Ethernet-APL vs 4–20 mA: The Real Economics of Brownfield Migration
The slide in the project review said what Ethernet-APL slides usually say: fewer cabinets, faster commissioning, lower cost. The question from the plant manager was shorter. “Lower than what?”
It is the right question, and the honest answer does not fit on one slide. Lower than a classic 4–20 mA installation with marshaling cabinets and multicore homerun cables? In the most-cited cost study so far, yes, by roughly ten percent of total investment. Lower than 4–20 mA/HART wired into remote I/O in the field? About equal. Lower than leaving a working analog unit alone for another ten years? Often not at all.
Brownfield is where this decision actually gets made. At a NAMUR update meeting in Ludwigshafen in June 2026, NAMUR chairman Tobias Schlichtmann of BASF described the European task as getting more out of existing plants, and BASF automation engineer Kai Krüning pointed out that brownfield migration needs a clear business case — one that is often still missing (CHEMIE TECHNIK report).
This article is about building that business case without cheating on either side: where 4–20 mA remains the rational answer, where Ethernet-APL starts to pay back, and how to read the published numbers without turning them into something they never claimed. It is written from the device side of the transition, based on our project experience with 10BASE-T1L/Ethernet-APL field devices as well as 4–20 mA and HART devices.
Teams typically assume: Ethernet-APL is cheaper than 4–20 mA, so a migration will pay for itself once the devices are available.
In reality: Ethernet-APL is roughly cost-neutral against well-designed analog installations on investment, clearly cheaper than legacy homerun-and-marshaling designs, and its biggest savings sit in engineering and commissioning — phases a brownfield migration only partly repeats. Whether it pays back depends on which cables must be replaced anyway, whether the control system already has an Ethernet path to the field, and how much the plant will actually use the extra diagnostics.
Quick Overview
Question: When does replacing 4–20 mA/HART loops with Ethernet-APL make economic sense in an existing plant?
Short answer: When the migration coincides with work that has to happen anyway — a control system upgrade, exhausted marshaling space, end-of-life fieldbus segments, or a unit extension — and when the device count per area is high enough to fill field switch ports.
Key evidence: TH Köln PLT lab study (1,570 signals): Ethernet-APL at cost parity with remote I/O and about €600,000 below a €6.2 million homerun design; BASF Ludwigshafen: about 28% less loop-check time on two world-scale plants.
Where 4–20 mA still wins: Small additions with spare I/O, safety loops for now, units with short remaining life, and control systems with no Ethernet-APL path.
What Is Actually Being Compared
Most “Ethernet-APL vs 4–20 mA” arguments compare two architectures. There are three, and leaving out the middle one is how the economics get exaggerated.
Classic 4–20 mA/HART with marshaling. Each transmitter runs a shielded pair to a field junction box, a multicore homerun cable carries the signals to the control room, and a marshaling rack cross-wires them to I/O cards. This is the installed base of most older plants.
4–20 mA/HART into remote I/O. The analog loops terminate in remote I/O stations in the field, which talk to the controller over a fieldbus or Ethernet. The homerun and much of the marshaling disappear, but every loop is still wired, documented and checked individually.
Ethernet-APL. Field devices connect point-to-point over spurs to APL field switches; the switches connect upward over a powered trunk or fiber. Discrete signals without an APL equivalent still need remote I/O.
The physical layer underneath is 10BASE-T1L from IEEE 802.3cg: full-duplex 10 Mbit/s Ethernet over a single twisted pair. Ethernet-APL adds what process plants need on top — port profiles and power classes (IEC TS 63444), the 2-WISE intrinsic safety concept (IEC TS 60079-47), and trunk-and-spur topology with trunks up to 1,000 m and spurs up to 200 m on the best cable category. A 4–20 mA loop, by comparison, carries one process variable plus a HART digital channel at 1,200 bit/s.
Power is the less-discussed difference. Analog Devices puts the power available to a typical 4–20 mA field device at around 36 mW (ADI). An Ethernet-APL Class A spur delivers about 0.54 W at the device, Class C about 1.1 W, and the second edition of IEC TS 63444 added Power Class B for devices up to about 1.16 W, as the SDOs announced in November 2025 (FieldComm Group). That is an order of magnitude more headroom — and still not enough for every instrument, a point we return to below. For the physical-layer details, see our breakdown of 10BASE-T1L power classes, PHY options and migration paths.
The Study Everyone Quotes — and What It Actually Says
The number that circulates in presentations comes from the PLT laboratory at TH Köln (Prof. Norbert Große). It is often attributed to BASF or Krohne, because it was presented again at the June 2026 NAMUR update that Krohne and ten partner companies hosted at BASF’s Gesellschaftshaus in Ludwigshafen. The study itself was commissioned by Bayer, Covestro, Currenta and Lanxess, used a reference quantity structure of 1,570 signals defined by the NAMUR APL Task Force, and was published in atp magazin 5/2025 (atp summary; CHEMIE TECHNIK).
It compared exactly the three architectures above and counted hardware, control room, engineering, installation, integration and commissioning.
| Cost block | Result reported by the study | Why it matters for brownfield |
|---|---|---|
| Hardware | Ethernet-APL about 5% more expensive than remote I/O (device premium plus field switches); homerun variant most expensive | In a migration you pay the device premium on every replaced instrument |
| Control room | Homerun variant far more expensive than remote I/O or APL | Only a saving if cabinet space is needed or the I/O is obsolete; existing racks are sunk cost |
| Engineering | APL cheaper: no loop diagrams, polarity-insensitive wiring, an instrument list defining switches and ports is enough | Partly repeated in brownfield: you still re-engineer every migrated loop once |
| Installation | Small difference between remote I/O and APL; APL requires shield grounding at both ends | Old plants may need bonding work before APL cable is allowed in |
| Integration | APL clearly cheaper, driven by data rate and uniform device access | Requires a DCS/PLC with a PROFINET (or other Ethernet) path to APL |
| Commissioning | Largest relative saving: about 85% with APL | Brownfield commissions only the migrated loops, often inside a turnaround window |
| Total | APL at parity with remote I/O; both almost €600,000 below the homerun design at about €6.2 million | The gap is mostly the price of the homerun and marshaling you may already own |
Read correctly, the study supports four statements and not more.
- It is an investment comparison. Operating benefits — diagnostics, device replacement, NOA data access — were listed but not priced. TH Köln plans to monetize them in a follow-up with pilot users, including safety functions on a reference plant enlarged by 20%.
- The headline saving is shared. Both remote I/O and Ethernet-APL beat the homerun design by roughly the same amount. Most of the €600,000 comes from removing multicore cables and marshaling, not from Ethernet as such. Against a good analog design, APL is cost-neutral, not cheaper.
- The scale is modest per signal. On our own arithmetic, the homerun variant works out to roughly €3,950 per signal and the gap to about €380 per signal, around 10%. These are reference-plant figures, not transferable prices, but they show how easily site-specific costs can swallow the difference.
- The reference is a new build. A 1,570-signal quantity structure priced from scratch has no sunk cable, no racks with spare slots and no outage constraints. A brownfield unit has all three.
None of this makes the study weak. atp columnist Thomas Tauchnitz drew the reasonable conclusion from it: new plants and larger extensions should go APL, because the engineering and commissioning advantage repeats with every later change while device premiums shrink (atp). That is a greenfield-and-expansion recommendation. It is not a proof that every analog loop in an operating plant should be replaced.
Field Evidence From BASF: Ludwigshafen and Zhanjiang
Studies model; plants report. The two BASF projects discussed at the same meeting are useful precisely because they include the costs that models leave out.
In Ludwigshafen, BASF maintenance manager Philipp Beckmann described the start-up of two new world-scale plants with around 1,750 Ethernet-APL and PROFIBUS PA field devices in an MRP ring. The plants started on schedule and under budget. Loop-check time fell by about 28%, installation errors dropped, and a topology scan showing every device at its switch and port replaced classic loop and terminal diagrams. Device replacement became plug-and-play through neighborhood detection, and troubleshooting moved from the terminal strip to software (CHEMIE TECHNIK).
The same report lists the other side. The young technology needed numerous device firmware updates. Engineering shifted toward system understanding, with a high entry barrier and real training needs in PROFINET and networking. And the safety functions of those plants were still executed in classic 4–20 mA.
Note what the 28% is: loop-check time on a mixed APL/PROFIBUS PA installation. It is not the 85% commissioning saving from the TH Köln model, and the two numbers should not be quoted as if they confirm each other one-to-one. They point in the same direction; they measure different things.
BASF’s greenfield site in Zhanjiang, China, adds a second lesson. Loop-check effort was visibly lower, but some planning assumptions did not hold — not because of Ethernet-APL, but because more was wired conventionally on site than planned. The savings scale with the share of devices that are actually on APL. A plan that assumes 80% APL and builds 50% will not see the modeled result.
Where the CAPEX Really Sits in a Brownfield Migration
A brownfield comparison is not APL price versus analog price. It is the incremental cost of migration versus the incremental cost of staying, on top of an installation that already exists. The line items look like this.
| Cost item | Stay on 4–20 mA/HART | Migrate to Ethernet-APL |
|---|---|---|
| Field devices | Sunk; replace like-for-like at end of life | APL device premium on every replaced instrument; four-wire devices still need separate power |
| Device-to-junction-box cable | Existing | Reusable only if it passes requalification against the APL cable categories |
| Multicore homerun | Existing | Cannot carry APL; replaced by a powered trunk or fiber to field switches |
| Marshaling and I/O cards | Existing, possibly with spare capacity | Freed — a real saving only if space is short or the I/O is obsolete |
| Field switches and power switch | None | New, with port reserve, Ex rating matched to the zone, auxiliary power if fiber-fed |
| Remote I/O for discrete signals | Existing | Still required for binary signals without an APL equivalent |
| Control system interface | Existing | Controller may need a PROFINET/APL interface or version upgrade |
| Engineering documentation | Loop diagrams maintained | Switch and port list; one-time re-engineering of migrated loops |
| Commissioning | Per-loop checks for new loops only | Topology scan plus functional checks, usually inside a turnaround |
| Lifecycle items | HART tools, calibration routines | Firmware update management, certificates and security updates, network training |
Two rows decide most brownfield cases. The first is the homerun: if the multicore cable and marshaling must be replaced anyway — corrosion, capacity, a new control room — the migration inherits most of the study’s savings. If they are healthy and have spare capacity, the migration has to pay for replacing something that works. The second is the controller: the Ethernet-APL Engineering Guideline notes in each migration scenario that the controller needs a version with an Industrial Ethernet interface. If a DCS upgrade is already planned, that cost is shared; if not, it lands on the APL budget.
Field instruments add their own constraint. Krohne, for example, introduced Ethernet-APL first on its four-wire MFC 400 converter for the OPTIMASS X400 Coriolis meter, explaining that the APL power budget, while larger than 4–20 mA, does not yet cover the full two-wire range for the largest diameters and hardest fluids (Krohne). A four-wire instrument keeps its separate power cable whichever network it speaks.
Can You Reuse the Existing Two-Wire Infrastructure?
Sometimes. This is the most oversold point in Ethernet-APL marketing, and the guideline itself is more careful than most brochures.
10BASE-T1L does not specify a cable; it specifies a channel through insertion loss and return loss. The reference cable for Ethernet-APL is fieldbus type A cable per IEC 61158-2, the shielded twisted pair used for PROFIBUS PA and FOUNDATION Fieldbus H1. The guideline defines four APL cable categories. Category IV allows 1,000 m trunks and 200 m spurs; categories I to III drop to 750/150, 500/100 and 250/50 m. For installed cable, it requires a measurement per ISO/IEC 11801-3, with a correction factor applied for spurs, and advises checking the age of the cable before reuse (Engineering Guideline v1.14).
Applied to the three installed-base scenarios in the guideline, that produces a clear pattern:
- FOUNDATION Fieldbus H1 and PROFIBUS PA plants: the installed type A cable is the strongest candidate for reuse, because its parameters are close to the APL reference cable — but it still has to be measured and requalified, not assumed fit. Junction boxes are a separate question — FieldComm Group notes that APL switches tend to be larger than H1 couplers, so some enclosures will not take the new hardware (FCG migration concept).
- Central I/O with marshaling: APL signals must not be routed through multicore bulk cables, so the homerun is replaced. The individual device cables can stay only if they meet a cable category. Older analog installations frequently used unshielded pairs, which will not.
- Remote I/O on PROFIBUS DP: the DP backbone cable has a different impedance and cannot be reused for APL. Device cables again depend on measurement.
Two practical items catch brownfield projects late. A powered trunk loses voltage over distance, so its real maximum length depends on load and conductor cross-section, not only on the 1,000 m category limit. And the study’s note on grounding matters: APL cabling calls for shield bonding at both ends, which in older plants raises equipotential bonding questions that a single-ended analog shield never did.
On the Ex side, there is a real engineering simplification — within limits. For compatible 2-WISE devices and segments, intrinsic safety verification becomes a port-to-port check: the guideline treats a powered spur as intrinsically safe when it connects one 2-WISE source port, one 2-WISE load port, up to two 2-WISE auxiliary devices and a compliant cable of up to 200 m, with the lowest protection level in the chain defining the system. That removes much of the entity calculation work analog and FISCO segments required. It is not automatic, though: the guideline explicitly notes that a permitted combination of power classes does not by itself mean the combination is permitted from an intrinsic safety viewpoint. That still has to be checked under the 2-WISE concept and documented for the installation.
So “reuse the existing two-wire infrastructure” usually means reusing the last few hundred meters in a fieldbus plant, sometimes reusing device cables in an analog plant, and almost never reusing the homerun.
Diagnostics and Maintenance: The TCO Items Nobody Has Priced Yet
HART already offers device diagnostics. The honest comparison is not “no diagnostics versus diagnostics” but “diagnostics that require a handheld, a multiplexer or an asset management layer” versus diagnostics that arrive on the same network as the process value, mapped to NAMUR NE 107 status signals and a uniform device model (PA-DIM).
Where this changes cost is in recurring work. At the Ludwigshafen meeting, Beckmann described a recurring safety proof test today: a technician connects a diagnostic tool at the device and works through vendor-specific HART menus, easily 20 minutes per sensor. With PROFIsafe over Ethernet-APL that check could be automated, and a device in a control loop could later be configured into a safety function without new wiring. That is a significant lever — and it is a future one, because those same plants still run their safety loops on 4–20 mA.
The counter-items are also operating costs. Firmware updates were the main teething cost in Ludwigshafen, so update management has to be budgeted. Evonik’s Thomas Scherwietes pointed at the IT side: once data flows freely, plants need automated certificate management for thousands of field devices and security updates in line with the EU Cyber Resilience Act. Our article on IEC 62443 and OT/IT convergence at the device level covers what that means for embedded architecture.
The practical rule: treat operating benefits as upside unless you can quantify them for one asset class. Valve positioners, Coriolis meters and analyzers generate diagnostics that maintenance teams already pay to collect. A temperature transmitter on a stable loop usually does not.
Greenfield vs Brownfield: Same Technology, Different Answers
| Factor | Greenfield or major extension | Brownfield migration |
|---|---|---|
| Cable and marshaling | Priced from zero; APL avoids them | Often sunk; saving only if replacement is due |
| Engineering and commissioning | Full saving applies | Applies only to migrated loops |
| Control system | Specified with APL from the start | Upgrade may be required and must be justified |
| Coexistence | Minimal | Analog, HART, fieldbus and APL side by side for years |
| Outage constraints | None before start-up | Cutover limited to turnarounds or hot-swap windows |
| Typical verdict | APL is often the stronger option for new plants and major extensions | Case-by-case, driven by triggers |
In brownfield, the economics flip when a trigger arrives that forces spending anyway: a DCS replacement, marshaling capacity exhausted, fieldbus hardware at end of life, I/O card obsolescence, a unit extension, or a digitalization program that needs field data the analog loops cannot deliver. Without a trigger, a migration competes against a working installation with zero incremental cost, and it usually loses.
When 4–20 mA Is Still the Rational Choice
- Adding a handful of instruments to a unit with spare I/O and marshaling capacity. The incremental cost of another analog loop is close to the device alone.
- Safety instrumented loops, until PROFIsafe over Ethernet-APL is proven on your site with the SIL-rated devices you need. The plants that pioneered APL still kept safety on 4–20 mA.
- Control systems with no PROFINET or APL path and no upgrade planned in the unit’s horizon.
- Units with a short remaining life, where there is no time to recover engineering and training costs.
- Aged or unshielded field cabling that would need replacing, combined with a healthy homerun — you would be paying for new cable and new infrastructure at once.
- Teams without network skills and without a training budget. BASF named the learning curve openly; smaller sites feel it more.
Where Ethernet-APL Starts to Pay Back
- New units and larger extensions inside an existing site, where cable and cabinets would be bought new anyway.
- Dense device clusters. A field switch with 8 to 24 spur ports amortizes well; a switch serving three instruments does not.
- FOUNDATION Fieldbus H1 or PROFIBUS PA segments due for renewal, where the type A cable is a strong reuse candidate after requalification, and the move replaces couplers rather than cables.
- Exhausted marshaling or control room space, where APL releases capacity that would otherwise cost a building extension.
- High commissioning volume in a short window, where loop-check time is the schedule risk.
- Assets whose diagnostics are already worth money: valve positioners, Coriolis meters, analyzers, rotating-equipment monitoring.
- Units that change often. Every later modification repeats the engineering and commissioning advantage.
A Practical Decision Sequence for a Brownfield Unit
Inventory and classify every signal
Analog, HART-used, fieldbus, discrete, safety-related, and Ex zone. The APL share you can realistically reach defines the ceiling on savings.
Check the control system path
Does the controller support PROFINET or another Ethernet protocol toward APL, and is an upgrade already planned?
Measure the cable, do not assume it
Requalify candidate cables against the APL categories and record the age and shielding. Budget replacement for what fails.
Choose the right network topology
A powered trunk from a power switch, or fiber to field switches with local auxiliary power. The choice changes cable, power and Ex design.
Price three variants, not two
Staying analog, analog into remote I/O, and Ethernet-APL. If APL only beats the homerun design, you have learned that the homerun is the problem.
Separate CAPEX from operating upside
Justify the investment on CAPEX and triggered spending; show diagnostics and proof-test automation as quantified upside where you can.
Plan coexistence and cutover
Loop-by-loop or area-by-area replacement inside turnarounds, with analog and APL devices running side by side.
Budget the device lifecycle
Firmware updates, certificate handling, security patches, and training are part of the TCO from day one.
What the Plant-Side Economics Mean for Field Device Manufacturers
Plants migrate loop by loop, over years. For instrument makers that means selling both worlds from one product line: a 4–20 mA/HART variant for the unit that stays analog, and an Ethernet-APL variant for the extension next door — ideally around the same measurement core, calibration and firmware base.
That is the kind of Ethernet-APL device development Promwad does from inside the device. On the connectivity side, our team integrated the Analog Devices ADIN1110 10BASE-T1L MAC-PHY into an industrial predictive maintenance sensor, from schematic through bring-up to a working PROFINET and OPC UA stack on FreeRTOS. On the legacy side, we have built devices around 4–20 mA and HART: an electromagnetic flow meter on an STM32F7 that combines PROFINET, HART and Modbus RTU in one device, and FreeRTOS firmware that added HART and Modbus RTU to an existing interface module for WIKA Italia.
The plant economics above translate into concrete device requirements:
- Power budget discipline. Class A leaves about half a watt for PHY, MCU, sensor front end and diagnostics. Designing to Class B or C widens the product range but narrows compatibility with installed switches.
- A robust update path. BASF’s biggest teething cost was device updates. A secure bootloader with fail-safe updates (MCUboot in our designs) is part of the customer’s TCO, not a firmware nicety.
- Diagnostics that map cleanly. NE 107 status and a PA-DIM-friendly data model are what make the operating benefits real for the plant.
- Ex and conformance planned early. 2-WISE compatibility and protocol conformance testing shape the hardware and the schedule, so they belong in the architecture phase, not at the end of the project.
Related projects: Electromagnetic flow meter design with PROFINET, HART and Modbus RTU; HART and Modbus RTU firmware for an industrial interface module.
Adding Ethernet-APL to an instrument that already sells on 4–20 mA?
Real Trade-Offs to Expect
- Reusing cable vs replacing it. Reuse saves installation money but adds measurement cost and the risk of aging cable failing later. A cable that barely passes category II today limits you to 500 m trunks and 100 m spurs forever.
- Powered trunk vs fiber with auxiliary power. A powered trunk needs no field power but limits length and device count through voltage drop. Fiber reaches further and supports rings, but every field switch needs local power.
- Area-by-area vs loop-by-loop. Migrating whole areas fills switch ports and captures commissioning savings; loop-by-loop fits turnaround windows but leaves partly used switches and mixed documentation.
- Early adoption vs waiting. Early projects absorb firmware updates and training costs; waiting keeps the analog cost structure and delays the data. Device maturity improved markedly since the first pilots, but the learning curve at site level remains.
- Dual-variant vs APL-only products (for OEMs). Two variants protect sales into analog plants at the cost of two certification and test paths. APL-only is simpler to maintain and arrives too early for much of the installed base.
You May Be Facing This Decision If:
- ✓ Your marshaling cabinets or control room are out of space and an extension is being discussed.
- ✓ A DCS upgrade or replacement is on the roadmap for the next three to five years.
- ✓ FOUNDATION Fieldbus H1 or PROFIBUS PA hardware in a unit is approaching end of support.
- ✓ A business case shows APL cheaper than 4–20 mA, but nobody has priced the remote I/O alternative.
- ✓ Your cable records do not say whether the installed pairs are shielded, what type they are, or how old they are.
- ✓ Maintenance spends real hours on handheld HART diagnostics and proof tests.
- ✓ Your product roadmap includes an Ethernet-APL variant, but it is unclear how quickly your customers’ plants will actually migrate.
The Short Version
Ethernet-APL is not always cheaper than 4–20 mA, and nobody who has read the numbers carefully claims it is. On investment, it matches a well-designed remote I/O installation and beats a legacy homerun design by around ten percent, with the largest gains in engineering and commissioning. In brownfield those gains are real but partial, and they arrive when the migration rides on spending that was coming anyway. The operating benefits — diagnostics, faster device replacement, and eventually automated safety proof tests — are the reason to go, but they still need to be priced for your plant rather than borrowed from a slide.
For instrument makers, the conclusion is simpler: plants will buy both interfaces for years, and the products that win are the ones designed for that coexistence from the start. That is the focus of our Ethernet-APL & 10BASE-T1L field device engineering work — from MAC-PHY selection and power budget to PROFINET, OPC UA and NE 107 diagnostics.
FAQ
Is Ethernet-APL cheaper than 4–20 mA?
Can I reuse my existing 4–20 mA cables for Ethernet-APL?
Who actually did the Ethernet-APL cost study, BASF or Krohne?
What savings did BASF report from real Ethernet-APL plants?
Should safety instrumented functions move to Ethernet-APL now?
Does a field device manufacturer need both 4–20 mA and Ethernet-APL variants?
Related Promwad Expertise
- Ethernet-APL & 10BASE-T1L Field Device Engineering: Promwad’s commercial page: MAC-PHY selection, PCB, firmware, PROFINET and OPC UA for field devices.
- Two Wires to the Field: What 10BASE-T1L Changes for Industrial Connectivity: physical layer, power delivery and protocol background.
- Electromagnetic Flow Meter Design: PROFINET, HART and Modbus RTU in one industrial instrument.
- Firmware Development for Industrial Protocols: HART and Modbus RTU for a WIKA Italia interface module.
- Explosion-Proof Device Development (ATEX, IECEx): hardware development with hazardous-area requirements for process automation.
- Industrial Sensor Development: sensors and transmitters with 4–20 mA, HART, IO-Link and single-pair Ethernet.
Plan Your Ethernet-APL Device Roadmap
Share the instrument, its current interfaces, the plants you sell into, and where the Ethernet-APL question sits on your roadmap. We will help define the next engineering step — from a connectivity architecture review to a device prepared for certification.