IMDHMIControls LibraryShop main store

Comparison

PROFINET versus DeviceNet for machine and panel I/O

Industrial networks Published 7 min read

Short answer

Choose by controller ecosystem first: PROFINET is industrial Ethernet at 100 Mbps with line, star and ring topologies, while DeviceNet is a CAN-based trunk running 125 to 500 kbps. Speed only decides when the panel needs dense or fast I/O. PROFINET for new high-speed machines, DeviceNet where the Rockwell-era devices and wiring already exist.

On this page
  1. Which network fits the controller you already have?
  2. How much faster is PROFINET than DeviceNet?
  3. What cable and topology does each one need?
  4. Can a machine use both networks at once?
  5. How do I check whether a 10 ms update fits on each network?
  6. Which network should I choose for a new panel?
  7. Questions asked

PROFINET and DeviceNet are not two speeds of the same wire: one is industrial Ethernet, the other a CAN-based fieldbus trunk, and the controller you already have usually settles the choice before performance is argued. PROFINET runs at 100 Mbps in line, star and ring topologies; DeviceNet runs at 125 to 500 kbps on a single multi-drop trunk. Match the installed ecosystem first, then let update-time arithmetic decide the rest. The standards, the numbers and a worked 10 ms check follow, with the decision for each case.

Which network fits the controller you already have? #

The controller platform decides this choice before speed does: a Siemens, CODESYS or other PROFINET-capable controller pulls you toward PROFINET, and an SLC 500 with its scanner family or an installed DeviceNet trunk keeps you on DeviceNet.

PROFINET came out of Siemens and is an open standard administered by Profibus & Profinet International (PI); controllers from Siemens and CODESYS-based platforms carry native ports, and I/O comes from many vendors. DeviceNet came out of Allen-Bradley, now Rockwell Automation; scanners such as the 1747-SDN put it on the SLC 500 family, and although multiple vendors build devices, the ecosystem is Rockwell-centred. Both are international standards: PROFINET sits in the IEC 61158 and IEC 61784 family as a real-time Ethernet profile, DeviceNet in the IEC 62026 series for low-voltage device interfaces.

If the controller already speaks one of them natively, that is the decision; adding the other means a gateway or new hardware. On the PROFINET side, the day-one task is picking the port and master — select a CODESYS fieldbus master for industrial Ethernet walks through that. On the DeviceNet side, it is scanner mapping; map a 1747-SDN scanner for DeviceNet drive control shows the shape of the job.

How much faster is PROFINET than DeviceNet? #

PROFINET moves 100 Mbps over industrial Ethernet while DeviceNet tops out at 500 kbps on its CAN trunk, so the raw bandwidth ratio is 200 to 1 at DeviceNet's fastest rate and 800 to 1 at 125 kbps.

Decision point PROFINET DeviceNet
Physical layer Industrial Ethernet, twisted pair or fiber CAN-based multi-drop trunk
Line rate 100 Mbps 125, 250 or 500 kbps
Topology Line, star, ring, tree One trunk with drops
Traffic Real-time cyclic I/O plus TCP/IP parameter traffic Cyclic, acyclic and event-driven CAN messaging
Standards IEC 61158 / IEC 61784 profile IEC 62026 series
Cyclic capacity for planning About 12,500 bytes per 10 ms at one-tenth of the wire About 150 bytes per 10 ms at the top rate
Typical fit Motion, dense or fast I/O, new machines Small slow panels, installed Rockwell-era fieldbus

Read the ratio as headroom, not as speed you will use. PROFINET carries real-time cyclic I/O and ordinary TCP/IP traffic on the same infrastructure, and the wire is almost never the constraint. DeviceNet exchanges cyclic and acyclic data with event-driven messaging over CAN framing, and at its top rate the wire is the constraint — the update-time check below converts the ratio into bytes per millisecond so you can see exactly where that happens.

What cable and topology does each one need? #

PROFINET gives you line, star, ring and tree topologies on Ethernet twisted pair or fiber behind switches, while DeviceNet is one shielded trunk with drops and both ends terminated — simpler to wire, fixed forever after.

On PROFINET, a switch gives you a star for free, devices daisy-chain into a line along the machine, and a ring buys media redundancy when the switches support it; fiber covers distance and electrical noise. The planning work is switch placement and cable routes, and later changes are incremental: add a port, add a drop.

DeviceNet is one trunk, full stop. Topology planning is drop placement, because the trunk is the network: screw-terminal drops tap the shielded pair, both ends take terminators, and every node runs one shared bit rate. Because it is CAN underneath, its stubborn faults are physical rather than logical — the dead-segment kind traced in why a 4-node CAN bus has no communication at 250 kbps.

Can a machine use both networks at once? #

Yes, and it is routine: a linking device maps DeviceNet nodes onto a PROFINET controller as cyclic I/O, so a mixed machine keeps its CAN drops while the controller side moves to industrial Ethernet.

The linking device keeps the trunk alive and presents its nodes as one cyclic I/O image, so the panel wiring stays put while the controller side moves to Ethernet. The same move works when the controller is the outsider: connect PROFINET IO to a CompactLogix through a gateway covers the reverse direction on a Rockwell controller. Commit to the mapping once and treat the gateway image like a remote rack with a fixed layout — every later device change has to fit inside it, which is the real cost of the hybrid.

How do I check whether a 10 ms update fits on each network? #

Count your cyclic bytes and divide by the wire rate: 16 nodes at 32 bytes each is 512 bytes, which needs about 16.6 ms of pure wire time on a DeviceNet trunk at its fastest rate but under 0.1 ms on 100 Mbps PROFINET.

Assumptions for the check, stated so you can rerun them with your own numbers: 16 I/O nodes, 32 bytes of cyclic data each, a 10 ms target, DeviceNet running at its fastest rate, 8 data bytes per CAN frame, and 130 bits of wire per frame — a standard 8-byte CAN frame is about 111 bits before stuffing, and 130 covers stuffing plus interframe space.

DeviceNet side:

  • Data: 16 nodes × 32 bytes = 512 bytes.
  • Frames: 512 bytes ÷ 8 bytes per frame = 64 frames.
  • Wire bits: 64 × 130 bits = 8,320 bits.
  • Wire time: 8,320 bits ÷ 500,000 bps = 16.6 ms.

That is wire time only, before scanner polling and response turnaround, so the 10 ms target fails at every DeviceNet rate; dropping to 250 kbps only doubles the 16.6 ms.

PROFINET side:

  • Per node: 32 data bytes + about 38 bytes of Ethernet, IP and UDP framing ≈ 70 bytes = 560 bits.
  • Total: 16 × 560 bits = 8,960 bits.
  • Time at 100 Mbps: 8,960 bits ÷ 100,000,000 bps = 89.6 µs ≈ 0.09 ms, better than 100 to 1 margin against the target.

Turn it into a planning number. At the top rate, one millisecond carries 500 bits, which is 500 ÷ 130 ≈ 3.8 frames, or about 31 bytes of payload; halve that for scanner overhead and plan on 15 bytes per millisecond — roughly 150 bytes per 10 ms of update time. Below that line DeviceNet is comfortable: 8 nodes at 4 bytes is 32 bytes in 4 frames, 520 bits, 1.0 ms of wire time at the fastest rate and 4.2 ms at 125 kbps, both trivial inside a 100 ms update. Above the line, the arithmetic has already made the choice for you.

Which network should I choose for a new panel? #

Choose PROFINET for any new machine with motion or with cyclic I/O above about 15 bytes per millisecond of update time; choose DeviceNet where an installed scanner, an existing trunk or a small slow panel dominates the design.

  • Motion on a new machine: PROFINET, no discussion. Axis control is cyclic telegram work, the pattern shown when you control a PROFINET servo drive with cyclic telegram data.
  • An I/O-only panel under the line — say 12 nodes at 8 bytes on a 100 ms update, under 1 byte per millisecond: either network carries it, so take the one your controller speaks natively and your electrician can terminate without a course.
  • A working DeviceNet installation on a Rockwell-era platform: stay on it until you cross the 15 bytes per millisecond line, because the trunk you own is worth more than bandwidth you are not using.
  • A mixed or phased machine: linking device, mapped once, treated as a remote rack.

One boundary holds for both: plain PROFINET and plain DeviceNet are not safety networks — machine safety functions ride the safety profiles (PROFIsafe over PROFINET, CIP Safety over DeviceNet) and get their own validation before anything depends on them.

Questions asked

What bit rate should I set on a small DeviceNet trunk?

Set the fastest rate the trunk length allows, because the permitted trunk length falls as the bit rate rises. Set every node, scanner included, to the same rate, terminate both physical ends of the trunk, and retest communication after any rate change. If drops are long or the environment is noisy, drop to 250 kbps and accept the slower cycle.

Do I need special switches for PROFINET?

Basic line and star layouts run on ordinary Ethernet switching, but an industrial cabinet argues for managed industrial switches: they carry ring redundancy, report port diagnostics, and survive the panel. Reserve switch hardware rated for the deterministic classes only when you are doing motion; plain I/O scanning does not need it.

Can I convert DeviceNet devices to PROFINET without rewiring?

No. The CAN trunk is not Ethernet media, so there is no cable-level conversion. Either replace the nodes with PROFINET I/O and pull new cable, or keep the trunk and put a linking device between it and the controller. The linking device maps the DeviceNet I/O into the PROFINET cyclic image, so the drops and field devices stay where they are.

Is DeviceNet the same as CANopen?

Both ride the same CAN physical layer, so they share the trunk rules: one bit rate, termination at both ends, multi-drop wiring. Above the wire they differ. DeviceNet is its own device profile tied to the Rockwell ecosystem, while CANopen is the profile common on servo drives and embedded nodes. Devices on one trunk must all speak the same one.

Why does a DeviceNet trunk stop communicating after a rate change?

Every node on a CAN trunk must run the same bit rate, and the trunk must be terminated at exactly its two physical ends. Set the scanner to the fastest rate while one drive still sits at 125, or leave a terminator in the middle after moving a node, and the whole segment goes quiet. The fault follows the wiring, not the software.

Keep reading

  • Selection guide

    How to select a CODESYS fieldbus master for industrial Ethernet

    Select the master around the device, not the connector: the 750-340 is a PROFINET IO coupler, and a PROFIBUS station needs a different coupler. Confirm three matches — device protocol, runtime master package, supported network adapter — then size the bus cycle from the process budget; the worked example clears a 10 ms sensor-to-actuator budget at a 2 ms cycle and fails it at 4 ms. Bench-test a laptop NIC only if the runtime documents it.

    Industrial networks

  • Selection guide

    How to choose between Modbus, Profibus and CANopen

    Match the protocol to the PLC master that already exists, not to the datasheet that impressed you. Then fit the cable run to the bit rate, because CAN-based trunks shorten as speed rises, and mix protocols one segment at a time through gateways or the PLC itself. A fourteen-device example shows the arithmetic.

    Industrial networks

  • How-to

    Choosing the DeviceNet configuration connection path

    RSNetWorx for DeviceNet cannot reach a 1769-SDN through the MicroLogix 1500 serial port: that port talks to the controller, not to the trunk. Configure the scanner from a direct DeviceNet interface such as a 1770-KFD or 1784-PCD, or route through a Logix controller port such as Ethernet on an L32E. RSNetWorx is still required for scan-list work; Auto-Configure covers only simple I/O blocks.

    Industrial networks