03 — PROFINET
Industrial Ethernet built for the factory floor.
Estimated 30 minutes.
What is Profinet?
Profinet is the Siemens-led industrial Ethernet standard. It is to the European factory floor what EtherNet/IP is to North America: the dominant choice for connecting PLCs, remote I/O, drives, and sensors over a single Ethernet network.
Standard Ethernet, mostly.
Profinet uses ordinary Ethernet cables (Cat5e+), ordinary RJ45 connectors, and ordinary switches — at least at the lowest performance class. The "industrial" parts come from how Profinet uses Ethernet, not from custom hardware.
But not always TCP/IP.
For high-frequency I/O, Profinet skips TCP and even IP. Real-time frames go straight on top of Layer 2 Ethernet using their own Ethertype (0x8892). That is the trick that gets cycle times down into milliseconds — and microseconds — on standard wire.
Three performance classes.
Profinet has three real-time classes: NRT (best effort), RT (Class 1, milliseconds), and IRT (Class 3, microseconds with hardware sync). A fourth — TSN — is the modern open replacement for IRT. You will see all four on real plants.
Compare the four classes.
Tap each to see cycle time, jitter, hardware needs and where it is used.
Which class for what?
You are connecting eight servo drives that must move in synchronised coordination for a high-speed printing press. Which Profinet class do you specify?
- NRT — TCP/IP is fine.
- RT — milliseconds is fast enough.
- IRT — synchronised motion needs sub-microsecond jitter.
- It does not matter, Profinet is Profinet.
Two kinds of traffic.
Inside any Profinet device, two completely different data flows happen at the same time: cyclic data (the I/O image — sent every cycle, push) and acyclic data (parameters, diagnostics, alarms — request-response, on-demand).
Share the wire — cyclic + acyclic.
Set the cyclic interval, fire acyclic events. Watch the two flows coexist on a 1-second timeline.
Why two modes?
Cyclic data is small and frequent — there is no time for a request, so the device just sends it. Acyclic data is large and rare, so a TCP/UDP request-response is fine. Mixing them on the same wire is exactly what Profinet was designed to do.
Device names, not IPs.
In Profinet, every device has a human-readable name (e.g. "io-station-east") that the PLC uses to find it. The IP address is assigned dynamically by the controller — you almost never type IPs by hand.
DCP = Discovery & Configuration.
Profinet uses a Layer 2 protocol called DCP (Discovery and Configuration Protocol) to find devices, set their names, and assign IPs. DCP runs at Layer 2, so it works even on devices that have never been given an IP.
Run a DCP scan.
A DCP "Identify-All" broadcast asks every Profinet device on the segment to announce itself. Tap to see what comes back.
Why names instead of IPs?
If a drive fails, you replace it with a brand-new one out of the spares box. The new drive has no IP and no configuration. You assign it the failed drive's name over DCP, and the controller pushes the configuration and IP automatically. No engineering laptop needed for the swap.
GSDML: the device datasheet.
Every Profinet device ships with a GSDML file — an XML document that describes everything the device can do: its modules, its parameters, its diagnostic codes. The engineering tool reads the GSDML to know how to configure the device.
Configure an ET 200SP from the GSDML.
Drag head + I/O modules from the GSDML catalogue into the rack. Wrong slot or undeclared module = TIA Portal rejects it.
No GSDML, no commissioning.
If a vendor has not given you the GSDML, you cannot add the device to your project. Always check the vendor support site — or look in the box — for the current version. Mismatched GSDML versions are a common reason a device "appears online but does not exchange data."
Controller, device, supervisor.
Profinet uses three roles. Controller — the PLC, runs the cyclic exchange. Device — the I/O block, drive, sensor, etc. Supervisor — the engineering tool (TIA Portal, etc.) that commissions and diagnoses. Multiple controllers on one network is allowed; multiple supervisors is normal.
A device is not exchanging data.
A new ET 200SP I/O station is plugged in, link lights are green, DCP discovers it — but the PLC reports "module fault" and no I/O is exchanged. What's the most likely cause?
- Bad Ethernet cable.
- Wrong device name — the PLC is looking for a name the device does not have.
- IRT licence missing on the PLC.
- TCP port 502 is firewalled.
Gotcha #1: Topology matters.
For RT, any switch works. For IRT, every switch in the cyclic path must be IRT-capable and configured into the topology — otherwise the timing breaks. Mixing a non-IRT consumer switch into an IRT line is a classic mistake.
Gotcha #2: Update time.
Each device has a configured update time (e.g. 4 ms). If too many devices are configured too aggressively, you will run out of bandwidth on the controller. The engineering tool will warn you — pay attention to that warning.
Gotcha #3: Plaintext, again.
Like Modbus TCP, Profinet has no native authentication or encryption. Keep it on an isolated OT network, behind a firewall, away from the office LAN. Profinet Security extensions exist but are rarely deployed.
You now know Profinet.
- Standard Ethernet wires; non-standard use of them.
- Three classes: NRT (best effort), RT (ms), IRT (µs hardware-sync). TSN is the open future.
- Two parallel data flows: cyclic (push, every cycle) and acyclic (request-response).
- Devices are addressed by name. DCP discovers them and assigns IPs.
- GSDML is the XML "datasheet" the engineering tool needs to commission a device.
Profinet complete.
- You can pick the right class for a given application.
- You know why Profinet uses names instead of IPs.
- You understand cyclic vs acyclic and where each lives.
- You can debug "online but no data" — start with the device name.