The Power of Ethernet

Key takeaways

• Industrial Ethernet now carries hard real-time traffic — if you specify the right Time-Sensitive Networking (TSN) sub-standards.

• Single-pair Ethernet (10BASE-T1L) is replacing 4-20 mA at the field layer, with Advanced Physical Layer (APL) for hazardous areas.

• Power over Ethernet (PoE) Type 4 delivers up to 90W per port — heat and cable selection are not optional at that wattage.

• IT/OT convergence is real and useful, but breaks in predictable places. Plan the seams.

• Ethernet is the right default for the plant floor. It is not the only answer.

A controls engineer walks into a 20-year-old plant on Monday morning. The fieldbus network in cabinet 7 still runs Modbus RTU over RS-485, but the project spec on her desk asks for cloud telemetry, machine-vision quality checks, and a single dashboard the operations director can read from his phone.

She has a choice. Pull more shielded twisted pair and bridge to a gateway. Or pull Cat6a and start over.

This is the real question behind "the power of Ethernet" in 2026: what does it actually do for engineers on the plant floor right now, and where does it still fall short? Ethernet is no longer the office network you tolerate in the control cabinet. It is the physical layer carrying control traffic, instrument signals, and DC power to devices that used to need three separate cables.

That shift deserves a clear-eyed look — including the parts most blogs leave out.

What Ethernet does on the plant floor

Industrial Ethernet is standard IEEE 802.3 Ethernet with adjustments for plant conditions. Rugged connectors. Wider temperature ranges. Switches built for vibration and electrical noise. The frames are the same. The expectations are not.

Four protocols carry most industrial Ethernet traffic today:

• EtherNet/IP — the dominant choice in North American discrete manufacturing. Runs the Common Industrial Protocol (CIP) over standard TCP/IP. Strong fit for Rockwell-heavy plants.

• PROFINET — the European counterpart, common in Siemens architectures. Three real-time classes (NRT, RT, IRT), from standard TCP/IP to hard real-time.

• Modbus TCP — Modbus you already know, wrapped in TCP. Simple, free to implement, and still the easiest path off a legacy serial network.

• OPC UA over TCP — primarily a data layer, though OPC UA PubSub with TSN is moving the protocol into control territory. Carries information from controllers up to a Manufacturing Execution System (MES) and the cloud with a published information model.

Pick the wrong one and integration costs climb. Pick the right one and the rest of the design choices get easier.

"Ethernet is not deterministic enough for real-time control"

This is the objection most engineers carry from training a decade ago, and it deserves a direct answer.

In the 1990s the objection was correct. 100BASE-T used best-effort delivery, switch buffers introduced jitter, and motion-control vendors built proprietary fieldbuses because Ethernet could not promise sub-millisecond response.

Time-Sensitive Networking (TSN) changed the math. TSN is a family of IEEE 802.1 sub-standards — 802.1AS for sub-microsecond clock synchronization, 802.1Qbv for scheduled traffic, 802.1Qbu for frame preemption. When the right ones are present in every switch on the path, standard Ethernet now carries control traffic with bounded latency.

CIP Sync, PROFINET IRT, and EtherCAT solve the same determinism problem with different tradeoffs — some predate TSN, some now build on top of it. None of them are magic.

The honest read for engineers specifying hardware: do not buy a switch because the data sheet says "TSN-ready." Ask which 802.1 sub-standards it implements, on which ports, at what speeds. The phrase covers a wide range of capability.

Single-pair Ethernet — the shift worth watching

If you only track one development in industrial networking this year, make it single-pair Ethernet.

Single-pair Ethernet (SPE), particularly the 10BASE-T1L variant, does to 4-20 mA loops and the Highway Addressable Remote Transducer (HART) protocol what regular Ethernet did to RS-485. One twisted pair. Up to 1,000 meters. Power and data on the same conductors. Each device gets an IP address.

For process plants this is the biggest physical-layer change in a generation. A pressure sensor in the field can sit on the same routable network as the historian — no gateway, no protocol translation, no signal-conditioning card in the rack.

Advanced Physical Layer (APL) takes the same idea into hazardous areas. Ex-rated SPE with intrinsically safe power delivery, ratified by a consortium of process-automation vendors. A field instrument in a Zone 1 area can run Ethernet directly.

The honest counterpoint: the ecosystem is maturing, not mature. Sensor selection is still narrower than 4-20 mA. Hazardous-area certifications take time to extend across vendors. Asset management tools that engineers already use for HART do not all speak SPE yet. Pilot the technology where it fits and let your sensor vendor's roadmap drive the timing on broader rollout.

Power over Ethernet — one cable for power and data

Power over Ethernet (PoE) has moved well past office IP phones.

The IEEE has now ratified four PoE Types:

• 802.3af Type 1 — up to 15.4W at the Power Sourcing Equipment (PSE), 12.95W at the Powered Device (PD)

• 802.3at Type 2 — up to 30W at the PSE, 25.5W at the PD

• 802.3bt Type 3 — up to 60W at the PSE, 51W at the PD

• 802.3bt Type 4 — up to 90W at the PSE, 71.3W at the PD (Class 8)

Within those Types, the IEEE defines nine power Classes (0 through 8) that specify the exact wattage handshake between PSE and PD. Type 4 Class 8 is the ceiling — 71.3W at the device — and it is enough to power a machine-vision camera, a small actuator, a wireless access point inside a hazardous-rated enclosure, or a Radio Frequency Identification (RFID) gate reader without running separate power.

The win is real. One cable cuts conduit fill, shrinks panel space, and shortens commissioning time. The catch is also real. Ninety watts over Cat5e in a 40°C cabinet generates heat the cable was not specified for. Bundled runs de-rate further. Specify Cat6a, calculate bundle de-rating against TIA-TSB-184-A, and verify the switch's per-port and total power budgets before assuming the design works.

The IT/OT convergence story, told honestly

Converging IT and operational technology (OT) networks looks like a clean win on the architecture diagram. The reality has sharp edges.

What convergence gets right: shared cabling and switches, common engineering skills, easier data flow from controllers to MES and cloud analytics. Vendors aligned around standard IP make every layer above the physical one cheaper to support.

What it gets wrong in practice:

• IT change windows assume nightly reboots. OT plants run continuously. Patching a core switch during a production run has caused more downtime than any cyberattack we have seen.

• VLAN segmentation that looks textbook to a network admin can break multicast-heavy protocols like EtherNet/IP when Internet Group Management Protocol (IGMP) snooping is misconfigured.

• Security teams pushing firmware updates without coordination can drop a Programmable Logic Controller (PLC) connection mid-batch.

The middle path is honest convergence: converged at the physical layer, segmented at the logical layer, with documented change-management boundaries between the two groups. The ISA/IEC 62443 standard is a sensible reference point for drawing those lines.

Where Ethernet is still the wrong choice

Being specific about tradeoffs is how engineers earn trust.

Ethernet is the right default, not the only answer. The cases where it loses are worth knowing:

• Very high-speed coordinated motion. EtherCAT and SERCOS III still beat standard Ethernet on cycle time for axes that need tight coordination. Both use Ethernet cabling, but the data-link layer is purpose-built.

• Long single runs without fiber or SPE. Standard 100BASE-T and 1000BASE-T cap at 100 meters of copper. If you cannot pull fiber and SPE does not fit the application, RS-485 still runs 1,200 meters off the shelf.

• Heavily mobile equipment. Automated Guided Vehicles (AGVs), overhead cranes, and gantries that move continuously are usually better served by industrial wireless or a private 5G slice than a slip ring or festoon cable.

• Cost-sensitive single-sensor links. A 4-20 mA loop to one transmitter is still cheaper, simpler, and easier for a field technician to troubleshoot than a managed switch port.

If a vendor tells you Ethernet is the answer at every layer of your architecture, ask harder questions.

When wireless wins: Wi-Fi 6/7 and private 5G

Mobile assets are where Ethernet runs out of road. Industrial Wi-Fi 6 and Wi-Fi 7 deliver multi-gigabit throughput with low latency and per-device scheduling — good enough for AGV fleet control, mobile Human-Machine Interfaces (HMIs), and operator tablets across a warehouse. Private 5G goes further: deterministic timing, dedicated spectrum, and clean handoff across large facilities for assets that move continuously across cell boundaries.

Neither replaces Ethernet at the cell. Both feed into it. The right architecture in 2026 puts fixed assets on copper, mobile assets on wireless, and converges both into the same managed switch fabric before data leaves the plant.

Cable, topology, and connector choices engineers underweight

A few practical points that rarely get airtime:

• Shielded vs unshielded. S/FTP is the default reflex in noisy environments, but shield grounding done wrong creates ground loops that produce more errors than unshielded cable would have seen. Bond the shield at one end unless the standard for your protocol calls for both.

• Cat6 vs Cat6a. Cat6 handles 1 Gbps to 100 meters and 10 Gbps to 55 meters. Cat6a extends 10 Gbps to the full 100 meters and gives you better PoE thermal performance. In a vibration-heavy environment, the stiffer Cat6a jacket also resists cable damage from cabinet movement.

• Ring topologies. Device Level Ring (DLR) on EtherNet/IP recovers in under 3 milliseconds on a 50-node ring. Media Redundancy Protocol (MRP) on PROFINET supports four configurable profiles — 10 ms (rings up to 14 switches), 30 ms, 200 ms (the common default), and 500 ms. Pick the profile that matches your ring size and uptime budget. IT teams sometimes resist ring topologies because they break spanning-tree assumptions. On the plant floor they are legitimate, well supported, and worth specifying.

• Connector choice. M12 X-coded supports 10 Gbps. D-coded tops out at 100 Mbps. RJ45 in a panel without a strain relief will eventually fail. Choose the connector for the next 10 years of the line, not the next 10 weeks.

Security on the floor, in short

The shift from air-gapped fieldbus to routable IP traffic opened the plant to the same risks the IT side has handled for decades — without the same hygiene built in.

The baseline: managed switches with port security enabled, Access Control Lists (ACLs) between cells, network segmentation aligned to ISA/IEC 62443 zones and conduits, and a change-management process that includes both control engineers and security staff. Vendor hardening guides for specific switches are a good starting point and worth reading before commissioning.

What to ask a vendor before you buy

Five questions worth asking out loud:

1. Which IEEE sub-standards does this switch actually implement — TSN profiles, PoE class, MRP, IGMP snooping?

2. What is the Mean Time Between Failures (MTBF) rated in a 60°C cabinet, not a 25°C lab?

3. Does configuration require a Windows-only tool, or a browser?

4. Is the firmware signed, and how do updates reach an offline plant?

5. Who answers the phone at 2 a.m. when a link flaps mid-production?

Answers to those five usually tell you more than the data sheet does.

Frequently asked questions

Is industrial Ethernet faster than fieldbus? For most modern industrial Ethernet protocols, yes. PROFINET IRT, EtherNet/IP with CIP Sync, and EtherCAT all deliver cycle times below 1 ms — faster than PROFIBUS DP, DeviceNet, or Modbus RTU at typical baud rates. The honest qualifier: speed only matters if the application needs it. Many sensor loops still run fine on slower protocols at lower cost.

Can I run a Programmable Logic Controller (PLC) on Power over Ethernet? Some small PLCs and remote I/O modules accept PoE Type 2 (30W) or Type 3 (60W) power. Most full-size PLCs still need a dedicated 24 VDC supply. Check the device data sheet for "PoE PD" (Powered Device) compliance and the required class before specifying.

Do I need Time-Sensitive Networking (TSN) for my application? Only if you need bounded latency for time-critical traffic — motion control, hard-coordinated drives, or safety messages — sharing a network with non-time-critical traffic. For sensor monitoring, supervisory control, and data historization, standard switched Ethernet is sufficient.

What is Single-pair Ethernet (SPE) and where does it fit? SPE is Ethernet that runs on one twisted pair of copper. The 10BASE-T1L variant covers up to 1,000 meters at 10 Mbps with power and data on the same conductors — fitting the role 4-20 mA loops play today. APL extends SPE into hazardous areas. Best fit: process plants, large-distance sensor networks, and brownfield retrofits where pulling four-pair cable is impractical.

Can Ethernet replace my existing fieldbus? Often, but rarely all at once. Most plants migrate cell by cell, using protocol gateways during the transition. Grid Connect's gateway products bridge Modbus, PROFIBUS, DeviceNet, and BACnet to Ethernet protocols, letting legacy equipment keep running while new lines go native IP.

The right next step

Ethernet on the plant floor has changed more in the last 5 years than in the 15 before. Determinism is solved if you specify it. SPE is opening process plants to native IP. PoE is collapsing power and data into one cable. Convergence is real, with sharp edges.

The right next step is matching the right Ethernet — protocol, cable, switch, topology, security posture — to the job in front of you.