EtherCAT posts 30–100 µs cycle times with sub-microsecond jitter, and PROFINET IRT reaches a hard floor of 31.25 µs on certified hardware [S3][S5]. For synchronized multi-axis servo work below 1 ms, EtherCAT is the consistent benchmark across 2025–2026 vendor guidance [S1][S2][S4].
PROFINET remains the right answer for plants already running Siemens S7-1500, ET 200, and TIA Portal, where 1 ms RT is fast enough and tooling cost dominates the decision [S1][S4]. The protocol gap is no longer academic: in a March 2026 conveyor benchmark, PROFINET IRT held 4 ms with 15 µs jitter across 96 diverters, while EtherCAT held 1 ms with 0.5 µs jitter on the same load [S2].
Cycle Time and Jitter: The Hard Numbers
Independent 2026 references converge on a 5–10x cycle-time gap between the two protocols at motion grade: EtherCAT at 30–100 µs, PROFINET IRT at 250 µs typical and 31.25 µs as a best-case hardware floor [S1][S3][S4][S5]. Jitter tells the same story: EtherCAT sits below 1 µs on a processing-on-the-fly ring, while PROFINET IRT lands in the 1–15 µs range depending on switch class and topology [S2][S5]. For a servo loop closing at 8 kHz, EtherCAT leaves 8x the timing headroom of a typical IRT install, and roughly 125x the headroom of a 1 ms RT install.
The architectural cause is the "frame-on-the-fly" read/write in EtherCAT: one Ethernet frame traverses all slave nodes in sequence, with each device extracting its data and inserting outputs in nanoseconds, instead of waiting for a polled request [S1][S3][S4]. PROFINET IRT instead relies on time-division multiplex scheduling across IRT-certified switches, which is why its lowest cycle times demand specific switch silicon [S4][S5]. If your motion controller sits inside a motion controller tier that publishes a distributed clock below 100 µs, EtherCAT preserves that capability natively; PROFINET IRT can match it only with switch-class compliance audited at build time.
Topology and Cabling Constraints
EtherCAT's mandatory line (daisy-chain) topology is the single biggest source of commissioning pain in retrofit plants, and it is the reason brownfield projects without free conduit paths default to PROFINET [S4]. You can branch with EK1100/EK1500 couplers, but the trunk must remain a line; ring redundancy exists via EtherCAT P and dual-port slaves, yet it costs more per node than a PROFINET ring with managed switches [S3][S4].
PROFINET accepts star, line, and ring topologies on standard managed Ethernet switches, and MRP (Media Redundancy Protocol) gives a sub-200 ms failover that is good enough for most discrete lines [S4][S5]. Device replacement without engineering tools is a PROFINET-native feature using LLDP topology discovery, which is meaningful on plants where you staff for line stoppages, not for laptop reconfigurations [S5]. For high-density servo cabinets, see how the control cabinet layout and cable routing interact with the chosen topology before locking the BOM.
Selection Criteria by Application
For sub-millisecond, high-axis-count motion (more than 10 coordinated servos, CNC, semiconductor handlers, delta robots, high-speed printing), EtherCAT is the only protocol of the three that delivers cycle time and jitter out of the box on commodity Ethernet cabling [S1][S4][S5]. For Siemens S7-1500 plants with single-axis or loosely coupled motion, PROFINET IRT's 250 µs class is more than sufficient and avoids forcing a second vendor's controller into a Siemens line [S1][S3]. For process plants dominated by analog loops, valve actuation, and slow I/O, PROFINET RT or even Modbus TCP wins on cost and integration simplicity [S1][S4].
A useful proxy is the published data point from a divert-and-conveyor test: PROFINET IRT at 4 ms handled 96 diverters without loss, which means any application below roughly 4 ms per node is comfortably inside PROFINET IRT's envelope; the moment you need to close a current loop at 16 kHz or synchronize a flying cutoff, you are in EtherCAT territory [S2]. For multi-protocol retrofit where a single control cable trunk already exists, weigh whether the cabinet layout supports a second line topology before commissioning starts.
Who It Is For vs Who It Is Not For
Choose EtherCAT if you are standardizing on Beckhoff TwinCAT or any PC-based control architecture, or if you need 10+ synchronized axes with deterministic sub-100 µs delivery; you also benefit if the OEM machine builder supplies it pre-commissioned, which most CNC, packaging, and robotics OEMs do in 2026 [S1][S3][S4][S6]. Choose PROFINET IRT if your plant floor is S7-dominated, your motion axes are below ten, and your engineering team writes in TIA Portal; IRT class A hardware is mandatory, and so is the IRT switch certification audit [S4][S5].
Skip EtherCAT if you must integrate into a Rockwell/Allen-Bradley Logix line at the controller level; CIP Motion on EtherNet/IP at 1 ms is the native Rockwell answer, and a gateway to EtherCAT adds a real-time seam you will fight for years [S5]. Skip PROFINET IRT if you need sub-100 µs jitter, because the protocol physically cannot deliver it without bespoke switch silicon that you would not buy just for one cell [S2][S4]. For access-controlled production halls that still need cycle-time discipline, the access control layer rarely interferes with these industrial Ethernet choices, but it can pin the Ethernet switch stack you are allowed to mount.
Comparison Matrix Across Decision Criteria
Comparing the two protocols on the four criteria that actually drive a 2026 retrofit or greenfield decision: cycle time (EtherCAT 30–100 µs, PROFINET IRT 250 µs class A, PROFINET RT 1–10 ms); jitter (EtherCAT below 1 µs, PROFINET IRT 1–15 µs, PROFINET RT milliseconds); switch hardware cost (EtherCAT standard managed switches, PROFINET IRT certified switches at a premium, PROFINET RT standard switches); and topology freedom (EtherCAT mandatory line, PROFINET star/line/ring) [S1][S2][S3][S4][S5]. On each axis, EtherCAT wins on speed and jitter, PROFINET wins on topology and switch cost, and the only criterion where PROFINET IRT beats EtherCAT is the existing Siemens tooling base [S4][S5].
Limitations, Failure Modes, and Interoperability
The most common failure mode in the field is mismatched switch class: PROFINET IRT cell running on RT-only switches silently degrades to RT cycle times, which usually shows up as a servo fault under load rather than at commissioning [S4][S5]. The second is EtherCAT drop-out on a single cable break in a non-redundant line, which takes the entire segment down until the slave with the break is bypassed, and that is why brownfield retrofits without free trunk paths default to PROFINET [S3][S4].
For multi-protocol plants, a PROFINET-to-EtherCAT gateway exists from several vendors, but expect 100–500 µs of added latency per hop, which collapses the EtherCAT advantage at the cell boundary [S3][S5]. Time-Sensitive Networking (TSN) is the long-promised convergence layer in IEC 61158 revision work, but the 2026 reality is that TSN still requires careful profile selection per vendor and is not yet a drop-in replacement for IRT in production cells [S4]. For projects where the control valve loop has to be closed at high rate over the same plant network, the control valve decision and the protocol decision are usually made together, because analog 4–20 mA on a fast EtherCAT ring is wasted bandwidth and a fast PROFINET ring with HART is the more common pair.
Standards Reference and Sourcing
Both EtherCAT and PROFINET sit inside the IEC 61158 family of real-time Ethernet fieldbus standards, which is why the architectural decision is the vendor's to make, not the integrator's [S4][S5]. For PROFINET, the cycle-time floors are anchored in PROFIsafe and PROFINET IRT conformance classes, and for EtherCAT in the ETG (EtherCAT Technology Group) conformance test. When validating a 2026 specification, ask the vendor for the conformance class number and the jitter measured on a representative 16-axis cabinet, not the data sheet single-axis number [S3][S4].
For machine builders comparing servo pick-and-place units to robotic welders, the same protocol decision recurs, and a side-by-side read of EtherCAT vs PROFINET for robot motion control confirms that the dominant axis remains servo-loop rate, not number of axes, as the trigger to switch protocols [S6]. For plants where motion control sits next to a process skid, the protocol choice usually follows the controller family already in linear motion actuators on the same line.
Track the next protocol-class revisions announced at SPS 2026 (Nuremberg, late November) and the ETG quarterly conformance reports for any new 31.25 µs PROFINET IRT class B devices; both will materially affect the 2027 build cycle.
This topic is covered further in Epoxy Reactor Batch Control: Viscosity Endpoint Detection in 2026.