Choosing a fieldbus for a new line usually comes down to one question: how much timing precision does the process actually need. Modbus TCP, EtherNet/IP and Profinet all move data over standard Ethernet, but they handle timing, addressing and hardware requirements in different ways.
Which fieldbus should you choose for a new industrial line?
Short answer: Modbus TCP for simple I/O and monitoring where a few milliseconds of jitter do not matter, EtherNet/IP if the plant already runs on Rockwell/Allen-Bradley hardware, Profinet if it runs on Siemens hardware or the process needs hard real time control such as motion or drive synchronization.
| Modbus TCP | EtherNet/IP | Profinet | |
|---|---|---|---|
| Determinism | Best effort, depends on network load | Predictable via cyclic I/O messaging | Sub-millisecond with IRT on certified hardware |
| Hardware needed | Any standard managed switch | Standard Ethernet, DLR switches for redundancy | Profinet-certified switches for IRT |
| Typical ecosystem | Open, vendor-neutral | Rockwell/Allen-Bradley | Siemens |
| Best fit | Monitoring, simple I/O, retrofit reporting | Machine control on Rockwell-based lines | Motion control, drive synchronization |
Modbus TCP trades determinism for simplicity
Modbus TCP is a request/response protocol: a client polls a server, the server replies. No native real time channel, no built in redundancy, timing depends entirely on the underlying TCP/IP stack and network load. What it offers instead is simplicity: a well documented, vendor-neutral protocol that almost any PLC, sensor or gateway can speak with a small software library, over any standard managed switch.
EtherNet/IP layers real time I/O over standard Ethernet
EtherNet/IP runs the Common Industrial Protocol (CIP) over standard Ethernet frames. It supports two message types: explicit messaging for configuration and diagnostics, and implicit (I/O) messaging for cyclic, time critical data. Implicit messaging uses UDP with a configured request packet interval, giving predictable update rates without needing special hardware. Device Level Ring (DLR) topology adds network redundancy. It is the default choice in Rockwell-centric plants because the entire ecosystem, drives, safety relays, HMIs, already speaks it.
Profinet reaches determinism through hardware level timing
Profinet defines three performance classes. RT (real time) runs on standard Ethernet hardware with prioritized frames. IRT (isochronous real time) needs Profinet-certified switches with hardware timing support (ASICs) to guarantee sub-millisecond, jitter-free cycles, the class used for motion control and drive synchronization. Profinet networks are also topology-aware: engineering tools use GSD files and a defined ring or line topology, which makes commissioning more rigid but diagnostics more precise.
SCADA and MES integration rarely use the same protocol as the machine network
Very few plants run a single protocol end to end. A common pattern: Profinet or EtherNet/IP on the machine network for real time control, Modbus TCP or MQTT one layer up for SCADA, historian and MES integration, because that layer does not need microsecond timing, it needs broad compatibility. Gateways that translate between the two, reading Profinet or EtherNet/IP on one side and exposing Modbus TCP or MQTT on the other, are usually where a retrofit or IIoT project actually starts, without touching the machine network itself.
Retrofitting an existing line: Modbus TCP is the fastest way to add an open source based PLC
An Arduino-based or Raspberry Pi-based PLC with onboard Ethernet, an M-Duino with an Ethernet port, for example, speaks Modbus TCP natively through open source software libraries, which makes it the fastest way to pull data out of a line regardless of what the machine network itself runs on. For lines already running EtherNet/IP or Profinet, a gateway such as GateBerry sits between the machine network and the reporting layer, so the retrofit adds a Modbus TCP or MQTT source without requiring a native EtherNet/IP or Profinet stack on the controller. For a new line built from scratch, the fieldbus choice should still follow the existing hardware ecosystem; for reading data off machines that already run something else, this is the pragmatic entry point.