10BASE-T1S and PLCA: Multidrop Ethernet Explained

If a zone controller in your architecture needs eight switch ports for eight small nodes, you have met the problem 10BASE-T1S exists to solve. It is the one member of the T1 single-pair Ethernet family that keeps a shared medium, and PLCA is what makes that shared medium predictable enough to put in a vehicle.

Most writing about 10BASE-T1S is organised around one part number, which helps after you have chosen the part and not before. This is the vendor-neutral version, and every timing figure, node count and segment length below carries a public IEEE or OPEN Alliance source. For the physical layer generally, start with our complete guide.

What 10BASE-T1S is

10 Mbit/s, one balanced pair, half duplex, and the mixing segment

10BASE-T1S is Clause 147 of IEEE Std 802.3cg-2019: 12.5 MBd, 4B/5B encoded differential Manchester, no forward error correction, one unshielded balanced pair, half duplex by default.

The rate is not the interesting part. IEEE 802.3cg defines two kinds of segment for this PHY, and the second is a shared medium. The OPEN Alliance interoperability test suite states both. A point to point link segment supports up to four in-line connectors, for at least 15 m of reach in the IEEE 802.3cg objectives; the OPEN Alliance interoperability suite builds its point to point test channel at 25 m. Multidrop is optional: half duplex on a mixing segment, one balanced pair interconnecting up to at least 8 PHYs to a trunk of up to at least 25 m, PHYs attached in line or on stubs up to 10 cm long. An overall effective rate of 10 Mbit/s is shared among the nodes.

Point to point versus multidrop with the same PHY

The same silicon does both. Point to point it is an ordinary short-reach link; where PLCA is enabled on one, the interoperability suite configures it as a two-slot cycle. Multidrop is where the design decisions start, because the pair is now a bus: node capacitance loads it, stubs reflect on it, and termination becomes a system property rather than a part inside each PHY. The OPEN Alliance register specification names the roles. An end node sits at either end with nothing between it and the edge termination, a drop node is any node between the two end nodes, and a head node is the highest level application node, typically implementing switch or gateway access to the network beyond the segment.

Where it sits in the T1 family

At the bottom, and alone in being designed to be shared. Every other member is point to point and full duplex, with a switch port per node; our T1 family explainer sets the family side by side across five IEEE standards. Settle one distinction first: 10BASE-T1L is Clause 146 of the same amendment, full duplex point to point, aimed at process automation. No mixing segment, no PLCA, and half the confusing search results are about it.

What 802.3cg standardises, and what OPEN Alliance adds

IEEE 802.3cg gives you the PHY, the segment definitions, and PLCA as the optional Clause 148 reconciliation sublayer with management objects in Clause 30. It does not give you a register map or a test plan.

Those come from the OPEN Alliance, whose published 10BASE-T1S set includes PLCA management registers, a PLCA conformance test suite, a half-duplex interoperability test suite, PMA and PCS test material, advanced diagnostic PHY features, EMC specifications, sleep and wake-up, and topology discovery. When someone says a PHY is compliant, ask against which of those.

The register map places PLCA control and status in MMD 31 of the Clause 45 address space. The reconciliation sublayer sits above the MII in the layering model, but many implementations embed PLCA in the PHY so it can be bolted to an MCU or switch that only exposes an MII. That is why you configure PLCA over MDIO, not through your MAC driver.

Why a 10 Mbit/s Ethernet exists at all

Ten megabits is not a bandwidth story, it is a stack story. A 10BASE-T1S node carries the unchanged IEEE 802.3 frame, so it runs IP, runs your existing service layer, and is captured with the same tools as everything else. The alternative at that edge is a different protocol plus a translation function in every gateway that touches it. The natural home is the leaf of a zonal architecture: lighting, seats, simple sensors, small actuators. Our domain versus zonal article covers why aggregating those into a zone controller is the structural argument, and multidrop is what stops that aggregation needing a switch port per leaf.

CAN and CAN FD stay at the leaf for a specific reason. CAN resolves contention before a frame goes out, with bit-wise arbitration that guarantees latency for the highest priority traffic, and it puts fault confinement in the data link layer. PLCA gives fairness and bounded access latency, a different property. Our Ethernet versus CAN comparison works that decision through function by function.

The wiring argument for multidrop is structural: one pair with stubs replaces N pairs into N switch ports, and the ports go away too. The cost argument gets no number here. Per-node figures for real programmes are OEM internal, so any percentage saving quoted against CAN came either from a bill of materials you cannot inspect, or from nowhere.

PLCA, arbitration without collisions

The coordinator, node ID 0, and the beacon

One node is the coordinator: the node configured with PLCA local node ID 0, responsible for the periodic beacon and for configuring the number of transmit opportunities between each beacon. Every other node is a follower with an ID from 1 to 254, synchronising its transmit opportunity counter to that beacon.

Node ID 255 is not a node. It is the register default, and the specification states that this special value causes PLCA functions to be suspended; the ethtool manual page describes it as disabling the node. A PHY powered up and left alone is both PLCA-disabled and node-suspended. Nothing about PLCA happens by accident.

Transmit opportunities and how the cycle advances

After the beacon, opportunities are counted off in node ID order, and a node transmits only during the one matching its own ID. A node with nothing queued releases its opportunity quickly, which stops the cycle degenerating into fixed slots.

The knob is the transmit opportunity timer, to_timer, in bit times. The register specification gives it a default of 32, requires it be set equally across all nodes for collision-free operation, and states that the default guarantees proper operation of conformant PLCA-enabled nodes over a 25 m twisted-pair cable in line topology. The ethtool manual page documents the same default as enough to cover a link of roughly 50 metres.

That number sets the cost of an idle node. In the interoperability suite’s own configuration, a to_timer of 32 bit times with a maximum slot number of 15 gives 16 opportunities per cycle, so a fully idle cycle costs 16 x 32 bit times plus the 20-bit beacon, 532 bits, about 53.2 microseconds at 10 Mbit/s. Loaded, the same document computes about 1.44 ms per cycle with 64-byte frames and about 24.7 ms with 1518-byte frames. Worked examples, not limits, but the lesson holds: cycle time is dominated by what is sent, not by how many nodes exist.

Node count on the coordinator

This is the setting people get wrong. The register specification is explicit that on the node with PLCA ID 0, node count must be set at least to the number of nodes that may be plugged into the network for PLCA to operate properly. A silicon vendor’s application note puts it in the form engineers use: transmit opportunities equal the maximum node ID in the segment plus one.

The register default is 8, and the ethtool manual page adds the consequence: node count regulates the minimum length of the cycle. Too high and every cycle carries idle opportunities nobody uses. Too low and the top node never gets a turn.

Burst settings

By default a node sends one frame per transmit opportunity. Maximum burst count sets the additional packets it may send during its own opportunity, defaulting to 0, which means burst mode off. When non-zero, the burst timer sets how long the reconciliation sublayer waits for the MAC to deliver a new packet before yielding, defaulting to 128 bit times and required to exceed the local MAC interpacket gap plus any additional transmit latency.

Burst is a fairness trade, paid for out of every other node’s worst case. Decide it per node and record it beside the node IDs.

Neither TDMA nor CSMA/CD

Not TDMA: an idle opportunity collapses to the to_timer value while a busy one lasts as long as the frame. Not CSMA/CD replaced either: Clause 148 defines PLCA as working in conjunction with CSMA/CD. That is what makes the failure mode below possible.

Topology and wiring rules

Mixing segment, stubs, termination

Two terminations, one at each physical end of the trunk, nominally 100 ohm. The register specification defines the end node as the node with nothing between it and the 100 ohm edge termination, which it may itself contain.

The interoperability suite’s reference network gives concrete geometry rather than round numbers: 8 nodes on a trunk whose total wire length between the two differential end terminations is 25 m, each node on a 10 cm stub, on automotive grade unshielded twisted pair. The splice distances between adjacent nodes are not uniform. They are 4 m, 2 m, 4 m, 0.5 m, 0.5 m, 13.5 m and 0.5 m, with the end terminations 0.1 m out from the end nodes. A real harness is lumpy, so the reference channel is lumpy.

Two details for a design review. Common mode termination is optional and only for drop nodes, and in the suite’s own harness the differential end termination is a stand-alone unit rather than resistors populated inside the end nodes, even though a product end node is allowed to contain it. The suite also tests termination mismatch deliberately: one end termination missing, both missing, a spurious third present. If your bring-up has no step verifying that both terminations exist and only two do, add one.

Read in physical order along the trunk, that topology’s node IDs are 4, 7, 6, 11, 8, 0, 15 and 10: node ID is a schedule position, not a place on the wire.

Connector and cable practice at the edge

Stub length is specified in centimetres, which makes the connector, the pigtail and the board entry a meaningful fraction of it. Our cables and connectors article covers the families in use and what their specifications do and do not tell you. Sleep and wake-up on this PHY has its own timers, covered in our TC10 article.

Bring-up checklist

1. Assign node IDs contiguously from 0. Contiguity is not required, but gaps become idle opportunities costing to_timer each, every cycle. Record the assignment under version control with the rest of the network configuration, as our configuration as code article argues at the whole-network level.
2. Set node count on the coordinator to the highest node ID in service plus one. Not the number of nodes present. With contiguous IDs those are the same number, which is the real reason to keep them contiguous.
3. Set to_timer identically on every node. The register specification requires equality for collision-free operation, and this is the commonest asymmetry in a mixed-vendor segment, because defaults are per-part.
4. Decide burst settings deliberately, then write them down. Off is the default and a legitimate answer. If you turn it on, keep the burst timer above your MAC’s interpacket gap plus transmit latency.
5. Decide who is the coordinator, once. The ethtool manual page states there must be exactly one coordinator on a PLCA network. Coordinator redundancy is a question for your silicon vendor, not a behaviour to assume from the base specification. A dynamic PLCA extension exists, allowing IDs including node 0 to be assigned during operation, but the interoperability suite places dPLCA outside its scope.
6. Read back the PLCA status bit on every node. It reads 1 when the coordinator is transmitting the beacon regularly, or a follower is receiving it regularly, and 0 otherwise. On a Linux test host, ethtool --get-plca-status reports the same thing, documenting off as PLCA temporarily disabled with the PHY in plain CSMA/CD mode.
7. Confirm with a capture, not a ping. A ping succeeds under CSMA/CD too. Our capture encapsulation article covers getting bus traffic into one time-aligned path.

Troubleshooting PLCA

The organising fact: PLCA can stop without the link going down. Frames keep flowing, every interface reports up, and the segment is now a contention network, which is why the symptoms below look like application problems.

Throughput collapses under load but everything looks up. The beacon stopped and the followers reverted. The interoperability suite spells out the sequence: the PLCA status bit stays set while beacons arrive regularly; when they stop, the node keeps incrementing its transmit opportunity counter until it reaches the maximum count of 255, then stops and starts a timer; if no beacon arrives before that timer expires, the status bit is cleared and the node reverts to CSMA/CD with PLCA deactivated. A returning beacon sets the bit again and normal PLCA resumes. A silicon vendor’s application note describes the same fallback and recommends an interrupt on the status-change bit rather than polling. The fallback exists to keep communication running, which is why nothing alarms.

PLCA status reads 0 on one node while the rest of the segment is fine. That node is not participating. Check its ID is not 255, that PLCA enable is set, and that its to_timer matches the others. The enable bit defaults to 0, meaning plain CSMA/CD, so a node that missed its configuration sequence lands here.

Duplicate node IDs. Two nodes claim the same opportunity and transmit into each other. The receive in assigned transmit opportunity bit is set when the PHY detects the beginning of a packet during its own assigned opportunity, which the OPEN Alliance diagnostic specification says indicates another node with a duplicate ID. It is mandatory for both roles and cleared only by an intentional host write, so it survives until you read it.

Node count set below the highest node ID in use. The top node never gets a turn, and the followers can tell you. The beacon received before transmit opportunity bit is set on a follower when the cycle completes before its assigned opportunity occurred, which the same specification says indicates either multiple coordinators or a node count smaller than the number of nodes present.

No coordinator at all. No beacon, so every node runs the sequence above and the whole segment ends up in contention. Distinguishable from a stopped beacon only by whether one was ever transmitted.

Two coordinators. The unexpected beacon bit, mandatory for the coordinator role, is set on a coordinator that detects a beacon it did not transmit, indicating another coordinator on the segment. Treat it as a configuration fault to remove, not a redundancy mechanism to tune.

Mixed PLCA and non-PLCA nodes. A legitimate configuration the suite tests, and one that needs care: a node running plain CSMA/CD transmits whenever the medium looks idle, so a run of empty opportunities can launch a packet into a PLCA node’s slot. With a to_timer of 32 bit times, three consecutive empty opportunities amount to 9.6 microseconds, equal to the 96 bit-time interpacket gap at the MAC, which is why the suite advises avoiding gaps of one, two or three.

Working method. Read the diagnostic registers first, on every node, and clear them before the run so what you read afterwards belongs to it. Capture second, physical layer included, because a stopped beacon is visible on the pair and an ID collision has a signature. Replace hardware last. Our 100BASE-T1 link bring-up checklist applies the same discipline, with one difference: on a mixing segment the fault can live in a node that is not the one showing symptoms.

Interoperability and conformance

TC14 is the OPEN Alliance technical committee that owns 10BASE-T1S interoperability and compliance testing, and its documents are listed on the OPEN Alliance specification index.

The Half-Duplex Interoperability Test Suite defines system tests. Its stress conditions include illegal transmissions, collisions, babbling idiots, ground shifts, clock drifts and harness failures, and its stated major objective is that disturbances do not lead to a hang-up or dead-lock state. Its groups cover normal communication and startup under those stresses, the diagnostic feature set, wake-up and sleep, and topology discovery.

What it does not do is stand in for PHY-level compliance. Its own note says results depend not only on the PHY or switch but on their configuration, the external filter, the MDI circuitry and the channel; PMA and PCS compliance live in separate documents. Our conformance article sets out how these layers relate.

For a lab, the consequence is that exercising a mixing segment is not exercising a link. You need a harness with the right geometry and fault injection into it, several link partners rather than one, concurrent register access across every node, and physical-layer observation on the shared pair. For where these segments sit relative to the backbone, see our in-vehicle Ethernet architecture article.

Where GSAS fits

GSAS Micro Systems is an engineering partner to teams building and testing vehicle networks in India. On 10BASE-T1S the conversation we have most often is not about which PHY to buy. It is about whether multidrop belongs in that part of the topology at all, and what a bench needs before a mixing segment can be signed off rather than merely observed working.

The value we try to add is making the silent things visible. A segment that has fallen back to contention still passes a ping, still shows every link up, and still moves traffic, but it no longer has the property you designed the architecture around. Instrumenting for that means reading PLCA status and the diagnostic bits on every node, capturing the shared pair rather than a switch port, and holding node IDs, node count, to_timer and burst settings as recorded configuration rather than remembered bring-up steps. Our engineers work with teams in Bengaluru, Pune, Chennai and Hyderabad on that gap between a segment that works and a segment you can prove.

If you are scoping a multidrop segment, sizing a bench that can stress one, or working out whether a CSMA/CD fallback explains a latency figure nobody can reproduce, our automotive Ethernet capabilities page covers what we do, and you can request a consultation.

References

• IEEE P802.3cg 10 Mb/s Single Pair Ethernet Task Force public area: https://www.ieee802.org/3/cg/
• IEEE P802.3cg objectives: https://www.ieee802.org/3/cg/objectives_3cg_0318.pdf
• IEEE P802.3cg tutorial, “10Mb/s Single Pair Ethernet: A guide” (Clause 147 at 12.5 MBd 4B/5B DME with no FEC, 15 m point to point and 25 m multidrop; Clause 146 10BASE-T1L at 7.5 MBd PAM-3 4B3T with 1000 m reach on 18 AWG): https://www.ieee802.org/3/cg/public/Jan2019/Tutorial_cg_0119_final.pdf
• OPEN Alliance, 10BASE-T1S Half-Duplex Interoperability Test Suite Specification, version 1.5 (segment definitions, node and stub figures, reference topology, cycle-time examples, beacon-loss fallback, termination mismatch, dPLCA scope): https://opensig.org/wp-content/uploads/2024/03/OA_10BASE-T1S_Interoperability_Test_Suite_v1.5.pdf
• OPEN Alliance, 10BASE-T1S PLCA Management Registers Specification, version 1.4 (MMD 31 register map, node roles): https://opensig.org/wp-content/uploads/2025/03/2024-12-10BASE-T1S-PLCA-Management-Registers-v1.4.pdf
• OPEN Alliance, Advanced diagnostic features for 10BASE-T1S automotive Ethernet PHYs, version 1.1 (diagnostic bits by role): https://opensig.org/wp-content/uploads/2024/01/OPEN_Alliance_10BASE-T1S_Advanced_PHY_features_for-automotive_Ethernet_V1.1.pdf
• OPEN Alliance automotive Ethernet specification index: https://opensig.org/automotive-ethernet-specifications/
• Silicon vendor configuration application note AN60001699 (transmit opportunities as maximum node ID plus one; CSMA/CD fallback on beacon loss): https://ww1.microchip.com/downloads/aemDocuments/documents/AIS/ApplicationNotes/ApplicationNotes/LAN8670-1-2-Configuration-Appnote-60001699.pdf
• ethtool(8) manual page (PLCA configuration and status options): https://man7.org/linux/man-pages/man8/ethtool.8.html
• ASAM CMP, Capture Module Protocol (10BASE-T1S symbols among supported payload types): https://www.asam.net/standards/detail/cmp/