Industrial Ethernet uses the same physical layer as office IT. The divergence is in how the network recovers from a link or device failure — and how fast. In a production environment, network recovery time directly maps to unplanned downtime.
Star topology
Every device connects to a central managed switch. A single device failure affects only that device. A switch failure takes down everything connected to it.
[Core Switch]
/ | \
[PLC-01] [HMI-01] [Switch-2]
/ \
[Drive-01] [Drive-02]Redundancy in star topologies is achieved with dual uplinks and Rapid Spanning Tree Protocol (RSTP). RSTP convergence time is typically 1–30 seconds depending on network depth and switch vendor implementation. That is acceptable for supervisory traffic; it is not acceptable for closed-loop control.
Use managed switches throughout. Unmanaged switches cannot participate in RSTP, create loops, and mask faults from network diagnostics.
Ring topology with MRP
Media Redundancy Protocol (MRP, IEC 62439-2) closes a physical ring and monitors link continuity. One device is designated the MRP Manager; all others are MRP Clients. Under normal operation, the Manager blocks one port to prevent a broadcast loop. On link failure detection, the Manager unblocks the redundant path.
[PLC] ── [Drive-01] ── [Drive-02] ── [I/O Block] ── [PLC]
└────────────────── ring ──────────────────────────┘MRP recovery time: ≤200ms (default). Configurable to ≤30ms on hardware that supports it.
PROFINET controllers (e.g., Siemens S7-1500) include dual integrated Ethernet ports with onboard switch silicon. Devices can participate directly in a ring without an external switch at each node. This reduces hardware cost and eliminates the external switch as a failure point.
Ring size limit: keep rings to ≤50 devices. Larger rings increase the Manager's fault detection time and complicate diagnostics.
DLR: Device Level Ring (EtherNet/IP)
EtherNet/IP implements ring redundancy via Device Level Ring (DLR), defined in the EtherNet/IP specification. DLR embeds ring supervisor capability into end devices — drives, I/O modules, and panels — without requiring a managed switch at every node.
Recovery time: <3ms, which qualifies DLR for synchronized motion control loops.
Allen-Bradley PowerFlex drives and 1734-series POINT I/O modules natively support DLR. Verify DLR support in the device's EDS file before designing a ring — not all EtherNet/IP devices implement it.
Fiber in industrial rings
Copper Cat5e/Cat6 has a 100m segment limit and is susceptible to ground loops in high-EMI environments — common near VFDs and large AC motors. Multimode fiber (OM3/OM4) eliminates both constraints: segments up to 550m–2km, galvanic isolation between nodes.
Use copper within the control cabinet. Use fiber for segments crossing the machine frame, traversing the plant floor, or installed near high-power drive wiring.
Topology selection guide
| Criterion | Star (RSTP) | Ring (MRP/DLR) |
|---|---|---|
| Fault recovery | 1–30 seconds | <200ms (MRP), <3ms (DLR) |
| Failure scope | Switch = all downstream | Single link = zero downtime |
| Cabling | Home-run to switch | Daisy-chain |
| Device count | Scales freely | ≤50/ring recommended |
| Motion control | Not suitable | DLR qualified |
| Hardware cost | Higher (more switches) | Lower (fewer switches, longer cable runs) |
Use ring when a fault recovery time requirement exists — any control application with a defined Maximum Allowable Downtime (MAD) under 30 seconds. Use star for supervisory layers (L2–L3) where RSTP convergence is acceptable and device counts are large.
Never mix managed and unmanaged switches in a ring. An unmanaged switch cannot forward MRP frames, breaks the ring's fault detection, and produces a split-brain condition on link failure.