What is real-time?
Real-time is the ability of a system to respond to an event or input within a predefined and guaranteed time, where a late answer is just as wrong as an incorrect one. So it does not necessarily mean “fast”, but rather predictable within a time limit, however small or large that limit may be. In industrial automation, real-time is the foundation of every PLC, every motion control application and every safety function.
🧠 Why does real-time matter in OT?
In industrial environments, processes are often time-critical. The control system works in a closed loop: measure, compute and actuate, over and over again.
- Stopping on a fault — a motor must stop immediately when a sensor detects a fault or an emergency stop is pressed
- Synchronised motion — a robot arm must grab a product from a moving conveyor at exactly the right moment
- Stable control — a PID controller can start to oscillate or become unstable when measurements arrive irregularly or late
- Cyclic processing — a control system reads Sensor data every cycle, computes and drives outputs
If the response arrives too late, this can lead to production errors, machine damage or unsafe situations for people. That is why real-time is a functional requirement in OT, not a performance feature.
🔧 What types of real-time systems are there?
| Type | Description |
|---|---|
| Hard real-time | Systems where every deadline absolutely must be met; a missed deadline is a system failure. e.g.: Safety systems (SIS), Motion Control. |
| Firm real-time | A late result is worthless and discarded, but an occasional miss causes no harm. e.g.: visual inspection that lets a product pass by. |
| Soft real-time | Deadlines are important, but occasional lateness is acceptable; quality degrades. e.g.: HMI updates, video streaming. |
| Non real-time | There is no guarantee of timeliness. e.g.: email, administration, ERP. |
Strictly speaking, “non real-time” is not a real-time class but the absence of a timing guarantee. The class depends on the consequence of a missed deadline, not on how short that deadline is.
⏱️ What makes a system deterministic?
Real-time requires deterministic behaviour: the same system responds within the same bounded time under all conditions. Four concepts are decisive:
- Latency — the delay between an event and the response
- Jitter — the variation in that delay from cycle to cycle; for motion control often more important than the latency itself
- Cycle time — the fixed period in which a controller reads, computes and writes
- Worst-case execution time (WCET) — the longest possible execution time of a task; real-time design works with the worst case, not the average
An RTOS (real-time operating system) such as VxWorks, QNX or FreeRTOS guarantees that high-priority tasks can interrupt running tasks within a bounded time (preemptive scheduling). A PLC executes its program according to the task model of IEC 61131-3 with cyclic tasks; a vendor watchdog puts the controller into stop with defined outputs when the maximum cycle time is exceeded. Since Linux kernel 6.12 (late 2024), the PREEMPT_RT real-time extension has been part of the mainline kernel, making soft PLCs and edge platforms on Linux easier to build.
🌐 How does real-time work in industrial communication?
Standard Ethernet is not deterministic: frames can be delayed in switch queues behind other traffic or dropped under overload, and there is no upper bound on that delay. Industrial protocols solve this in different ways:
| Technology | Approach | Typical cycle time | Jitter |
|---|---|---|---|
| ProfiNET RT | Priority for process data, bypassing the TCP/IP stack | 250 µs – 512 ms, usually 1–16 ms | A few µs to ms |
| ProfiNET IRT | Reserved time slot per cycle, dedicated switches (Conformance Class C) | 250 µs, down to 31.25 µs | < 1 µs |
| EtherCAT | A single frame passes all slaves “on the fly”, with distributed clocks | 100 servo axes in 100 µs | Well below 1 µs |
| TSN | IEEE 802.1 extensions to standard Ethernet: time synchronisation and time-scheduled queues | Configurable, µs range | < 1 µs achievable |
ProfiNET RT and IRT are maintained by PROFIBUS & PROFINET International; EtherCAT was introduced by Beckhoff in 2003 and has been part of IEC 61158 since 2007. TSN is the work of the IEEE 802.1 TSN task group, the successor of Audio Video Bridging. Its core building blocks are IEEE 802.1AS for time synchronisation (based on the Precision Time Protocol) and IEEE 802.1Qbv from 2015, the time-aware shaper. In June 2026, IEC/IEEE 60802 was published, the TSN profile for industrial automation. ProfiNET over TSN is called Conformance Class D.
Safety protocols such as PROFIsafe add a watchdog time on top: if a safety message does not arrive in time, the installation goes to its safe state.
🏭 Which cycle times are common in practice?
| Application | Type of real-time | Typical cycle time | Explanation |
|---|---|---|---|
| Servo and motion control | Hard real-time | 31.25 µs – 1 ms | Axes must run in sync, jitter matters |
| PLC motor control | Hard real-time | 1 – 20 ms | Must switch at exactly the right moment |
| Process control in a DCS | Hard real-time | 100 ms – 1 s | Temperature and pressure change more slowly |
| HMI visualisation | Soft real-time | 0.5 – 2 s | A small delay is acceptable |
| Historian data logging | Soft real-time | 1 s – 1 min | A small delay or buffer time is acceptable |
| MES | Non real-time | Minutes to hours | Works in shifts, hours and minutes |
| ERP system | Non real-time | Hours to days | Planning data does not need immediate processing |
These time scales match ISA-95: level 4 plans in months, weeks and days, level 3 works in days, shifts, hours and minutes, and levels 0 to 2 respond in seconds and milliseconds.
📊 What does real-time mean in an IT and data context?
In IT, marketing and data analytics, “real-time” usually means something else: near real-time. A dashboard that refreshes every few seconds, a Unified Namespace that forwards values via MQTT, or a streaming platform such as Apache Kafka is fast, but offers no guaranteed deadline. A delay of a few seconds is not a failure there.
| Aspect | Control real-time (OT) | Near real-time (IT/data) |
|---|---|---|
| Time scale | µs to ms | Seconds to minutes |
| Guarantee | Bounded worst-case response time | Fast on average, no guarantee |
| Consequence of lateness | Failure, damage or unsafe situation | Stale screen or report |
| Technology | RTOS, PLC, IRT, EtherCAT, TSN | Cloud, databases, message brokers |
In OT convergence, this difference often causes misunderstandings. An IT team promising “real-time data” rarely means the same requirement as a control engineer.
🧭 How do you determine the real-time requirements in a project?
- Determine the consequence of lateness — does a missed deadline lead to danger, product loss or just inconvenience? That decides hard, firm or soft.
- Derive the deadline from the process — the mechanics, process dynamics or safety distance determine the required response time, not the vendor’s preference.
- Calculate the whole chain — sensor, input filter, network, PLC cycle, output and actuator all add up.
- Choose the right technology — ProfiNET RT is sufficient for most machines; IRT, EtherCAT or TSN for synchronised axes.
- Separate critical and non-critical traffic — keep camera streams and data exchange with IT away from the control cycle.
- Measure in practice — check cycle time, watchdog events and jitter during the FAT and under full load.
An example from safety engineering: under ISO 13855, the total stopping time of a machine determines how far a light curtain must be placed from the hazard. With an approach speed of 2,000 mm/s, every additional 10 ms of response time means 20 mm of extra distance. A slow or unpredictable safety PLC literally makes the machine bigger.
🔄 What are the biggest misconceptions about real-time?
Real-time does not always mean “faster” than other systems. It means: always on time, regardless of whether that is 1 millisecond or 1 second, as long as it is predictable.
- “Fast hardware is real-time” — a fast server with an ordinary operating system can occasionally stall for hundreds of milliseconds
- “Fast on average is good enough” — real-time design is about the worst case, not the average
- “Gigabit Ethernet is deterministic” — more bandwidth shortens queues, but guarantees no upper bound
- “Real-time dashboards are real-time” — in OT terms they are near real-time
- “The cloud can do control” — control of physical processes stays local; the cloud is for analysis
❓ Frequently asked questions
Is a PLC a real-time system?
A PLC is a hard real-time system: its program runs cyclically within a fixed cycle time, usually between 1 and 20 ms. A watchdog monitors that cycle time and puts the PLC into a safe state if a cycle takes too long.
Is standard Ethernet suitable for real-time communication?
Standard Ethernet with TCP/IP gives no guaranteed upper bound on delay and is therefore not deterministic. Real-time industrial Ethernet such as ProfiNET RT and IRT, EtherCAT or Ethernet with TSN extensions adds priority, time slots and time synchronisation to communicate in real time.
Can Linux or Windows be real-time?
A standard Windows or Linux installation is not real-time, because tasks can wait for an unpredictable time. Linux with PREEMPT_RT, part of the mainline kernel since version 6.12, provides real-time scheduling. Soft PLCs on Windows usually run their own real-time kernel alongside Windows.
What is jitter and why does it matter for real-time?
Jitter is the variation in delay or cycle time from one cycle to the next. In real-time motion control, jitter causes axes to drift out of sync, even when the average delay is small. That is why IRT, EtherCAT and TSN achieve a jitter of less than 1 microsecond.
Can real-time control run in the cloud?
Real-time control cannot run reliably in the cloud, because the delay and availability of an internet connection are not guaranteed. The cloud is suitable for near real-time analysis and reporting, while control stays local in the PLC, DCS or edge systems.
What is the difference between hard and soft real-time?
In hard real-time, a missed deadline is a system failure that can lead to damage or danger, as in motion control or a safety system. In soft real-time, a late result only lowers quality, such as an HMI screen that briefly lags.
📌 In summary
Real-time systems respond within a predetermined, guaranteed time: predictability matters more than speed. In industrial automation, RTOSs, PLC cycles and deterministic networks such as ProfiNET IRT, EtherCAT and TSN make this possible, and that is crucial for guaranteeing Safety, quality and performance.