PROFINET divides time into repeating send cycles. Standard RT traffic just queues onto shared Ethernet — fast, but it jostles with everything else, so arrival times wobble. IRT carves out a reserved, scheduled slot at the very start of every cycle that nothing else may touch — buying the rock-steady timing that synchronised motion needs. Watch the two side by side.
Each grey band is one send cycle. The coloured blocks are device frames landing on the wire. In IRT mode the green frames lock to the start of every cycle — identical position, cycle after cycle (zero jitter). In RT mode the blue frames drift around inside the cycle as they contend with other traffic. The jitter meter tracks the spread.
Hit 🌩️ Inject network load and watch what it does to each mode: RT frames scatter further (their timing degrades as the wire gets busy), while IRT frames don't budge — their slot is reserved by schedule, immune to other traffic. That immunity is the entire point of IRT, and why it's reserved for motion and the tightest loops.
Cyclic data only flows after a startup handshake. The pieces:
📄 GSDML — the XML file describing a device's modules, data and capabilities. You import it so TIA knows what the device is.
🏷️ Device name + DCP — at startup the controller uses the DCP protocol to find the device by name and hand it its configured IP. No name, no connection.
🔁 Provider/consumer — once connected, each side "provides" its data and "consumes" the other's, every send cycle, with a watchdog on freshness.
🗺️ Topology — IRT needs to know which port connects to which (the planned topology) so it can schedule the reserved phase across the line.
You're synchronising 8 servo drives for a printing registration — they must all sample within microseconds of each other, every cycle. Which class?