29 Real-Time ISA-95: Determinism and Technology Mapping
More
Depth: Practitioner
Progress saves as you read. Mark sections complete as you go; your progress appears in the sidebar.
- Podcast audio coming soon
- Character narration audio coming soon
29.1 Give Each Factory Path the Right Promise
A fast average link can still miss a control deadline when timing varies. This second chapter of three explains how deadline consequences, jitter, and clock synchronization guide technology choices across ISA-95. The Timing Foundations chapter established the level timing ladder; here we distinguish hard, firm, and soft deadlines and use that evidence to place technologies.
Keep the existing factory example in view: a servo motor needs a 250-microsecond cycle, a SCADA HMI refreshes every 500 ms, and a cloud dashboard updates every 30 seconds. Supervisory control and data acquisition (SCADA) presents equipment conditions and alarms; its human–machine interface (HMI) lets operators see them. These three paths serve different jobs. A dashboard update cannot establish that a servo command arrived on time.
The problem is choosing a timing promise for each job before choosing its protocol. We will first classify missed deadlines, then examine arrival variation and synchronization, and finally map protocols, computing platforms, and network boundaries to their duties.
KEY CONCEPTS
- Deadline consequence
A missed deadline can fail control or degrade usefulness.
- Timing variation
The same mean delay can hide very different arrivals.
- Shared timing
Coordinated axes need consistent command timing.
- Technology placement
Match each path to its timing duty and authority.
29.2 Learning Objectives
- Distinguish hard, firm, and soft real-time deadlines.
- Map jitter, synchronization, and network technology to ISA-95 levels.
29.3 Decide What a Missed Deadline Means
Determinism means predictable, bounded timing. Throughput describes how much work a system handles; it cannot establish the worst-case response of the servo. The deadline class instead describes what happens when a result arrives late.
In hard real-time, missing the guaranteed deadline is a system failure. Worst-case execution time (WCET), the longest execution time for the task, must be bounded, and the complete response path must fit the deadline. Safety systems such as emergency stops, coordinated-axis motion control, and process control of exothermic reactions illustrate why that guarantee matters.
Hard real-time implementations commonly use preemptive, priority-based scheduling and may need specialized hardware. Options include EtherCAT or PROFINET IRT networks, real-time operating systems (RTOS) such as VxWorks, QNX, or RTAI, FPGA-based control, and time-triggered architectures. Synchronized motion can require jitter below 1 μs; the application sets the actual limit.
The upper timeline shows even intervals with a bound at each deadline. The lower timeline stretches and shrinks around an average. In soft real-time, occasional misses degrade performance rather than constitute system failure. HMI updates, data logging, trend analysis, and operator notifications can use this model. Standard operating systems may be acceptable, with statistical quality of service evaluated under expected load and jitter; an average alone does not describe all delays.
A firm deadline makes a late result unusable, although an occasional miss may be tolerated. Best effort offers no timing guarantee at all: work completes when resources are available. Historical analysis, business reporting, email notifications, and non-critical analytics fit best-effort paths. Keep those paths separate from the servo’s bounded control duty.
29.4 Preserve Coordination Despite Delay
For coordinated motion, jitter, variation in arrival timing, can matter more than absolute latency. A consistent delay can be scheduled around; unpredictable variation can leave axes at different positions. Jitter is neither mean latency nor packet loss.
The upper timeline’s five arrivals stay near their expected times, keeping the axes aligned. The lower timeline has the same average but includes a 32 ms arrival beyond the deadline. Its coordination panel shows the resulting positional mismatch. Reducing average delay would not, by itself, establish consistent command arrival.
The application examples below pair cycle time with a separate jitter limit and node count. They are design examples, rather than universal requirements for every machine in each category.
| Application | Cycle time | Maximum jitter | Nodes |
|---|---|---|---|
| Simple I/O | 10 ms | 1 ms | 10–100 |
| Process control | 100 ms | 10 ms | 100–1000 |
| Packaging machinery | 1 ms | 100 μs | 10–50 |
| Printing press | 125 μs | 1 μs | 20–100 |
| Semiconductor handling | 62.5 μs | 100 ns | 10–30 |
Simple I/O tolerates milliseconds of variation in this example, while printing needs microsecond precision and semiconductor handling needs sub-microsecond precision. At a 125 μs printing cycle, the 1 μs budget protects registration; occasional slow cycles can smear or misregister colors.
Synchronization aligns clocks so distributed devices can coordinate. Institute of Electrical and Electronics Engineers (IEEE) 1588 Precision Time Protocol (PTP) supports sub-microsecond synchronization in suitable implementations. EtherCAT distributed clocks provide hardware-based synchronization. Time-triggered protocols schedule messages deterministically, while GPS timing supplies an absolute reference for wide-area systems. Clock alignment and predictable delivery both matter to the servo path.
29.5 Jitter Impact Calculator
Adjust actual jitter for each application and compare it with the cycle and jitter budgets.
Open Jitter Impact Calculator29.6 Map Technologies to Their Duties
With deadline consequence and variation separated, return to the factory’s three paths. EtherCAT is a candidate for the 250 μs servo cycle at Level 0–1; PROFINET can support supervisory factory traffic for the 500 ms SCADA display; MQTT fits best-effort distribution for the 30-second cloud dashboard. Each choice still needs validation of its complete response path. A protocol name alone cannot prove the servo deadline.
Read the figure upward from sensors and actuators to controllers, supervision, operations, and enterprise planning. Its lower interface is measured in microseconds, followed by a few milliseconds, tens to hundreds of milliseconds, and slower enterprise exchanges. Measure response end to end and keep control that must not slip with the function that owns it.
| Level boundary | Typical protocols | Illustrative latency | Timing role |
|---|---|---|---|
| 0–1 | EtherCAT, PROFINET IRT | <100 μs | 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–1 s | Best effort |
These are selection examples, not ISA-95 mandates or fixed capabilities of every implementation. Level 0 is the physical process; Level 1 performs basic control with programmable logic controllers (PLCs) and controllers. Level 2 supervises equipment and presents alarms to operators. Level 3 manages manufacturing operations through a Manufacturing Execution System (MES), historians, and batch coordination. Level 4 supports enterprise planning and analytics.
For the example connecting 100 PLCs to MES production scheduling, OPC UA provides structured interoperability through the intervening Level 2–3 supervisory boundary. Its semantic context serves MES and historian consumers; that exchange does not own the drive cycle. REST APIs and MQTT support Level 3–4 planning, analytics, and dashboards when the control boundary remains protected.
Protocol selection also needs a platform and network design. A programmable automation controller (PAC) or industrial PC (IPC), like a PLC, can host low-level control when its timing is validated.
| Level | Platform | Operating system | 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 acceptable |
| 4 | Cloud/enterprise | Any | Containerization, serverless |
The network examples pair redundancy with segmentation. An air gap physically separates networks; a demilitarized zone (DMZ) separates zones through an intermediate network.
| Level | Network | Redundancy | Segmentation example |
|---|---|---|---|
| 0–1 | Dedicated industrial | Ring, dual-port | Air-gapped from IT |
| 2 | Industrial Ethernet | RSTP, PRP/HSR | VLAN separation |
| 3 | Converged IT/OT | Standard high availability | DMZ between zones |
| 4 | Corporate/cloud | Internet standards | Firewall protection |
An air gap is one low-level isolation option, not an ISA-95 requirement. Where integration is needed, the chosen boundary must protect critical control from IT disruption. Neither redundancy nor segmentation alone establishes a timing guarantee.
29.7 ISA-95 Level Selector Tool
Vary response time, missed-deadline impact, and operator interaction; review the suggested level against the function's duty.
Open ISA-95 Level Selector ToolSummary
Hard real-time bounds the response because a missed deadline is failure. Firm deadlines make late results unusable; soft real-time tolerates some degradation, and best effort promises no completion time. Jitter describes variation, while synchronization aligns distributed clocks.
The factory’s servo, SCADA display, and cloud dashboard need different promises. Use deadline, authority, and failure consequence to select the protocol, platform, and network boundary, then validate the complete path.
Choose the timing promise before the technology. A fast average cannot prove a bounded deadline or synchronized motion.
Carry these decisions into Real-Time ISA-95: System Design and Practice, beginning with the Assembly Line Design Case.

