Single Pair Ethernet

Key takeaways

• Single Pair Ethernet (SPE) is Ethernet over one twisted pair instead of two or four. The frames are ordinary Ethernet frames; only the physical layer changes.
• SPE is a family of IEEE standards, not one speed: 10BASE-T1S and 10BASE-T1L (802.3cg), 100BASE-T1 (802.3bw), 100BASE-T1L (802.3dg, standardised June 2026) and 1000BASE-T1 (802.3bp). Connectors are defined separately in IEC 63171.
• IEEE 802.3bw specifies 100BASE-T1 for 15 m unshielded and 40 m shielded. Perinet measurements between two periCORE Development Boards reached 180 m over Cat 5e and data-grade telephone cable, 95 m over KNX cable and 100 m over unshielded doorbell wire.
• Power can travel with the data, either on the data pair (PoDL, IEEE 802.3bu) or on extra conductors in a hybrid cable. It is not automatic and has to be specified.
• SPE is transport. IP addressing, MQTT, security and structured data come from the device's network stack, not from the cable. Perinet's components supply that stack.
• Existing fieldbuses and PLCs stay in place. SPE adds a direct path for field data into IT systems next to them.
• Perinet's SPE portfolio covers the sensor-to-IT path end to end: periNODE adapters for analogue, digital and Modbus sensors, periSWITCH and periSTART for the network, periSNOOP for existing 4-20 mA loops, periMICA at the edge and periCORE for building SPE connectivity into your own device.

What is Single Pair Ethernet?

Single Pair Ethernet is Ethernet transmitted over one twisted pair, which allows smaller cables and connectors to bring Ethernet connectivity closer to sensors, actuators and other field devices.

More precisely, SPE is a family of Ethernet physical layer standards that carry Ethernet over one twisted pair rather than the two or four pairs used by conventional Ethernet. Depending on the standard and the cabling, the same connection can also supply power.

Two sentences worth keeping in mind for everything below, because most SPE explainers blur them:

SPE is transport. It carries Ethernet frames to places Ethernet could not previously reach.

Everything above transport is a separate layer. IP addressing, MQTT, security and structured data do not arrive with the cable. They come from what the device itself contains, which is where Perinet's components do their work.

Why SPE exists

In most installations, Ethernet means RJ45, and RJ45 is typically IP20. That one fact explains most of why the field level never got Ethernet. Industrial-grade Ethernet connectors are either too large for a compact device or too expensive to justify on one, and four-pair cable is thick, stiff and awkward to route to something that reports a single value once a second.

So the field level developed its own networks. Fieldbuses did a genuinely good job of the thing they were built for: getting deterministic signals to a controller over cheap two-wire cabling. Millions of installed devices still do exactly that, reliably, and there is no reason to rip them out.

What they were not built for is moving field data directly into IT systems. Doing that means adding a conversion step, and every conversion step costs engineering hours, adds a component to maintain, and gives data one more chance to lose its meaning on the way.

SPE takes a different route. Two wires are easy to terminate, which is why SPE connectors can be small enough for a compact device, or wired in the field by hand, the way Modbus RTU wiring always was. It brings the Ethernet boundary out of the control cabinet and closer to the device.

How SPE works

SPE transmits Ethernet over a single twisted pair. The frames are ordinary Ethernet frames; what changes is the physical layer that carries them.

That single change has knock-on effects worth understanding:

• Cabling gets smaller and lighter. One pair instead of four means thinner cable, a tighter bend radius and smaller connectors, which can also reduce installation and cable-routing costs.
• Topology no longer has to be a star. Conventional Ethernet gives you one switch port per device and about 100 m to reach it. With switched SPE devices, the connection can continue from device to device, so lines run in daisy chains and trees instead of every run returning to a cabinet. Distance matters less here than people expect: 10BASE-T1L stretches to roughly 1,000 m, but the last few metres to a sensor rarely need a kilometre. What they need is a route that does not demand its own port and its own conduit.
• Ethernet reaches closer to the device. Where a device includes the necessary networking stack, it can be addressed and managed with standard Ethernet and IP tooling rather than through a bus-specific master.

Speed, distance and topology vary by standard, and the trade-offs are real. There is no single "SPE speed."

What SPE provides, and what Perinet adds

SPE gets Ethernet to the sensor. It does not make the sensor an IP device. Here is what that means in practice.

An SPE cable reaching a sensor does not make that sensor an IP device. A conventional analogue, digital or serial sensor has no network stack; it produces a signal. SPE gives that signal a physically viable path. Something still has to give it an address, a protocol and a security identity.

What SPE provides

• Ethernet over one twisted pair
• Smaller cabling and connections that suit compact field devices
• A standards-based physical and data-link foundation
• A choice of speeds and network architectures

What Perinet adds

• A full network stack, so a device becomes addressable and manageable rather than a signal source
• Per-device security and identity
• MQTT and HTTPS REST at the device itself
• Interfaces for the sensors people actually have: analogue, digital and Modbus
• A direct route for field data into IT systems

Stated plainly: SPE makes Ethernet physically practical at the field level. Perinet makes the connected data usable.

How far does 100BASE-T1 actually reach?

The reach figures in the standards table are design points rather than physical limits. The 15 m for 100BASE-T1 was set for automotive wiring harnesses, with a 40 m shielded case for industrial use; the 1,000 m for 10BASE-T1L was set for process plants. Each describes the conditions the standard was written against.

Industrial buildings are already full of cable that was never specified for any of this. So the useful question is what SPE does on it.

In Perinet testing (periCORE Cable Performance Application Note PRN.100.900 rev 2, August 2026), 100BASE-T1 remained stable at 180 m over Cat 5e and data-grade telephone cable, 95 m over KNX cable and 100 m over unshielded doorbell wire (YV 2 × 0.6 mm, two solid copper conductors, PVC insulation), though doorbell wire has no standardised impedance so that figure varies by product and should be treated as indicative. Those measurements were taken between two periCORE Development Boards; the endpoint devices matter, and the same Cat5e measured 90 m with a periMICA at one end. Switched devices can create additional segments.

Full method, test setup and all measured cable types: periCORE Cable Performance Application Note.

Two details that trip people up:

IEEE 802.3cg defines both 10BASE-T1S and 10BASE-T1L. They are not interchangeable. T1S is short-reach and can run multidrop; T1L is long-reach and point-to-point.

The letters mean something. In a name like 100BASE-T1, the "T1" identifies twisted-pair Ethernet over a single pair. In 10BASE-T1L, the L marks the long-reach variant; the S in 10BASE-T1S marks the short-reach one.

For connector detail, see the SPE connector mating faces documentation.

SPE topologies

Topology is where SPE stops being a single thing and becomes an architectural choice.

Point-to-point

One device, one cable, one port. Simple, dedicated, and the basis of most higher-speed SPE deployments.

Multidrop

Several devices share one line. It is a feature of 10BASE-T1S, and it means devices share bandwidth and run half-duplex. A shared line also needs proper bus termination, which is a footnote on a drawing and a recurring nuisance in a running plant.

Switched branching

A small switch sits out in the field instead of every cable routing back to a cabinet. periSWITCH is a three-port Single Pair Ethernet switch that acts as a branch point close to the devices, so a line can fan out where the machine actually is. 100BASE-T1 is point-to-point at the physical layer, so this is switched branching rather than a shared medium: each segment is its own link, and the switch passes traffic between them.

The practical consequence of branching is less cable, fewer long runs, and a cabinet that does not need a port for every sensor in the building.

Power over SPE: PoDL and hybrid cabling

"SPE carries power" is true but loosely stated, and the distinction matters when you are specifying a system.

PoDL (Power over Data Line)

PoDL superimposes power onto the same twisted pair that carries the data, defined in IEEE 802.3bu. One pair, both jobs.

Hybrid cabling

Hybrid cabling is a cable and connector architecture rather than an SPE standard. Power travels on additional conductors alongside the data pair. The cable is slightly larger, and in exchange it supports devices whose power requirements a single pair cannot serve, including devices of different power classes sharing the same cable network.

Perinet's components use hybrid cabling, supplying 2A at 24V.

The trade-off only becomes obvious once you build something. PoDL requires the power class of the device and of the supplying equipment to be matched, so a network of devices with different loads can complicate switch selection and network design. It also costs space: the circuitry needed to put power on the data pair makes the device bigger, in some cases bigger than the same device would be with ordinary Ethernet. In compact devices, that extra circuitry can reduce or even erase the size advantage that SPE was intended to provide.

Hybrid cabling avoids both problems. Data stays on the SPE pair, power runs on separate conductors, and devices with different load requirements share one architecture within the system's overall power budget. The cost is a slightly larger cable.

SPE and Ethernet-APL: related, not the same

These two terms get used as synonyms and they are not.

SPE is the broad family of single-pair Ethernet physical layers described above.

10BASE-T1L is one member of that family: 10 Mbit/s, long reach.

Ethernet-APL (Advanced Physical Layer) is built on 10BASE-T1L and adds the requirements and profiles needed for process-industry environments, including intrinsic safety for hazardous areas such as chemical and oil and gas plants, where long cable runs and explosion protection dominate the specification.

So Ethernet-APL is not a separate alternative to SPE. It is a process-industry profile built on one SPE physical layer, and not a superset of the family.

Perinet works in a different part of the landscape, on 100BASE-T1, aimed at digitalisation broadly and at industrial and building applications in particular, rather than at process control in hazardous areas.

Perinet does not offer 10BASE-T1L or Ethernet-APL products, and that is a scoping decision. Sensor manufacturers rarely ship a sensor with more than 25 m of cable, and in a real installation there is always a concentration point, a switch or a cabinet, within tens of metres of the device. 100 Mbit/s at that distance is what carries encrypted MQTT, firmware updates and a full IP stack at the sensor. Where a run genuinely has to cover hundreds of metres, fibre between concentration points avoids the galvanic problems of long copper and offers far more bandwidth, and in hazardous-area process loops 10BASE-T1L and Ethernet-APL are the right fit. Because every SPE variant carries ordinary Ethernet frames, a 10BASE-T1L segment in the process area and a 100BASE-T1 segment on the packaging line end up as IP traffic on the same plant network.

SPE alongside your existing automation

The common framing is SPE versus fieldbus. It is the wrong frame, because it implies a choice you do not have to make.

Your PLC exists to control the machine, and its inputs and outputs are wired for that job. It does it well, it is validated, and interfering with it to extract a temperature reading for a dashboard is a poor trade. In most plants the reason field data never reaches IT is not that the control system is bad. It is that a controller built to run tight processes is far more engineering-heavy than monitoring needs to be, so every request to surface one more value turns into engineering work, revalidation, and risk on the production side.

So do not replace that path. Add a second one.

SPE makes it practical to run a parallel route for data: sensors connected directly to the IP network, feeding IT and analytics systems, while the existing control path keeps doing control. Two paths, two purposes, neither compromising the other.

One path for control. One path for data.

For machine control itself, a PLC with directly wired sensors is already the right design, and SPE adds little there. Its value is on the operational side: getting data from machines you already run into the systems where decisions are made, without touching the control program.

That reframing also explains where the savings come from. It is not that Ethernet is cheaper than a fieldbus per metre. It is that a direct path removes the translation layer, and translation layers are where the hidden costs accumulate: the gateway to specify and maintain, the mapping to document, the engineer who understands it, the day lost when a tag name changes.

We have measured that difference. PLC versus periSNOOP for Sensor Data Monitoring compares extracting the same sensor data through a PLC with EtherCAT against a periSNOOP clipped onto the existing loop.

SPE is not automatically the right answer for every device network. Where the job is parameterisation and diagnostics of devices through a master, IO-Link remains a strong fit, and the useful question is which layer you want to remove rather than which vendor is right. Where conventional signals are already concentrated in a cabinet, the honest comparison is SPE against remote I/O rather than against a fieldbus.

The endpoint changes. Each sensor gets its own IP-addressable network endpoint with its own security identity, reached over SPE, with no programming or address mapping to get it there. No master, no gateway in between. The trade-off is deterministic real-time control, which stays where it belongs, on the PLC.

Where SPE is used

• Brownfield retrofit. Get data from machines already in production onto the IP network, without redesigning the machine around them. periSNOOP joins an installed 4-20 mA loop in-line while the PLC keeps its measurement. periNODE adapters read existing digital signals and Modbus buses, or add off-the-shelf sensors where nothing is measured yet. See how to get sensor data without changing the PLC.
• New measurement points. Add an off-the-shelf sensor where nothing was measured before. No I/O card, no PLC change, no automation project to get one number out of a machine.
• Condition monitoring and predictive maintenance. Continuous field data in a form analytics tools can use, rather than a manual reading logged once a shift.
• Machine and line performance. Live availability, performance and quality data from equipment already on the floor. See OEE monitoring for packaging lines.
• Smart buildings. Distributed, low-bandwidth sensing across a large site, where cable volume is the dominant cost and topology is the lever that reduces it.
• Energy and utility monitoring. Put pressure and flow from existing utility lines onto the network to find waste. See compressed air monitoring.

See these applications in practice across our SPE use cases.

How to put a sensor on an SPE network

Most deployments add an off-the-shelf sensor for a new measurement; where a sensor already feeds a PLC, an in-line monitor copies its signal instead. Either way, a Perinet deployment can follow this path:

1. Connect the signal. A periNODE adapter takes the analogue, digital or Modbus output of an off-the-shelf sensor added where the measurement is needed and puts it on the network as structured data, with no existing system required. It can also read an existing digital signal or listen to an existing Modbus bus. For a 4-20 mA sensor already wired to a PLC, periSNOOP sits in the loop instead.
2. Branch where it makes sense. periSWITCH fans a line out close to the devices, so cabling follows the machine rather than the cabinet.
3. Bridge to standard Ethernet. periSTART converts between the SPE network and conventional Ethernet infrastructure.
4. Process on site. periMICA, a modular edge computer, runs monitoring, dashboards and analytics locally, and passes on what belongs upstream.

In a retrofit of this kind, the control program and the PLC stay exactly as they are. If you are designing a new device rather than connecting one, the route runs through periCORE instead, described under SPE components from Perinet.

SPE components from Perinet

For connecting what is already on the floor, the Smart Components do the work. periNODE Single Pair Ethernet adapters bring analogue, digital and Modbus sensors onto the network as structured data. periSWITCH branches SPE lines close to the devices. periSTART converts between the SPE network and standard Ethernet. periSNOOP sits inline in a 4-20 mA loop and reads it without affecting the signal the controller receives.

Above them, the periMICA modular edge computer turns connected sensor data into monitoring, analytics and AI-ready information on site.

To try it, start with a kit.

Building connectivity into your own device?
periCORE gives device manufacturers the same network identity, per-device security, MQTT and HTTPS REST functionality in an embedded module.

Open standards, no lock-in

Everything in a Perinet network is a published standard. The physical layer is 100BASE-T1 (IEEE 802.3bw). The connector is the M8 hybrid mating face defined in IEC 63171-6:2021. periSTART connects the SPE segment to any standard 100BASE-TX Ethernet network. Devices use IPv6 and TLS, and data leaves each device as MQTT and HTTPS REST, so any broker, database or platform that speaks those protocols can read it without Perinet software in between. periMICA runs open-source software in standard Linux containers.

You can start with one periNODE and a periSTART, and nothing you add later depends on replacing what is already there.

When does this matter?

If you are trying to get machine or sensor data into MQTT, SQL, an MES, an ERP or an analytics tool, and doing so today means PLC engineering, a protocol gateway or a proprietary master, the problem may not be the application. It may be where your IP network starts. With an IP-native SPE device, the network can start at the sensor, next to the control system you already run, so the data has its own path into IT and the control program stays untouched.

The quickest way to check is to look at one machine: which signal do you need, where does it have to end up, and how many boxes sit between the two today.

See an example architecture

Or tell us about one machine and we will sketch a practical path.

Tell us which signal you have and where the data needs to end up. We will show you a practical architecture for getting it there without unnecessarily changing the control system. No sales call unless you want one.