Ethernet-APL Reference Architecture: Two-Wire Ethernet for Process Automation (2026)

For forty years the process industry ran its field instruments on 4-20 mA analog loops, and for the last twenty it bolted digital fieldbuses on top of them. Both fell short of the same goal: real Ethernet, running all the way down to a temperature transmitter sitting in a Zone 0 hazardous area, over the same two wires that already carry its power. A sound ethernet-apl architecture finally closes that gap. It takes the 10BASE-T1L physical layer, wraps it in a rigorously defined intrinsic-safety envelope, and lets PROFINET, OPC UA, or EtherNet/IP reach the last meter of the plant without a gateway in the path.

This is not a marketing refresh of fieldbus. It is a different electrical and topological contract, with new components — power switches and field switches — and new planning rules for segment length, power budget, and explosion protection. Get those rules wrong and devices drop off the bus or, worse, an intrinsic-safety certificate becomes invalid.

What this covers: the 10BASE-T1L physical layer and IEEE 802.3cg, the 2-WISE intrinsic-safety concept from IEC TS 60079-47, a full trunk-and-spur reference architecture with power and field switches, hazardous-area zoning, protocol integration over APL, migration from 4-20 mA and fieldbus, and the segment-budget and failure-mode math you need before you specify hardware.

Context and Background

The physical layer has always been the process industry’s bottleneck. Factory automation moved to 100 Mbit/s industrial Ethernet — PROFINET, EtherNet/IP, EtherCAT — a decade ago, but those variants assume four-wire or two-pair cabling, mains-adjacent power, and non-hazardous cabinets. None of that survives a trip into a refinery’s Zone 1 rack row or a Zone 0 vessel nozzle. Field instruments stayed on 4-20 mA with HART riding as a low-bandwidth digital overlay, or on FOUNDATION Fieldbus H1 and PROFIBUS PA at 31.25 kbit/s. HART tops out around 1.2 kbit/s of usable digital throughput; H1 gives you 31.25 kbit/s shared across a segment. Neither can stream the diagnostics, waveforms, and asset data a modern digital twin expects.

Why now, and not a decade ago? Two things had to converge. First, single-pair Ethernet PHY technology matured to the point where 10 Mbit/s over 1000 m of a single twisted pair became electrically feasible with acceptable power. Second — and this is the harder problem — the intrinsic-safety community had to agree on a way to certify such links without drowning every installation in entity calculations. Analog loops and fieldbus each solved half the problem: loops gave you two-wire simplicity, fieldbus gave you a digital bus, but neither gave you high-bandwidth switched Ethernet into Zone 0. APL is the first field technology to hold all three properties at once, and it took a rare standards-body alliance plus a purpose-built safety concept to get there.

Ethernet-APL — Advanced Physical Layer — was launched in 2015 by a coalition that rarely cooperates: FieldComm Group, ODVA, and PROFIBUS & PROFINET International, working with the IEEE 802.3 group and a dozen instrument vendors. The deliverable was 10BASE-T1L, ratified as part of IEEE 802.3cg-2019, plus a companion safety concept (2-WISE, IEC TS 60079-47:2021) and an engineering guideline that turns the raw PHY into an installable network. The result is 10 Mbit/s full-duplex over a single twisted pair, power and data on the same conductors, reach of up to 1000 m on a trunk, and certified operation into Zone 0. That is roughly 320 times the digital bandwidth of an H1 segment on cabling with the same ruggedness and the same intrinsic-safety guarantees. For how this couples to information-model transport, our companion piece on the OPC UA FX field-exchange reference architecture covers the layer that rides on top; the IEEE 802.3cg standard is the authoritative physical-layer source.

The Ethernet-APL Reference Architecture

An ethernet-apl architecture is a two-level trunk-and-spur topology: a high-power trunk carries data and up to ~57.5 W of power from a power switch in the safe area out to field switches distributed in the hazardous area, and each field switch fans the trunk out into short, low-power, intrinsically safe spurs that terminate on individual field devices. Power flows outward on the same pair that carries bidirectional 10 Mbit/s data. Zoning, power class, and segment length are chosen per branch, not globally.

Figure 1: Layered Ethernet-APL reference architecture from controller and edge down to Zone 0 field devices.

Figure 1 shows the canonical flow: controller and OPC UA edge gateway connect through a plant backbone switch to an APL power switch in the safe area; the power switch drives a trunk of up to 1000 m to a field switch mounted in Zone 1 or Zone 2; the field switch splits into spurs of up to 200 m that reach the individual transmitters and valves, which may sit in Zone 0.

The trunk: high power, long reach, 2.4 Vpp

The trunk is the backbone of a field segment. It is a 10BASE-T1L link operating in the 2.4 Vpp transmit mode — the higher of the two amplitude modes 802.3cg defines — because it must drive up to 1000 m of cable and deliver enough power to run several downstream field switches. A single trunk power source port sources up to 57.5 W. In practice the trunk is not itself intrinsically safe in the general case; it is run in Zone 2 or the safe area, or protected by increased-safety (Ex e) or flameproof (Ex db) enclosures where it enters Zone 1. The trunk uses heavier cable (Type A fieldbus cable, ~0.8 mm² conductors) so that resistive drop over 1000 m still leaves adequate voltage at the far field switch.

The trunk carries one crucial architectural property: it aggregates. Every spur behind a field switch shares the trunk’s 10 Mbit/s pipe and its power budget. A trunk is not a point-to-point cable; it is a shared resource you must budget deliberately.

A trunk segment also has structural limits beyond raw length. The port-profile rules cap a trunk at a bounded number of terminal connections and permit at most a small number of auxiliary (in-line, trunk-powered) devices along its run. In practical terms you plan a trunk to reach a cluster of field switches in a marshalling location, then fan out locally, rather than snaking one trunk device-to-device across a whole plant. The trunk’s job is transport and bulk power; the fan-out and the intrinsic-safety conversion happen at the field switch.

The spur: short, intrinsically safe, 1.0 Vpp

A spur is a 10BASE-T1L link in the 1.0 Vpp transmit mode, reaching up to 200 m, and — this is the whole point — engineered so its stored and available energy is low enough to be certified intrinsically safe. The field switch converts the trunk’s non-IS energy into IS-limited energy per spur. Spur power classes for IS operation cap delivered power at roughly 0.5 W (Class A) or 1.0 W (Class C) at the field device, chosen so the ignition-energy limits of the target gas group are never exceeded even under fault. That power ceiling is why an APL field device’s electronics budget is measured in fractions of a watt, and why a two-wire radar level transmitter on APL looks a lot like its loop-powered ancestor in power discipline, but with 10 Mbit/s of bandwidth instead of a 4-20 mA needle.

Power switch and field switch as distinct roles

The two active components are not interchangeable. The APL power switch lives in the safe area (or a suitably protected enclosure), takes standard industrial Ethernet uplink and mains-derived DC, and energizes trunks. The APL field switch lives out in the field, draws its own power from the trunk, and redistributes energy-limited power to spurs. A field switch is typically an Ex-rated device: Zone 1 units use flameproof (Ex db) or increased-safety (Ex ec) protection for the switch electronics while presenting Ex ia spur ports. This division — non-IS bulk power in, IS-limited power out — is the structural heart of the design and the reason a single certificate can cover a whole segment.

Physical Layer, Power, and the 2-WISE Safety Concept

The deeper you go, the more the numbers matter. This section walks the 10BASE-T1L PHY, the power-over-two-wires mechanism, and the 2-WISE intrinsic-safety rules that make Zone 0 operation certifiable rather than merely hopeful.

Figure 2: Power distribution and intrinsic-safety energy-limiting from the 57.5 W trunk source to Ex ia field devices.

Figure 2 traces energy flow: the power switch sources 57.5 W into the trunk; an auxiliary or terminating field switch draws its own supply from that trunk; an IS barrier stage inside the field switch limits energy per spur; and Class A/C spurs deliver 0.5-1.0 W to Ex ia devices in Zone 0.

10BASE-T1L: the PHY in detail

10BASE-T1L is single-pair Ethernet at 10 Mbit/s, full duplex, defined in IEEE 802.3cg-2019 (which also standardized the shorter-reach 10BASE-T1S). It uses PAM3 line coding with a 4B3T-style mapping and a 7.5 MBd symbol rate, and it is designed for up to 1000 m of reach — an order of magnitude beyond ordinary short-reach automotive SPE. Two transmit amplitude modes exist: 1.0 Vpp for short, low-power (and IS) links, and 2.4 Vpp for long trunks that need signal margin over 1000 m. It is full duplex on one pair, using echo cancellation, so there is no half-duplex collision domain and no shared-media arbitration; every APL link is a switched point-to-point segment. That single fact — switched, full-duplex, per-device links — is what separates APL fundamentally from H1 and PA, which were multi-drop shared buses.

Because it is standard IEEE 802.3 Ethernet at the MAC layer, 10BASE-T1L carries ordinary Ethernet frames. Nothing above the PHY has to know it is running on a single twisted pair in a hazardous area. PROFINET, OPC UA, EtherNet/IP, HART-IP, and Modbus TCP all ride unchanged.

It helps to put the bandwidth in perspective. A FOUNDATION Fieldbus H1 or PROFIBUS PA segment runs at 31.25 kbit/s shared across every device on the bus, of which useful cyclic payload is a fraction after framing and scheduling overhead. A 10 Mbit/s APL link is switched and per-device, so each instrument gets the full pipe. That is a raw factor of roughly 320x in signalling rate, and much more than that in effective per-device throughput once you account for the shared-bus arbitration H1 imposes. The practical consequence is that an APL transmitter can stream high-resolution diagnostics — vibration spectra, echo curves from a radar level gauge, full HART command sets, firmware images — that were simply impossible to move over a 31.25 kbit/s shared segment. This is what makes field-level data rich enough to feed a digital twin rather than just a control loop.

Power and data on the same two wires

APL uses simultaneous power and data delivery: a DC bias carrying power is superimposed on the AC-coupled 10BASE-T1L signal on the same pair. The field device and switch each present a coupling network — inductors to pass DC while blocking the signal, capacitors to pass the signal while blocking DC. This is conceptually similar to loop power on 4-20 mA or the phantom power of PA/H1, but engineered for a 10 Mbit/s signal and a much larger power envelope. The port profile defines how much voltage and current a source may present and how much a load may draw, and — critically for IS — caps both.

2-WISE: intrinsic safety made segment-wide

The genius of the safety model is 2-WISE — 2-Wire Intrinsically Safe Ethernet, standardized as IEC TS 60079-47:2021. Classic intrinsic safety requires you to compute, for every pairing of source and load, whether the combination’s stored energy (in cable capacitance and inductance plus device reactances) can produce an ignition-capable spark or hot surface. The traditional “entity concept” makes you check four inequalities per drop: the source’s maximum output voltage Uo against the device’s maximum input voltage Ui, output current Io against Ii, output power Po against Pi, and — the tedious part — the source’s allowed external capacitance Co and inductance Lo against the sum of the device’s internal reactances plus the cable’s capacitance and inductance over its full length. Doing that pairwise for every instrument on a network, and re-doing it whenever anyone changes a cable run, is a combinatorial nightmare.

2-WISE replaces the pairwise proof with a small set of port profiles: each APL port declares a profile that fixes maximum supply voltage, current, and power, and a rule set for cable parameters. If a compliant source port connects to a compliant load port with cable inside the stated limits, the segment is intrinsically safe by construction — no per-installation entity calculation. The port profile has already done the worst-case energy math for a bounded cable length, so the installer’s job collapses to two checks: are both ends the right profile, and is the cable within the stated length and per-metre reactance limits. This is the practical enabler that lets an integrator wire a plant without an IS engineer re-certifying every drop, and it is why APL can be rolled out by ordinary instrument technicians rather than a scarce specialist. For a wider treatment of naming and addressing the resulting device fleet, see our unified namespace architecture guide. The authoritative source is IEC TS 60079-47:2021.

Hazardous-area zoning and Ex ratings

Figure 3: Hazardous-area zone mapping from Zone 0 field devices to the safe-area power switch.

Figure 3 maps the architecture to hazardous-area zones. In the ATEX/IECEx scheme, Zone 0 is where an explosive gas atmosphere is present continuously or for long periods, Zone 1 where it is likely in normal operation, and Zone 2 where it is unlikely and short-lived; the North American NEC Division scheme (Div 1, Div 2) is analogous. 2-WISE spurs achieving type of protection “ia” are suitable for Zone 0, Zone 20 (dust), and Division 1. The mapping is direct:

• Zone 0 / Div 1 — field devices on Ex ia spurs only. The spur’s energy limiting makes even a continuous gas presence safe.
• Zone 1 — field switches, typically Ex db (flameproof) or Ex ec electronics presenting ia spur outputs.
• Zone 2 — trunk cabling and auxiliary switches.
• Safe area — the power switch and the plant backbone.

Because the zone assignment is per branch, a single field switch can present some spurs into Zone 0 and route its trunk back through Zone 2 to a safe-area power switch, all under one coherent certification story.

Protocol and data flow over APL

Figure 4: Frame and power flow across an Ethernet-APL segment from field device to OPC UA controller.

Figure 4 shows a cyclic exchange. The field device brings up its 1.0 Vpp link to the field switch; the field switch forwards the frame over the 2.4 Vpp trunk to the power switch; the power switch hands it to the controller as an ordinary PROFINET or OPC UA frame; the controller’s cyclic setpoint returns by the reverse path; and power flows continuously in the opposite direction over the same pairs. Because APL is switched Ethernet, PROFINET over APL supports RT cyclic exchange and, on capable switches, TSN-based scheduling — the same real-time toolbox described in our PROFINET IRT and TSN guide. OPC UA can run as a server on the device itself or as PubSub over the segment. HART-IP lets legacy HART payloads tunnel to modern asset-management hosts without a multiplexer.

A worked segment budget

The abstract limits become concrete only when you count. Consider a trunk feeding four field switches. Each field switch draws roughly 5-8 W to run its own electronics and PHYs, and presents four spurs. If each spur delivers a Class C 1.0 W IS budget to its device, that is 4 W of spur load plus, say, 6 W of self-draw, for about 10 W per field switch. Four such switches demand roughly 40 W. Against a ~57.5 W trunk source, that leaves only ~17 W of headroom — and headroom is exactly what resistive drop over a long trunk consumes. Push to five field switches (~50 W) at 1000 m and you are riding the ceiling with almost no margin for cable loss or ageing. This is why the FieldComm Engineering Guideline publishes segment tables rather than a single “1000 m, full power” figure: the achievable combination of length, number of field switches, and per-spur power is a surface, not a point.

The bandwidth budget is usually the easier constraint. A trunk with four field switches and sixteen spurs is sixteen devices sharing 10 Mbit/s. Even if every device streamed 100 kbit/s of cyclic and diagnostic data, that is 1.6 Mbit/s of the 10 available — comfortable. Power, not bandwidth, is the binding constraint in nearly every real APL design. Plan the watts first; the bits look after themselves.

Comparing APL against what it replaces

Attribute	4-20 mA + HART	FOUNDATION Fieldbus H1 / PROFIBUS PA	Ethernet-APL
Digital rate	~1.2 kbit/s (HART)	31.25 kbit/s shared	10 Mbit/s switched per device
Topology	Point-to-point loop	Multi-drop shared bus	Trunk-and-spur, switched
Power + data on 2 wires	Yes (loop)	Yes	Yes
Reach	~1-2 km loop	~1.9 km segment	1000 m trunk + 200 m spur
Intrinsic safety	Barriers, entity calc	FISCO/FNICO	2-WISE port profiles, Zone 0
Native Ethernet protocols	No (needs gateway)	No (needs linking device)	Yes (PROFINET, OPC UA, EtherNet/IP)
Gateway to control/asset mgmt	Required	Required	None in path

The decisive rows are the last three: APL keeps the two-wire, loop-powered, intrinsically safe ergonomics the process industry already trusts, while removing the gateway and multiplying bandwidth by orders of magnitude. That combination is what no prior field technology offered.

Trade-offs, Gotchas, and What Goes Wrong

APL is powerful, but it trades away some of fieldbus’s simplicity, and several failure modes are specific to it.

Trunk oversubscription. The trunk is a shared 10 Mbit/s pipe and a shared ~57.5 W budget. Hang too many field switches and spurs off one trunk and you can run out of power long before you run out of bandwidth. A field switch drawing, say, 5-8 W for itself plus 4 spurs at 1 W each is ~10-12 W; five such switches on a trunk approach the source ceiling. Budget power first, bandwidth second — the opposite instinct from IT Ethernet.

Cable drop over 1000 m. Resistive voltage drop on the trunk is real. At 1000 m on Type A cable the far-end field switch sees materially less voltage than the near end. If you also load that far switch heavily, it may brown out. The engineering guideline’s segment tables exist precisely because length and load interact non-linearly; do not treat 1000 m and full power as simultaneously achievable without checking the budget.

IS is only valid within the port profile. 2-WISE’s by-construction safety holds only if every port on a spur is profile-compliant and the cable stays within the stated capacitance/inductance-per-length and total-length limits. Splice in the wrong cable, exceed the spur length, or connect a non-compliant device and you have silently invalidated the certificate. The convenience of skipping entity calculations is also a trap for the careless.

No multi-drop on a spur — mostly. Classic H1/PA let you daisy-chain many devices on one segment. APL spurs are generally one device per spur (short multi-drop variants exist but are the exception). This raises field-switch port count and cabling versus a shared bus. You gain determinism and per-device diagnostics; you spend ports.

Redundancy must be designed in. A trunk failure orphans every field switch and device behind it. Media Redundancy Protocol (MRP) rings, dual power switches, and ring-capable field switches address this, but ring topologies complicate the IS story and the power budget. Redundancy is not free at the physical layer.

Brownfield coexistence. You will rarely rip out an entire fieldbus plant at once. APL and legacy 4-20 mA/HART or PA segments coexist during migration, which means dual toolchains, dual spares, and dual competencies for years.

Grounding, shielding, and EMC discipline. Single-pair Ethernet at 7.5 MBd over 1000 m in an electrically noisy plant is unforgiving of sloppy shielding and grounding. The engineering guideline is specific about cable shield termination and single-point versus multi-point grounding because ground loops and pickup that a 4-20 mA loop would shrug off can raise the bit-error rate on a 10 Mbit/s link. A segment that “works on the bench” and fails intermittently in the field is almost always an EMC or grounding defect, not a protocol bug. Budget commissioning time for it.

Diagnostics are richer but so is the failure surface. Every APL device is now an addressable Ethernet node with a MAC, a link state, and often an on-board web server or OPC UA server. That is wonderful for diagnostics and dangerous for security: the attack surface of the field layer just grew from “a c