106 Real-Time ISA-95: System Design and Practice
106.1 Start With the Decision
An assembly line needs a local stop in milliseconds and a plant report in seconds. The design must keep those paths separate while preserving one auditable data flow.
106.2 Route Overview
This is part 3 of 3. Review Real-Time ISA-95: Determinism and Technology Mapping for the preceding evidence.
106.3 Learning Objectives
- Map assembly-line sensing, control, supervision, and planning to ISA-95.
- Calculate bandwidth and verify end-to-end deadline evidence.
106.4 Chapter Roadmap
- Assembly Line Design Case
- Mobile Figure Summary: Level 0 Data Flow
- Bandwidth Calculator for Case Study
- Checkpoint: Assembly-Line Evidence
- Putting Numbers to It
- Real-Time Systems Basics
- Map Tech to ISA-95 Levels
- Real-Time and ISA-95 Links
- Interactive Quiz: Match ISA-95 Concepts
- Quiz: ISA-95 Design Sequence
- Common Pitfalls
- Not Every Message Is Control
- Respect Layer Boundaries
- 3. Measuring Only Network Latency
- Label the Diagram
- Code Challenge
- Summary
- See Also
- In 60 Seconds
- Knowledge Check
- Quiz: Real-Time ISA-95
- What’s Next
106.5 Assembly Line Design Case
The next scenario puts the timing rules into one design record. Treat each subsystem as a separate row before choosing a protocol.
You are designing the control system for an automotive assembly line with:
- 100 robotic arms: Each with 6 axes (600 servo motors total)
- 1,000 quality inspection sensors: Vision systems, force sensors, laser scanners
- 50 AGVs: Automated guided vehicles delivering parts
- 10 operator stations: HMIs for monitoring and manual intervention
Requirements:
- All robots must be synchronized within 1 ms
- Quality data must be logged for 10-year traceability
- AGVs must avoid collisions with <100 ms response time
- Operators need real-time production status
- Enterprise ERP system needs hourly production counts
- Predictive maintenance for all critical assets
106.5.1 Solution: Latency Requirements
Robotic motion control: <1 ms hard real-time, <1μs jitter for synchronization
- Justification: 6-axis coordinated motion requires deterministic timing
- ISA-95 Level 1 (Basic Control)
Quality sensors: <100 ms soft real-time
- Vision processing and data logging can tolerate slight delays
- ISA-95 Level 2 (Supervisory)
AGV collision avoidance: <100 ms hard real-time
- Safety-critical, must respond to obstacles deterministically
- ISA-95 Level 1 with safety rating
Operator HMIs: <1s soft real-time
- Human perception doesn’t require sub-second updates
- ISA-95 Level 2 (Supervisory)
ERP production counts: Hourly batch updates
- No real-time requirement
- ISA-95 Level 4 (Enterprise)
106.5.2 Solution: Protocol Selection
For robot control: EtherCAT or comparable deterministic motion network
- Sub-millisecond cycle-time capability can support a 1 ms motion requirement when the drive, controller, topology, and load test all meet the timing budget.
- Tight jitter control enables coordinated multi-axis motion when clocks, task scheduling, and diagnostics are validated.
- The release record should include measured cycle time, jitter, fault response, and replacement-device behavior, not only the protocol name.
For quality sensors: PROFINET or EtherNet/IP
- Less stringent timing requirements
- Standard industrial Ethernet sufficient
For AGV coordination: local safety plus supervised fleet coordination
- Obstacle detection and stop behavior should remain local to the vehicle, safety scanner, controller, or certified safety system.
- Wi-Fi, private 5G, OPC-UA, MQTT, or REST can carry fleet state, dispatch, and telemetry only where bounded local safety behavior is preserved.
For IT/OT integration: OPC-UA
- Bridges PLCs to MES/ERP
- Provides semantic data model for quality traceability
106.5.3 Solution: Data Integration Strategy
Pause at Figure 106.1 before carrying solution: data integration strategy forward. Its visual vocabulary joins ISA-95 Level 0: Physical Process Layer to Sensors, which frames isa-95 level 0 physical process layer.
Locate ISA-95 Level 0: Physical Process Layer on Figure 106.1 before checking Sensors. The visual’s third anchor, Temperature, completes isa-95 level 0 physical process layer. Carry ISA-95 Level 0: Physical Process Layer into solution: data integration strategy; use Temperature as its limiting condition.
Level 0 values and commands are closest to the physical process, so stale data and misplaced authority can change machine behavior directly.
106.6 Bandwidth Calculator for Case Study
Try it: Adjust the number of motors, signals per motor, sampling rate, feature-reduction factor, and illustrative transfer price to see how bandwidth assumptions change. Treat the output as a planning comparison, not a current cloud-price quote.
Key architectural decisions:
- Edge processing: FFT, statistical features, and event extraction at Level 2 reduce the data published beyond the control zone.
- Time-series storage: A historian or time-series database handles high-rate process values with timestamp, compression, and quality rules.
- OPC-UA server: Provides a governed namespace for selected values and context.
- Data lake: Raw or summarized data retention follows quality, compliance, and cost requirements.
- Security: Network segmentation, OPC-UA encryption, VLANs separate control and IT networks
Bandwidth calculation:
- 600 motors × 10 signals × 1 ms sampling = 6 million samples/second
- At 4 bytes/sample = 24 MB/second raw data
- After edge processing: 1 Hz features = 6 kB/second to cloud
- Reduction factor: 4,000×
Checkpoint: Assembly-Line Evidence
You now know:
- The case has 100 robotic arms, 600 servo motors, 1,000 quality sensors, 50 AGVs, and 10 operator stations, so one shared timing path would be unsafe.
- Robot synchronization is under 1 ms with less than 1μs jitter, while ERP production counts are hourly batch updates.
- Streaming 600 motors x 10 signals every 1 ms creates 6 million samples/second, so edge reduction keeps raw control-loop data out of enterprise analytics.
An automotive assembly line with 600 servo motors, each reporting 10 signals (position, velocity, current, temperature, etc.) at 1 ms intervals generates massive data volumes:
At 4 bytes per sample (32-bit float):
Edge processing performs FFT and extracts 10 features per motor per second:
The reduction illustrates why raw control-loop data is usually not streamed unchanged to enterprise analytics. The exact cost depends on retention, compression, network path, cloud region, contract terms, and whether the data is stored, transferred, or queried.
What is “real-time”? It means the system must respond within a specific time limit. Think of it like a deadline:
- Hard real-time: Like a goalkeeper — you MUST save the ball in time, or it is a goal. Missing the deadline is a failure (e.g., emergency stop must activate within 10 ms).
- Soft real-time: Like a bus schedule — being 2 minutes late is annoying but acceptable. Missing the deadline degrades quality but does not cause failure.
- Best effort: Like postal mail — it arrives when it arrives, with no guaranteed delivery time.
What is ISA-95? It is a standard that organizes a factory into 5 floors (levels), each with different speed requirements. The bottom floors need lightning-fast responses (controlling machines), while the top floors handle slow business decisions (planning what to build next month).
The key rule: Always use the right tool for the right floor. You would not use a racing car to deliver furniture (too expensive), and you would not use a delivery truck to race in Formula 1 (too slow). Similarly, do not use cloud computing to control a robot arm, and do not use a PLC to run your ERP system.
When designing industrial IoT systems, the biggest mistake is using the wrong technology for the wrong level. This decision framework helps you match computing platforms, protocols, and architectures to automation levels.
| Level | Timing | Example Application | Right Technology | Wrong Technology (Why) |
|---|---|---|---|---|
| 0-1 | <1 ms | Servo motor control | PLC + EtherCAT | Cloud API (200 ms latency — motor crashes before alert) |
| 1-2 | 1-10 ms | Discrete I/O, safety | PLC + PROFINET | Wi-Fi + MQTT (packet loss breaks safety) |
| 2 | 100 ms-1s | HMI displays | Industrial PC + SCADA | Consumer laptop (not rated for factory floor) |
| 3 | Seconds | MES batch tracking | On-prem server + OPC-UA | Edge gateway (insufficient compute for MES) |
| 4 | Minutes-hours | ERP planning | Enterprise cloud | PLC (PLCs don’t run SAP) |
Decision Tree:
Q1: Does a missed deadline cause physical damage or safety risk?
- Yes → Hard real-time (Levels 0-1) → Use PLC/PAC with deterministic network (EtherCAT, PROFINET IRT)
- No → Proceed to Q2
Q2: Is response time under 100 ms required?
- Yes → Soft real-time (Level 1-2) → Use PLC with standard industrial Ethernet (PROFINET IO, EtherNet/IP)
- No → Proceed to Q3
Q3: Does it involve human interaction or visualization?
- Yes → Level 2 (SCADA/HMI) → Industrial PC with SCADA software (Ignition, WinCC)
- No → Proceed to Q4
Q4: Does it coordinate production workflows or track quality?
- Yes → Level 3 (MES) → Server-grade hardware with MES software (Siemens Opcenter, Rockwell FactoryTalk)
- No → Proceed to Q5
Q5: Is it business planning, finance, or supply chain?
- Yes → Level 4 (ERP) → Enterprise cloud or data center (SAP, Oracle)
Real Example — Assembly Line Robot:
- Level 0-1: Servo drives (EtherCAT, <100μs cycle) — controls arm position
- Level 1: Safety PLC (PROFINET IRT, 1 ms) — emergency stop circuit
- Level 2: HMI touchscreen (Ethernet/IP, 100 ms) — operator sees cycle count
- Level 3: MES server (OPC-UA, 10s) — tracks which product variant is running
- Level 4: ERP cloud (REST API, 1 hour) — orders more parts when inventory low
Key Principle: Never use higher-level (slower) tech for lower-level (faster) requirements. The opposite is OK but wasteful — using EtherCAT for HMI updates works but costs 10× more than needed.
- Hard Real-Time → EtherCAT/PROFINET IRT: Deterministic protocols guarantee sub-millisecond response for Level 0-1 safety and motion control.
- ISA-95 Levels → Technology Selection: Each level has distinct timing requirements: Field (μs), Control (ms), Supervisory (100 ms), Operations (seconds), Enterprise (hours).
- Jitter → Multi-Axis Synchronization: Timing variation between coordinated axes is more critical than absolute latency for precision motion.
- OPC-UA → Level 2-3 Boundary: Bridges real-time control (PLCs) to business systems (MES/ERP) with semantic data models and security.
Cross-module connection: Industrial Protocols explains how EtherCAT achieves <100μs cycle times with <1μs jitter using time-triggered communication and distributed clocks for Level 0-1 hard real-time control.
Common Pitfalls
A historian sample, an HMI alarm, an MES work-order update, and a drive command do not have the same deadline or authority. Classify each message by level, owner, deadline, stale-data behavior, and consequence before choosing a protocol.
MES, ERP, dashboards, and analytics systems may request, schedule, or record, but they should not bypass PLC, safety, or cell-level authority. Any write path from Level 3 or Level 4 into Level 0-1 needs explicit ownership, authorization, interlock, and rollback rules.
A real-time claim must include sensor delay, filtering, scan time, task priority, queueing, switch behavior, output update, actuator response, and safe fallback. A fast link can still miss a machine deadline if any queue or scheduler in the path is unbounded.
106.7 Code Challenge
106.8 Summary
Real-time requirements and ISA-95 levels provide a systematic framework for industrial system design:
ISA-95 pyramid: Five levels from field devices (sub-millisecond) to enterprise (hours-days) with distinct timing requirements and technologies appropriate to each level. The standard (IEC 62264) provides the definitive interface model between enterprise and control systems.
Determinism matters: Hard real-time systems require guaranteed worst-case timing with bounded jitter; soft real-time tolerates occasional delays with statistical QoS; best-effort has no timing guarantees. Choosing the wrong class can cause safety incidents (under-specifying) or wasted budget (over-specifying).
Jitter vs latency: For synchronized motion control, consistent timing (low jitter) is often more critical than absolute speed. IEEE 1588 PTP and EtherCAT distributed clocks provide sub-microsecond synchronization for multi-axis coordination.
Technology mapping: Protocol, computing platform, and network architecture choices must align with the timing requirements of each ISA-95 level. Use EtherCAT/PROFINET IRT for Level 0-1, PROFINET/EtherNet/IP for Level 1-2, OPC-UA for Level 2-3 bridging, and REST/MQTT for Level 3-4 cloud integration.
Design principle: Never use higher-level (slower) technologies for lower-level (faster) requirements. Cloud cannot control motors; PLCs should not run ERP. Design from the bottom up — secure the hardest real-time constraint first, then relax requirements as you ascend the pyramid.
106.9 See Also
- Industrial Protocols — Comparison of EtherCAT, PROFINET IRT, EtherNet/IP with latency and jitter specifications for each ISA-95 level
- Real-Time Operating Systems — VxWorks, QNX, RTAI architecture for guaranteed worst-case execution time (WCET) at Level 0-1
- Time-Sensitive Networking — IEEE 802.1 TSN extensions for deterministic Ethernet suitable for industrial control
- OPC-UA Standard — Level 2-3 integration protocol that bridges real-time control to business systems with semantic data models
106.10 Knowledge Check
106.11 What’s Next
| Direction | Chapter | Description |
|---|---|---|
| Next | Predictive Maintenance | Condition monitoring and ML-based failure prediction |
| Related | OPC-UA Standard | The unifying standard for industrial interoperability |
| Related | Industrial Protocols | Protocol selection for different applications |
| Index | Industry 4.0 Fundamentals | Overview of all IIoT topics |
- Next: Predictive Maintenance — Condition monitoring and ML-based failure prediction
- Related: OPC-UA Standard — The unifying standard for industrial interoperability
- Related: Industrial Protocols — Protocol selection for different applications
- Index: Industry 4.0 Fundamentals — Overview of all IIoT topics
106.12 Continue Your Route
This final part closes the route from Assembly Line Design Case through What’s Next. Return to Real-Time ISA-95: Determinism and Technology Mapping or continue from the applications module index.