You order a remote I/O module for a machine upgrade. It arrives with two RJ45 ports, a DIN-rail clip and a datasheet that says Modbus TCP. You wire it to an S7-1200, open TIA Portal, and nothing appears in the hardware catalog. The PLC never sees a device.
The module is not faulty. It is simply not PROFINET.
The confusion is common because the two terms are used as if they were alternatives. They are not. Remote I/O answers where the I/O lives. PROFINET IO answers how the data travels. One is an architecture, the other is a protocol standard — and a single device is often both at once.
The short answer
| Remote I/O | PROFINET IO | |
|---|---|---|
| What it is | A deployment architecture and product form | A communication standard |
| Defined by | Each manufacturer | PI (PROFIBUS & PROFINET International), IEC 61158 / 61784 |
| Bound to one protocol | No — Modbus, EtherCAT, CANopen, PROFINET, wireless | Yes — only inside a PROFINET network |
| What you buy | A field device with terminals | A capability that device must implement |
| Fails when | Wrong protocol version ordered | Device lacks GSDML or conformance |
A PROFINET remote I/O is the intersection: the physical form of the first, the communication behavior of the second. Everything below explains why that distinction decides whether your project runs in an afternoon or a week.
What is remote I/O?
Remote I/O is input/output hardware installed near the sensors and actuators it serves, connected to the controller over a network cable instead of individual wires run back to the control cabinet. It is also called distributed I/O.
The reason it exists is arithmetic. A 4–20 mA transmitter 80 meters from the cabinet needs a shielded pair back to a marshalling terminal; forty of them need a cable tray, cabinet space and a week of labor. Move the terminals out to the machine and one Ethernet run replaces all of it. Analog signals also stop traveling long distances, which removes a whole class of noise and drift problems.
A remote I/O station has two layers:
- A head or bus coupler — the network adapter that talks to the controller.
- I/O slices — DIN-rail modules wired to the field: digital inputs for switches and proximity sensors, relay or transistor outputs for valves and contactors, analog inputs for 4–20 mA and 0–10 V, plus temperature, counter and pulse modules.
To the control program, a remote input looks exactly like one wired into the controller's own rack. That transparency is the whole point.
The critical detail: the protocol is chosen by the controller, not by the I/O. Remote I/O ships in versions. A Modbus TCP/RTU head such as the Ebyte M31-U series exposes 8DI/8DO with an RS485 port for expansion slices. An EtherCAT version (M31-U-CAT) and a CANopen version (M31-U-CAN) use the same housing and terminal layout. For sites where cable cannot be run at all, a LoRa-based wireless I/O station reports the same points over the air. Cabinet-mounted network I/O such as the ME31 series covers the case where the field device is only a few meters away but the PLC rack is full.
Nothing in the definition of remote I/O mentions a protocol. That absence is exactly why the term collides with PROFINET IO in so many procurement conversations.
What is PROFINET IO?
PROFINET IO is the part of the PROFINET standard that governs cyclic exchange of I/O data between a controller and field devices. It has been part of IEC 61158 and IEC 61784 since 2003 and is maintained by PI, the organization behind PROFIBUS. It runs on switched, full-duplex 100 Mbit/s Ethernet — standard IEEE 802.3 frames, not a proprietary physical layer.
It is the successor to PROFIBUS DP, but it is not the same thing. PROFIBUS DP is a 12 Mbit/s RS-485 fieldbus with a token-passing master. PROFINET moved the same data model onto Ethernet and added the diagnostics and topology functions that a switched network makes possible.
Three roles, and only three:
- IO-Controller — normally the PLC. Owns the connection and the output data.
- IO-Device — the field device: remote I/O station, drive, valve island, RFID reader.
- IO-Supervisor — an engineering or diagnostics tool that observes without controlling.
Three things you must have to commission a device:
- A GSDML file — the XML description of the device's modules, slot structure and data lengths. Without it the device does not exist in the engineering tool.
- A device name — assigned over the wire by DCP. PROFINET identifies devices by name, not by IP address, which is why a replaced module with a blank name is invisible.
- A topology view — built from LLDP neighbor discovery, which is what allows the controller to report where a cable broke rather than only that something broke.
Three performance classes define how deterministic the traffic is:
| Class | Channel | Typical cycle | Used for |
|---|---|---|---|
| CC-A | Cyclic, non-time-critical | ~10 ms and above | Simple monitoring, infrastructure |
| CC-B | Real-Time (RT) | 1–10 ms | Machine and process automation |
| CC-C | Isochronous RT (IRT) | down to 31.25 µs, ~1 µs jitter | Multi-axis motion, synchronized drives |
IRT reserves time slots in the switch hardware, so it needs IRT-capable switches on every hop of the path. Mixing a standard switch into an IRT segment does not slow it down; it breaks it.
Notice what this standard describes from beginning to end: frame formats, state machines, timing, diagnostics, naming. It never says where the I/O hardware should be installed. That question belongs to remote I/O.
Five differences that actually matter
1. They live on different layers
Remote I/O is a physical and topological decision: put the terminals here instead of there. PROFINET IO is a protocol decision: these bytes arrive on this cycle in this format. One is answered by a layout drawing, the other by a datasheet section.
2. Protocol flexibility versus ecosystem fit
A multi-protocol remote I/O can be re-pointed at a different controller family with a firmware or model change. A PROFINET device is committed: it works inside the Siemens ecosystem and with any third-party controller that implements PROFINET, which in practice means the engineering work happens in TIA Portal or STEP 7. If the plant standard is Siemens, that is an advantage. If your controller is a PC-based IPC from another vendor, it is friction.
3. Commissioning works differently
A Modbus TCP station is usually brought up in a browser: set an IP, map registers, done. A PROFINET device is brought up in the engineering tool: import GSDML, drag the device into the network view, assign a name, set the I/O addresses, download, then verify that the name actually reached the device. The second path is more structured and more diagnosable, but it is not faster for a two-point station.
4. Determinism is a design property, not a label
Modbus RTU over RS485 is polled: total update time is the sum of request and response frames for every slave, and it degrades linearly as you add stations. A single multi-register read at 19.2 kbit/s takes roughly 10–15 ms, so a full scan of a dozen slaves lands in the 100–200 ms range before any application logic runs. PROFINET builds a schedule at configuration time and holds a 1–10 ms cycle against it — the minimum achievable cycle still grows with device count and data volume, but it is known and repeatable rather than emergent. IRT goes further and reserves the time slot in hardware.
Neither is “better”. A tank level that changes once a minute does not need 1 ms. A flying-shear registration does not tolerate 40 ms.
5. Diagnostics differ by an order of magnitude
Polling a Modbus register tells you the value, and nothing about why it is wrong. A PROFINET device reports channel-level faults, wire break, short circuit, module pulled, and its own position in the topology — as structured alarms, in the controller's diagnostic buffer. On a line that stops at 3 a.m., that difference is measured in hours of troubleshooting, not seconds of cycle time.
Where the two overlap: PROFINET remote I/O
A PROFINET remote I/O station is both concepts in one housing, and it is the product most people actually mean when they say “remote I/O” in a Siemens plant.
The Ebyte M31-U-PN series is a working example. It is a remote I/O head in the sense above — DIN-rail mounted, 110 × 40 × 100 mm, 9–36 V DC, −40 to +85 °C, conformal coated, with variants from 8DI/8DO relay to 16AI and analog output — and it is also a PROFINET IO-Device: GSD-based integration, dual RJ45 ports with switch function for daisy-chaining, and no code required in the PLC to read or write the I/O. It also carries an RS485 port and an integrated PROFINET-to-Modbus RTU gateway, so a Modbus sensor chain can be pulled into the same station. The JGD111-P book-type series covers the cabinet-mounted case where slices are not wanted.
There is one reliable way to tell whether a module will work on your controller before you buy it: the datasheet must name the protocol and the vendor must supply a GSDML file. An RJ45 port proves nothing. Modbus TCP and PROFINET use the same connector, the same cable and the same switch — and share no software at all.
How to choose
Five questions, in this order. The first one usually settles it.
1. What controller is already installed?
Siemens S7-1200/1500 → PROFINET directly. Mitsubishi, Omron, Delta, or a PC-based IPC → Modbus TCP or EtherCAT is less work and less cost.
2. What cycle time does the process actually need?
Discrete interlocks and analog monitoring at 10–50 ms → Modbus TCP is sufficient and cheapest. Coordinated motion below 1 ms → PROFINET IRT or EtherCAT, and the network design becomes part of the project.
3. How far apart are the points?
Points spread across a line or a floor → one station per zone, daisy-chained through the second RJ45 port. Points inside one cabinet → local rack I/O is still cheaper than a network.
4. What is the environment?
Dust, coolant, washdown, no cabinet → IP67 field-mounted blocks. Clean control cabinet → DIN-rail IP20 stations like the M31-U series or ME31 network I/O.
5. Do you already have RS485 instruments and drives you are not replacing?
Then do not replace them. A PROFINET-to-Modbus RTU gateway such as the PN1-D25 series maps existing Modbus registers into PROFINET with a GSD file, supports 1× RS485 plus 1× RS422 with 32 command nodes per channel, and runs a PROFINET cycle down to 4 ms. It is usually the cheapest path from an old line to a new PLC, and it keeps the instruments that already work.
The short version: protocol first (the PLC decides it), form factor second (the site decides it), point count and spare capacity last.
Four misconceptions worth clearing up
“Remote I/O means wireless I/O.” Wireless is one transport option, used when cable cannot be run. The overwhelming majority of remote I/O in service is wired RS485 or Ethernet.
“PROFINET IO is the same as PROFINET.” PROFINET IO is one part of the standard. PROFINET also covers component-based automation, energy profiles (PROFIenergy) and application profiles; and PROFINET is not PROFIBUS.
“It has an Ethernet port, so it will work.” Without a GSDML file and a conformant stack, the controller will not accept the device. This is the single most expensive misunderstanding in this category.
“PROFINET is always faster than Modbus.” A PROFINET segment running through unmanaged switches with a careless topology can miss cycles. Determinism comes from the class you specified and the network you built, not from the logo on the box.
Frequently asked questions
Why does my remote I/O module not appear in the TIA Portal hardware catalog?
Two causes, in order of likelihood. First, the GSDML was never imported: TIA Portal → Options → Manage general description files → add the vendor GSDML, then the module appears under Other field > PROFINET IO. Second, the device has no PROFINET device name: assign it via Online access → assignable devices. PROFINET addresses devices by name, not by IP, so a module with a valid address that answers ping still stays invisible with a blank name. If the datasheet says Modbus TCP instead, there is no GSDML for it and it will never appear — that is a PROFINET IO-Device order, for example the M31-U-PN remote I/O head, not a configuration problem.
How do I connect existing Modbus RTU energy meters and VFDs to an S7-1500?
Gateway them; do not rewire them. A PROFINET-to-Modbus RTU gateway such as the PN1-D25 series sits in the middle: Modbus master toward the instruments, PROFINET IO-Device toward the PLC. It carries 1× RS485 plus 1× RS422 with 32 command nodes per channel and sustains a PROFINET cycle down to 4 ms. You set station number, baud and parity on the RS485 side, declare register ranges as input/output data in the GSD-based device configuration, and the mapping itself needs no PLC code. Check total Modbus scan time against your PROFINET cycle before you commit to the point list.
Can I daisy-chain PROFINET remote I/O stations along a conveyor instead of homering every node?
Yes, provided each node has a built-in 2-port switch. That is what the dual RJ45 ports on the M31-U-PN are for — one trunk runs the length of the line and each station drops out of the second port, which is the standard line topology for distributed I/O. Two limits: IRT (CC-C) needs IRT-capable switches on every hop, so a plain switch in that path breaks the isochronous domain rather than merely slowing it; and a line topology hides which node dropped unless LLDP neighbor discovery is enabled, so keep ring or star where the diagnostic map matters more than cable cost.
Which remote I/O module should I use for 4–20 mA sensors in a washdown area?
Mounting decides the part. No cabinet in the zone → IP67 field-mounted block, connectors only. Clean or splash-protected cabinet → DIN-rail IP20 with conformal coating: the M31-U family runs 9–36 V DC over −40 to +85 °C and offers analog input variants up to 16AI plus analog output; the ME31 covers the case where the cabinet is close but rack space is gone. On the analog side, confirm per-channel isolation, whether each channel is jumper- or software-selectable between 4–20 mA and 0–10 V, and reserve roughly 20 percent spare channels — field points always grow.
My Modbus RTU scan time is over 100 ms. How do I reduce it?
Pull the three levers that actually move polling time: raise the baud rate (9600 to 19200 or 115200), collapse point-by-point requests into contiguous register-block reads, and cut slaves per RS485 segment. Know the floor, though: one multi-register read at 19.2 kbit/s already takes about 10–15 ms, so a dozen polled slaves land in the 100–200 ms band no matter how tidy the polling is. If the process genuinely needs a repeatable cycle under 10 ms, that is a protocol change — Modbus TCP, or PROFINET CC-B with its configured 1–10 ms schedule — not further tuning.
The takeaway
Choosing wrong here is rarely a hardware failure. It is a category mistake made at the quotation stage: buying an architecture when you needed a protocol, or the reverse.
Decide the protocol from the PLC you own, decide the form factor from the site you have, and confirm that the datasheet names both the protocol and a GSDML file before the purchase order goes out.
If you are unsure which side of that line your application sits on, send us the PLC model, the point count and the distance between cabinets. We will tell you whether the answer is a PROFINET station, a Modbus station, or a gateway in front of instruments you already own.