is system failure.

Characteristics:

Worst-case execution time (WCET) must be bounded
Jitter must be minimal (<1μs for synchronized motion)
Preemptive, priority-based scheduling
Often requires specialized hardware

Examples:

Safety systems (emergency stop)
Motion control (coordinated axes)
Process control (exothermic reactions)

Implementation approaches:

Dedicated real-time networks (EtherCAT, PROFINET IRT)
Real-time operating systems (VxWorks, QNX, RTAI)
FPGA-based control
Time-triggered architectures

29.5.2 Soft Real-Time

Definition: Should respond quickly but occasional delays acceptable; results in degraded performance, not failure.

Characteristics:

Average response time matters more than worst-case
Some deadline misses tolerable
Standard operating systems acceptable
Statistical quality of service

Examples:

HMI updates
Data logging
Trend analysis
Operator notifications

29.5.3 Best Effort

Definition: No timing guarantees; response when resources available.

Examples:

Historical data analysis
Business reporting
Email notifications
Non-critical analytics

Checkpoint: Deadline Classes

You now know: Begin with hard real-time means a missed deadline is a failure, which is why safety systems and coordinated motion need bounded worst-case execution time. Next consider soft real-time fits HMI updates, data logging, trend analysis, and operator notifications where delay degrades usefulness. Then test best-effort paths belong to historical analysis, business reporting, email notifications, and other work where completeness matters more than millisecond delivery.

29.6 Jitter and Synchronization

For coordinated motion and distributed control, jitter (timing variation) is often more critical than absolute latency:

Figure 29.2 shows why an average latency figure can hide a missed deadline.

Two command-stream timelines have the same illustrative mean latency of 20 milliseconds. Low-jitter arrivals cluster near expected update times and preserve deadline margin, while high-jitter arrivals spread early and late and include a missed margin. A coordination panel shows aligned axes for consistent arrivals and positional mismatch for variable arrivals. Jitter is identified as timing variation, not mean latency or packet loss. — Figure 29.2: Two streams can have the same average latency but very different jitter and coordination outcomes.

The two axes in Figure 29.2 are aligned on purpose, and both streams average 20 ms. On the upper axis the five arrivals sit between 18 and 22 ms, so every one lands before the 30 ms deadline with margin to spare. On the lower axis the same average is built from arrivals at 8 ms and 32 ms, and the late one crosses the deadline. The panel on the right shows what that costs a machine: steady arrivals keep two axes in step, while variable arrivals push them out of position. The closing line is the definition to carry forward. Jitter is variation in arrival timing, not the mean and not packet loss.

Jitter requirements by application:

Application	Cycle Time	Max Jitter	Nodes
Simple I/O	10 ms	1 ms	10-100
Process control	100 ms	10 ms	100-1000
Packaging machinery	1 ms	100μs	10-50
Printing press	125μs	1μs	20-100
Semiconductor handling	62.5μs	100ns	10-30

Simple I/O: 10 ms cycle time, 1 ms max jitter, 10-100 nodes
Process control: 100 ms cycle time, 10 ms max jitter, 100-1000 nodes
Packaging machinery: 1 ms cycle time, 100μs max jitter, 10-50 nodes
Printing press: 125μs cycle time, 1μs max jitter, 20-100 nodes
Semiconductor handling: 62.5μs cycle time, 100ns max jitter, 10-30 nodes

Synchronization mechanisms:

IEEE 1588 (PTP): Precision Time Protocol for sub-microsecond sync
Distributed clocks: EtherCAT’s hardware-based synchronization
Time-triggered protocols: Deterministic message scheduling
GPS timing: Absolute time reference for wide-area systems

The chapter now has two timing ideas on the table: response time and consistency. The jitter calculator isolates the second one so you can see why a smaller average delay is not enough for synchronized axes.

29.7 Jitter Impact Calculator

Try it: Adjust the actual jitter value to see how different applications tolerate timing variation. Notice that printing presses and semiconductor handling require sub-microsecond precision, while simple I/O can tolerate milliseconds of jitter.

Knowledge Check: Real-Time Requirements

Question 1: A printing press requires a cycle time of 125 microseconds with maximum jitter of 1 microsecond. Which type of real-time guarantee does this require?

a) Best effort b) Soft real-time c) Hard real-time d) Near real-time

Answer

c) Hard real-time — A printing press with a 125 microsecond cycle time and 1 microsecond maximum jitter tolerance is a classic hard real-time application. Missing a deadline would cause visible print defects (color misregistration, smearing). Hard real-time requires guaranteed worst-case execution time (WCET) and minimal jitter, typically implemented with specialized hardware like EtherCAT with distributed clocks.

Question 2: At which ISA-95 level would you place a SCADA system that displays alarm conditions to plant operators?

a) Level 0 — Field Devices b) Level 1 — Basic Control c) Level 2 — Supervisory d) Level 3 — Operations

Answer

c) Level 2 — Supervisory — SCADA (Supervisory Control and Data Acquisition) systems operate at ISA-95 Level 2, which handles monitoring and supervision of production processes. Level 2 includes HMI displays, alarm management, historical data logging, and recipe management. The timing requirement is soft real-time (100 ms-1s), which is appropriate for human-operator interfaces.

Question 3: Why is jitter often more critical than absolute latency for synchronized motion control?

a) Jitter is easier to measure than latency b) Coordinated axes need to arrive at positions at exactly the same time, so timing consistency matters more than speed c) Jitter only affects wireless networks d) Lower jitter always means lower latency

Answer

b) Coordinated axes need to arrive at positions at exactly the same time, so timing consistency matters more than speed — In a multi-axis robot, all servo motors must receive their position commands at precisely the same instant. If one axis is consistently 5 ms late but all axes have the same 5 ms delay, the motion is perfectly coordinated. But if one axis varies randomly between 1 ms and 9 ms (high jitter), the axes become unsynchronized, causing mechanical stress, poor surface finish, or collisions.

29.8 Technology Mapping

The following diagram shows how protocols, platforms, and network architectures align with ISA-95 levels:

Read Figure 29.3 to see how the timing promise tightens as you move down the stack.

ISA-95 technology mapping showing physical process and control levels with PLC/RTOS and deterministic protocols, supervisory levels with SCADA and industrial PCs, operations levels with OPC UA and servers, and enterprise levels with REST or MQTT and cloud platforms. — Figure 29.3: ISA-95 technology mapping by level

Start at the bottom of Figure 29.3, where sensors and actuators meet controllers. That interface is measured in microseconds and labelled hard real-time, so the protocol choice there is not a preference. Climb one step and the budget opens to a few milliseconds with a soft real-time promise. Two steps higher, supervisory and operations traffic runs in tens to hundreds of milliseconds on best effort, and the enterprise link at the top is slower still. The pattern is the argument: each level buys reach by giving up timing. The notes add the working rule. Measure the response end to end, and keep control that must not slip at the level that owns it.

Technology choices should follow the level’s deadline, authority, and failure consequence rather than a single preferred protocol.

Technology Mapping Summary

Level 0-1: EtherCAT or PROFINET IRT on PLCs/RTOS for sub-millisecond hard real-time control
Level 2: PROFINET or EtherNet/IP on industrial PCs for supervisory monitoring and HMI workloads
Level 3: OPC-UA or Modbus TCP on servers for MES, historians, and batch coordination
Level 4: REST APIs or MQTT on cloud and enterprise platforms for planning and analytics

29.8.1 Protocol Selection by Level

Level	Typical Protocols	Latency	Determinism
0-1	EtherCAT, PROFINET IRT	<100μs	Hard real-time
1-2	PROFINET, EtherNet/IP	1-10 ms	Soft real-time
2-3	OPC-UA, Modbus TCP	10-100 ms	Best effort
3-4	REST APIs, MQTT	100 ms-1s	Best effort

Level 0-1: EtherCAT or PROFINET IRT, <100μs latency, hard real-time
Level 1-2: PROFINET or EtherNet/IP, 1-10 ms latency, soft real-time
Level 2-3: OPC-UA or Modbus TCP, 10-100 ms latency, best effort
Level 3-4: REST APIs or MQTT, 100 ms-1s latency, best effort

29.8.2 Computing Platform by Level

Level	Platform	OS	Processing
0-1	PLC, PAC, IPC	RTOS, bare metal	Deterministic scan cycle
2	Industrial PC	Windows, Linux	Standard scheduling
3	Server	Windows Server, Linux	Virtualization OK
4	Cloud/Enterprise	Any	Containerization, serverless

Level 0-1: PLC, PAC, or IPC running RTOS or bare metal for deterministic scan cycles
Level 2: Industrial PC running Windows or Linux for standard supervisory scheduling
Level 3: Server hardware running Windows Server or Linux with virtualization
Level 4: Cloud or enterprise platforms using containerized or serverless workloads

29.8.3 Network Architecture by Level

Level	Network	Redundancy	Segmentation
0-1	Dedicated industrial	Ring, dual-port	Air-gapped from IT
2	Industrial Ethernet	RSTP, PRP/HSR	VLAN separated
3	Converged IT/OT	Standard HA	DMZ between zones
4	Corporate/cloud	Internet standards	Firewall protected

Level 0-1: Dedicated industrial networks with ring or dual-port redundancy and air gaps from IT
Level 2: Industrial Ethernet with RSTP or PRP/HSR and VLAN separation
Level 3: Converged IT/OT networks with standard high availability and DMZ zoning
Level 4: Corporate or cloud networks using internet standards and firewall protection

Checkpoint: Technology Placement

You now know: Begin with level 0-1 control can need EtherCAT or PROFINET IRT below 100μs, while Level 1-2 traffic can often use PROFINET or EtherNet/IP in the 1-10 ms range. Next consider oPC-UA fits the Level 2-3 boundary because MES and historian consumers need semantic context, not drive-cycle ownership. Then test rEST APIs and MQTT belong at Level 3-4 integration when the path is planning, analytics, or dashboards rather than machine control.

29.9 ISA-95 Level Selector Tool

Try it: Adjust the response time requirement and safety criticality to see how the recommended ISA-95 level changes. The tool applies the decision framework to match your requirements to the appropriate automation level.

Knowledge Check: Technology Mapping

Question 4: An engineer needs to connect 100 PLCs to a Manufacturing Execution System (MES) for production scheduling. Which protocol and ISA-95 level boundary does this cross?

a) EtherCAT at Level 0-1 boundary b) OPC-UA at Level 2-3 boundary c) PROFINET IRT at Level 1-2 boundary d) MQTT at Level 3-4 boundary

Answer

b) OPC-UA at Level 2-3 boundary — Connecting PLCs (Level 1-2) to an MES (Level 3) crosses the Level 2-3 boundary, which is exactly where OPC-UA excels. OPC-UA provides the semantic data modeling needed for MES integration (production orders, quality data, genealogy), built-in security for the IT/OT boundary, and platform-independent communication.

Question 5: A factory floor network uses air-gapped segmentation. Which ISA-95 level does this practice primarily protect?

a) Level 4 — Enterprise systems need air gaps from the internet b) Level 0-1 — Critical control networks must be physically isolated from IT networks c) Level 3 — MES systems require air gaps from SCADA d) All levels equally use air-gapped segmentation

Answer

b) Level 0-1 — Critical control networks must be physically isolated from IT networks — Air-gapped segmentation is primarily used at Level 0-1, where dedicated industrial networks (EtherCAT, PROFINET IRT) carry safety-critical and hard real-time control traffic. These networks must be physically isolated from IT networks to prevent cyberattacks from reaching PLCs, safety controllers, and field devices. A compromised Level 0-1 network could cause physical damage, equipment destruction, or endanger human life.

29.10 Continue to the Next Part

Carry this evidence into Real-Time ISA-95: System Design and Practice, which begins with Assembly Line Design Case.