Plants that run EtherCAT motion segments alongside PROFINET process segments are now the default shape of a 2026 retrofit: one cabinet houses a Siemens S7 controller on PROFINET IRT for I/O and valve islands, the next cabinet runs a Beckhoff CX or third-party motion controller on EtherCAT for servo axes, and the two talk through a documented gateway such as a PROFINET-to-EtherCAT master coupler from Helmholz, Beckhoff, or HMS [S1][S2].
PROFINET held 27% of new industrial Ethernet node installations in 2025, EtherCAT 17%, and EtherNet/IP 23%, per the HMS Networks 2025 Industrial Network Market Shares report cited in May 2026 [S2]. With that mix of installed base, a single-protocol greenfield is the exception rather than the rule on multi-vendor shop floors, and the question for most controls engineers is no longer whether to mix, but where the gateway sits and what it can pass.
Where the Gateway Belongs: Segment Boundaries, Not Cable Sharing
The first rule of mixed EtherCAT and PROFINET plants is that the two protocols do not share a physical segment [S4]. EtherCAT's "processing on the fly" model, in which a single Ethernet frame traverses every slave in sequence and each node reads its data on the way through, is incompatible with PROFINET's switched-Ethernet model, where every device replies independently to a controller [S1][S5]. A unmanaged five-port switch sitting between an EtherCAT master and a PROFINET controller will not bridge them; it will simply break both networks, because EtherCAT slaves will not arbitrate CSMA/CD the way PROFINET devices expect.
Correct topology is a PROFINET segment terminated at a dedicated gateway (a PROFINET slave on one side, an EtherCAT master on the other), then a separate EtherCAT segment downstream [S1][S4]. A typical 2026 packaging-line sketch has the Siemens S7-1500 on PROFINET IRT at 31.25 µs, the gateway as a PROFINET IO device with a GSDML file, and a Beckhoff AX5000 servo drive on EtherCAT downstream of the gateway. Cycle-time budgets published in vendor documentation assume each segment is independently engineered; the gateway adds 1 to 3 ms of latency, which is the single largest number most engineers underestimate when they spec the line [S1][S2].
What Crosses the Gateway, and What Does Not
Standard process data (cyclic I/O, diagnostics, parameter channel) crosses cleanly through any vendor PROFINET-to-EtherCAT coupler, with a typical mapping of up to 512 bytes of process data in each direction on the higher-end units [S4][S6]. For everything else, the picture is more uneven. Safety stacks do not pass through a generic gateway: PROFINET safety (PROFIsafe) and EtherCAT safety (FSoE, Fail Safe over EtherCAT) both ride on a black-channel model under IEC 61508, and each safety frame must be terminated inside its own segment by a safety controller of the same family [S2][S5]. A PROFINET PROFIsafe frame entering an EtherCAT segment will be discarded by every FSoE-capable slave on the segment.
Three traffic classes deserve careful scoping at the gateway. First, real-time motion: keep EtherCAT motion on its own segment, and never try to ride PROFINET IRT frames through an EtherCAT slave, because IRT requires PROFINET-certified switches with cut-through latency that no EtherCAT node provides [S3]. Second, configuration: most gateways expose an acyclic parameter channel that lets TIA Portal read EtherCAT slave CoE (CAN application protocol over EtherCAT) objects, but the read latency is acyclic and not deterministic, so do not loop process-control tuning through it. Third, TCP/IP services such as HTTP, HTTPS, and FTP can be tunnelled through the gateway using the EtherCAT mailbox protocol, but each tunnelled session adds jitter and reduces the available process-data window [S5].
Comparison: PROFINET vs EtherCAT Across Five Selection Criteria
Side by side, the two protocols diverge on the dimensions that drive gateway sizing. PROFINET RT runs cyclic I/O at roughly 1 ms with jitter in the 100 µs range, while PROFINET IRT reaches 31.25 µs with sub-microsecond jitter when running on PROFINET-certified switches; EtherCAT publishes cycle times under 100 µs with sub-microsecond jitter, and benchmarked motion systems have hit 31.25 µs [S2][S3][S7]. Topology-wise, PROFINET supports line, star, and ring (with MRP media redundancy) on standard managed switches, while EtherCAT is line or tree with mandatory daisy-chaining and a hard practical limit of 65,535 nodes per segment [S3][S5]. On safety, both support SIL 3 per IEC 61508 via PROFIsafe and FSoE respectively, but the two safety frames do not interwork at the gateway level [S2]. For engineering, PROFINET ties to Siemens TIA Portal and a deep GSDML ecosystem, while EtherCAT ties to Beckhoff TwinCAT and the ESI (EtherCAT Slave Information) XML device descriptions; mixing the two means maintaining two engineering toolchains and two device libraries, which is where integration cost actually shows up in a 2026 project [S1][S3].
Limits That Bite in 2026 Field Installations
Three limits repeatedly show up in commissioning logs. Gateway latency is the first: published numbers for current-generation PROFINET-to-EtherCAT couplers sit between 1 and 3 ms per direction, which is fine for process I/O but unacceptable for cross-segment motion synchronisation, where engineers must keep the EtherCAT motion loop on one side of the gateway and treat the PROFINET side as supervisory only [S1]. Device-description drift is the second: a PROFINET GSDML revision and an EtherCAT ESI revision for the same physical drive (for example a third-party servo) do not always ship on the same schedule, and a mismatch between master firmware and slave description is the most common cause of "green light in TIA, red light in TwinCAT" complaints.
Shared infrastructure is the third limit. Industrial Ethernet captures 79% of new factory nodes in 2025 per HMS, but that share does not mean one cable can carry two protocols [S2]. Mixing EtherCAT and PROFINET on the same unmanaged switch fabric corrupts both networks, and the failure mode is silent: PROFINET will keep running at degraded cycle times, EtherCAT will keep slaves in PRE-OP, and the line will fault only on a high-priority event such as a safety stop. The fix is enforced segment isolation, either by physical switch port separation with managed PROFINET switches on the PROFINET side, or by a dedicated gateway appliance that owns the boundary. The same logic shows up in adjacent gateway problems, and the patterns in DeviceNet to EtherNet/IP gateway conversion are worth reading for engineers who also have legacy serial fieldbus on the line.
Who the Mixed Plant Is For, and Who Should Avoid It
Mixed EtherCAT and PROFINET plants are right for OEMs that already have a Siemens process-control standard and need Beckhoff-class motion density, for retrofits where the cabinet is staying Siemens and the new motion axes are third-party, and for systems integrators who can support two engineering toolchains without doubling the commissioning budget [S1][S2]. They are not right for lines with a single safety zone that must stay on one safety stack, for plants where the controls engineers have not been trained on both PROFINET IRT and EtherCAT distributed clocks, and for any project where the customer expects one SCADA HMI to poll both networks in real time without gateway buffering.
Engineers new to this stack should also understand the broader construction machinery and equipment and lamps and light fittings ecosystem views, because the same gateway-vs-direct-cable logic that applies on a packaging line also governs how mobile machinery integrates cabin controls. For a process-control view, the flow meter and pressure transmitter encyclopaedia pages show how field devices are addressed in mixed-segment commissioning when a Profibus PA legacy device sits on a PROFINET proxy and an EtherCAT valve island sits downstream.
Engineering Practice: What a Clean Mixed Plant Looks Like
A well-engineered 2026 mixed plant keeps PROFINET on the process side (valves, I/O, instruments, drives that only need 1 ms update), keeps EtherCAT on the motion side (servo axes, robots, CNC spindles), and uses a single PROFINET-to-EtherCAT gateway per motion cell, not one gateway per drive [S1][S3]. The gateway is configured as a PROFINET IO device with a vendor-supplied GSDML, and the EtherCAT side is configured in TwinCAT or the third-party master's ESI tool, with one CoE object per PROFINET slot so that diagnostic data round-trips [S6]. MRP redundancy on the PROFINET ring and cable redundancy on the EtherCAT segment are specified independently; the gateway is not in the redundancy path unless the vendor explicitly publishes hot-swap behaviour for the specific firmware revision in use.
Cable selection is the same Cat 5e/6 spec used elsewhere in industrial Ethernet, but the EtherCAT segment must use the daisy-chain ports on each slave rather than tapping into a star switch, and the PROFINET segment must use PROFINET-certified switches if IRT is enabled [S3]. Functional safety, where required, stays inside its own protocol family on each segment: PROFIsafe on PROFINET, FSoE on EtherCAT, and any cross-segment safety function is implemented at the controller level rather than at the gateway. The same discipline shows up in broader plant retrofits, and the lighting equipment and electric lamps and concrete batching plant reference pages are useful cross-checks for how gateway boundary decisions scale to non-continuous process plants.
Trackable Signals and Next Nodes
Two signals are worth watching through the rest of 2026. The first is TSN (Time-Sensitive Networking) profile convergence, with PROFINET over TSN and EtherCAT over TSN both moving through the IEC/IEEE 60802 working group; once both ride the same converged physical layer, the gateway problem collapses to a software profile, not a hardware boundary [S2][S3]. The second is the migration of PROFIsafe-on-PROFINET and FSoE-on-EtherCAT toward a shared CIP Safety or openSafety profile under IEC 61508, which would let one safety controller span both segments. Until those land in shipping firmware, the right working assumption is that a PROFINET-to-EtherCAT gateway in 2026 is a documented segment boundary with a 1 to 3 ms latency tax, not a transparent bridge, and the engineering budget should be planned around that ceiling.