What Is PROFINET?

Back to Blog

How PROFINET Achieves Real-Time on Ethernet

Standard TCP/IP is not deterministic - packets can be delayed by network congestion. PROFINET solves this with three communication channels layered on ordinary Ethernet. Non-time-critical data (configuration, diagnostics, web access) uses standard TCP/IP or UDP/IP. Real-Time (RT) traffic bypasses the TCP/IP stack and uses prioritized Ethernet frames for cycle times in the low milliseconds, sufficient for most process and factory I/O.

Isochronous Real-Time (IRT) goes further, using hardware time-slicing and scheduling in special switch ASICs to reserve bandwidth and synchronize devices, achieving sub-microsecond jitter for demanding motion control such as coordinated axes or high-speed packaging. Most oil and gas process applications use RT; IRT is reserved for machinery that needs tight mechanical synchronization.

PROFINET in a SCADA Architecture

Because PROFINET runs on standard Ethernet, it integrates cleanly with the rest of a plant network and with SCADA. A PROFINET IO controller (typically a PLC) exchanges cyclic data with distributed IO devices, drives, and instruments; the SCADA server then reads process values from that controller, most often over OPC UA or Modbus TCP, rather than participating in the PROFINET cyclic exchange itself.

For cloud-native SCADA, this separation is convenient: the deterministic control network stays local and isolated, while an edge gateway or the controller exposes a curated set of tags over a routable protocol. Merobix, for example, reads those tags over OPC UA or Modbus TCP and streams them outbound to the cloud, so a PROFINET-based plant gains remote dashboards, alarms, and history without altering the real-time control layer or opening the OT network to inbound traffic.

Conformance Classes: What to Put on the Spec Sheet

PROFINET bundles device and network capabilities into conformance classes, and they are the cleanest way to state what a project actually requires. CC-A covers basic cyclic RT exchange over standard Ethernet cabling and permits wireless media. CC-B adds network diagnostics and topology detection through SNMP and LLDP, which is what lets maintenance locate a failed device by switch and port instead of walking the cable tray. CC-C adds the hardware-supported bandwidth reservation and clock synchronization that IRT requires, and CC-D covers the newer TSN-based implementations. The conformance class you write into the specification quietly decides which devices, switches, and even cables qualify for the project, so settle it before the first purchase order, not after.

Beyond the class, the practical selection checklist is short. Confirm a current GSDML file is available for your engineering tool. Decide whether the network needs media redundancy such as an MRP ring, because retrofitting a ring into a line topology later means touching every switch. And decide whether the controller connection needs system redundancy - the S2 and R1 schemes exist for plants that cannot tolerate a gap during a controller failover. None of these choices change the protocol; they change which hardware belongs on the order.

Commissioning: Names First, Then Addresses

A PROFINET IO device is identified by its station name, not its IP address, and commissioning follows from that. The sequence is stable across tools: import each device's GSDML file, build the IO configuration and the expected topology, assign station names to the physical devices, and download - the controller then hands out IP addresses at startup using the discovery and configuration protocol. Getting the naming step wrong is the classic first-start failure: a device with a blank, mistyped, or duplicated name simply never joins the cyclic exchange, and the controller reports a missing station even though the hardware is powered and linked.

Plan for device replacement while you are still commissioning. If the expected topology has been configured, a controller can recognize a factory-fresh replacement by its neighbor relationships and assign it the correct name automatically, so a technician can swap hardware without a laptop. If topology has not been configured, every future swap becomes an engineering-tool visit. It is a decision made once, during commissioning, that shapes years of maintenance callouts.

Failure Modes and How the Diagnostics Model Helps

PROFINET devices report faults through a structured diagnostics model: a failing channel raises a diagnostic alarm that travels to the controller with the module, submodule, and channel identified, so the fault arrives already located. Each cyclic connection is also supervised by a watchdog - if a device misses more than the allowed number of update cycles, the connection is dropped and the controller flags a station failure rather than acting on stale data. That fail-silent behavior is a feature: the absence of data is announced instead of hidden.

The recurring field problems are worth knowing in advance: duplicated or mismatched station names after a swap, a damaged patch cord or port producing intermittent frame loss, a switch configuration that strips priority tagging so RT frames queue behind bulk traffic, and EMC coupling on runs near drives. Intermittent frame loss shows up as sporadic station failures long before it shows up as bad process data, so when a network drops cycles on and off, work through a structured approach to diagnosing jitter and dropped cycles instead of swapping hardware on suspicion.

The Three Traffic Classes at a Glance

Channel	Transport	What it carries
NRT (non-real-time)	Standard TCP/IP or UDP/IP	Configuration, parameter reads, diagnostics, web and file access
RT (real-time)	Prioritized Ethernet frames, bypassing the TCP/IP stack	Cyclic process I/O for most machine and process control
IRT (isochronous real-time)	Hardware-scheduled time slots in IRT-capable switches	Clock-synchronized motion and tightly coordinated axes

One consequence of this layering deserves emphasis: because the NRT channel is ordinary IP, every PROFINET device is also an IP host, and the usual OT rules about segmentation, firewalling, and exposure apply to it exactly as they do to any other industrial Ethernet node. Real-time behavior does not exempt a network from security design - it just raises the cost of getting that design wrong.

Frequently Asked Questions