Engineering Guide · CAN / CAN FD

Industrial CAN Bus Engineering: Isolation, CAN FD, Signal Integrity & Reliability

Updated October 17, 2026 · BITECH Engineering · Technical figures are taken from the cited standards and device datasheets; BITECH model data comes from the published product pages.

Industrial CAN Bus Architecture & Reliability Requirements

Start CAN hardware selection from the system boundary — which devices share power and ground, how far the cable runs and which frame formats must coexist — rather than from the protocol's maximum bit rate.

Why CAN Bus Still Matters in Industrial Automation

CAN remains common in machine controls, servo and motion networks, mobile robots, vehicle diagnostics and battery management systems. Its value comes from protocol mechanisms: non-destructive bitwise arbitration, CRC-based error detection, automatic retransmission and fault confinement that removes a persistently faulty node from the bus. Higher-layer protocols such as CANopen and SAE J1939 build on these mechanisms.

Arbitration guarantees that the highest-priority identifier wins access when the bus becomes idle. It does not by itself guarantee a fixed latency for every message. Worst-case response time depends on bus load, frame lengths, identifier assignment and error recovery, so it has to be calculated or measured for each network rather than quoted as a generic figure.

Common CAN Bus Failure Modes

Every CAN controller keeps a transmit error counter (TEC) and a receive error counter (REC). A node is error-active while both counters are below 128, becomes error-passive when either reaches 128 and goes bus-off when the TEC exceeds 255. A bus-off node stops transmitting until it is recovered. Most field faults show up as counter growth long before communication stops.

Key Engineering Parameters Before Hardware Selection

Galvanic Isolation: When Isolated CAN Is Necessary

Isolate a CAN port when the nodes it connects do not share a solid, low-impedance ground reference — separate cabinets or supplies, long cable runs, drive-heavy floors or vehicle chassis segments. A short segment inside one cabinet with a verified common ground often runs reliably without isolation.

Ground Loops and Ground Potential Differences

A non-isolated transceiver references CAN_H and CAN_L to the local logic ground. When two nodes sit on grounds that differ by several volts — because of motor return currents, long earth paths or separate power supplies — that offset appears as common-mode voltage at the receiver. Standard transceivers tolerate only a limited common-mode range; the transceiver discussed in an EEVblog design review later in this guide is checked against a ±12 V receiver range, while an isolated part such as TI ISO1042 specifies ±30 V. Beyond the receiver's range, bits are misread; beyond the absolute maximum ratings, parts are damaged. Ground-return current flowing through the CAN reference conductor or shield also injects noise.

Isolated vs Non-Isolated CAN Architecture

Mixing isolated and non-isolated nodes on one bus is electrically valid. Isolation protects the node that has it; it does not make the other nodes immune, and it is not required on every node by default.

Isolated DC/DC Power and Signal Isolation

A digital isolator or isolated transceiver only creates an isolation domain if the bus-side transceiver is powered from a supply that is also isolated from logic ground. TI's ISO1042 documentation, for example, describes its protection as effective when the device is used together with isolated power supplies. If the bus-side supply is derived from the logic rail, or if CAN_GND, the shield or a protection component is tied back to logic ground, the barrier is bypassed. Review the isolated DC/DC converter's own isolation rating, any Y-capacitors across the barrier, and where CAN_GND and the shield terminate.

Understanding Isolation Ratings, CMTI and Bus Protection

A single "isolation voltage" figure is not enough to compare interfaces. Device datasheets separate several ratings, measured under different conditions:

Product-level figures such as "4000 V isolated" are normally withstand-type ratings. When comparing industrial PCs, ask which rating the figure refers to, the test standard and duration, the working voltage, and whether ESD/EFT/surge immunity was tested on the CAN port. BITECH's AE-3588BT page, for instance, states that its 4000 V figure is a withstand rating and that working-voltage data comes from the isolation test report. For the underlying principles, see Industrial PC Galvanic Isolation and Isolated vs Non-Isolated I/O.

Classical CAN vs CAN FD: Compatibility & Failure Modes

Classical CAN and CAN FD nodes can share a bus only if every Classical node is FD-tolerant, or if the CAN FD nodes transmit Classical frames only. A standard Classical CAN controller answers every CAN FD frame with an error frame.

Classical CAN vs CAN FD — Frame Structure and Data Rates

The controller, the transceiver and the harness each impose a limit; the lowest one sets the usable data rate.

Why Legacy CAN Controllers May Fail on CAN FD Networks

A Classical controller expects a reserved bit where a CAN FD frame carries a recessive FDF bit. It treats that bit as a format violation and transmits an error frame, destroying the CAN FD frame on the bus. ISO 11898-1:2024 distinguishes nodes that support Classical frames only without tolerating CAN FD, nodes that tolerate CAN FD frames, and nodes that support CAN FD or CAN XL. A Classical-only node can communicate with CAN FD-capable nodes only while Classical frames are used.

Field example. In a March 2026 NXP Community thread, an engineer connected an S32K144 CAN FD channel and two Classical-only channels (one behind a TJA1051T transceiver) to the same bus. The first CAN FD transmissions produced 15 error frames; after that, CAN FD traffic passed and no further error frames appeared. NXP support explained that the Classical node sends active error flags until its error counters reach 128, after which it is error-passive or bus-off and stops disturbing the bus. The network appears to work only because the legacy node has silenced itself — it is no longer receiving reliably. NXP's recommended fix was a transceiver with CAN FD-passive support, such as specific variants of the UJA1169.

CAN FD Error Frames, Error Passive and Bus-Off

Diagnose mixed-network problems with error counters, not with "frames still arrive". Typical signs are a burst of error frames whenever CAN FD traffic starts, one node whose TEC/REC climbs quickly to the error-passive threshold, and repeated bus-off and restart events on that node. An error-passive node can still transmit but signals errors only passively, so it stops destroying frames while its own reception remains unreliable.

Mixed Network Design and Gateway Solutions

Diagnostic sequence: list every node's controller capability → capture error frames and counters while CAN FD traffic starts → identify the node whose counters rise → confirm whether its controller and transceiver are FD-tolerant → choose one of the options above.

CAN SIC Transceivers & Signal Integrity

CAN SIC (Signal Improvement Capability) is a transceiver feature that actively suppresses ringing at the dominant-to-recessive edge and tightens bit-timing symmetry. It does not change the CAN FD protocol, the frame format or the controller.

What Is CAN SIC (Signal Improvement Capability)?

CAN SIC was first specified by CAN in Automation in CiA 601-4 and is now part of ISO 11898-2:2024, after which CiA withdrew CiA 601-4. CiA has also published a corrigendum proposal, CiA 140, to address inconsistencies in the published ISO 11898-2:2024 text. SIC transceivers are interoperable with conventional high-speed Classical CAN and CAN FD transceivers on the same bus, according to NXP's TJA1463 documentation.

Ringing, Reflections and High-Speed CAN FD

Every stub, star point and unterminated branch is an impedance discontinuity that reflects energy. Reflections are most harmful at the dominant-to-recessive transition, because the recessive state is not actively driven — only the termination pulls the bus back. At 5 Mbit/s a data-phase bit lasts 200 ns, so ringing that would be harmless in a 2 µs bit at 500 kbit/s can consume the sample-point margin.

Conventional CAN FD PHY vs CAN SIC PHY

The 8 Mbit/s figure is a transceiver capability, not a guarantee for any harness. Usable data-phase rate still depends on the controller's timing, loop delay, node count, cable and stub lengths, which is why the full bus must be validated.

When to Consider CAN SIC for Industrial Networks

• Star or multi-stub topologies that cannot be rewired into a clean linear bus.
• Networks with many nodes that need a CAN FD data phase of 2 Mbit/s or more.
• Designs that may migrate toward CAN XL, where CiA lists CAN SIC XL transceivers with a FAST mode of up to 20 Mbit/s.

BITECH product pages do not list CAN SIC transceivers on current AX or AE models. Treat CAN SIC as an OEM/ODM evaluation item, not a standard feature.

CAN Bus Termination, Topology & Grounding

A linear high-speed CAN bus is terminated with one 120 Ω resistor at each physical end of the trunk — two in total, not one per node.

Why Two 120 Ω Terminators Are Required

The terminators match the characteristic impedance of the twisted pair, absorbing the wave at each end instead of reflecting it, and they discharge the bus into the recessive state after each dominant bit. Nodes in the middle of the trunk connect through short stubs and carry no termination.

Understanding the 60 Ω Resistance Measurement

With every node unpowered and no other parallel path, two 120 Ω terminators in parallel measure about 60 Ω between CAN_H and CAN_L. Each connected transceiver adds a high input resistance in parallel, so large networks read slightly lower.

For the full field procedure, see CAN Bus 60 Ω Check.

Split Termination and Switchable Termination

Split termination replaces each 120 Ω resistor with two 60 Ω resistors whose centre point is connected to ground through a capacitor, filtering common-mode noise. Switchable termination lets one hardware design serve as an end node or a mid-bus node, but the switch becomes part of the termination network.

A December 2025 EEVblog design review of a TCAN1044A interface with MOSFET-switched split termination shows the typical traps. A reviewer noticed that placing the termination behind 10 Ω series resistors made it look like about 140 Ω to the bus instead of 120 Ω. Another pointed out that the MOSFET gates were referenced to board ground while the sources followed the bus, so common-mode excursions within the transceiver's ±12 V range could switch the termination unintentionally, and stressed that the network's behaviour when unpowered matters. The designer moved to an opto-isolated solid-state-relay approach based on a TI reference design. Before adopting switchable termination, verify the effective resistance including switch on-resistance, the common-mode behaviour and the state during power loss.

Stub Length, Cable Shielding and Grounding Strategy

Allowed stub length shrinks as the data-phase bit rate rises, because reflections must settle well within a shorter bit time. Use the limits in ISO 11898-2 guidance and the transceiver vendor's application notes for the actual bit rates, rather than one fixed length for all networks. Carry a CAN reference conductor (CAN_GND) with the pair when nodes are isolated or on different supplies. For shields, single-point bonding avoids low-frequency ground-loop currents, while bonding at both ends performs better at high frequencies when an equipotential bonding system exists. Decide per installation and document the choice.

Dual CAN Architecture for Industrial PCs, AMRs & BMS

Two CAN ports are only as independent as their controllers, transceivers and isolation domains. Check all three before relying on dual CAN to separate traffic.

Independent CAN Controllers and Isolation Domains

Dual CAN for AMR/AGV Motion and Diagnostics

A common AMR pattern places drive controllers on one channel and BMS, sensors, diagnostics and firmware download traffic on the other. Diagnostic bursts and large transfers then cannot consume motion-bus bandwidth, and a fault on the service bus does not take the drive bus off-line. See AMR Controller Solutions for platform options.

CAN Networks for Battery Energy Storage and BMS

Battery racks and energy-storage segments may sit at different potentials and do not always share a ground plane. Isolation per channel lets each CAN segment keep its own reference. BITECH's AE-3588BT, for example, lists two CAN 2.0B channels with 4000 V isolation, positioned for BMS aggregation in ESS gateways.

Vehicle CAN Gateways and Power Management

In vehicles, supply transients and CAN line protection are separate design questions. Ignition sensing, controlled shutdown, cranking dips and load dump belong to the power input; ISO 7637-2 tests address that path. The input range also has to match the vehicle bus: BITECH's AX-238BT page notes that a 9–36 V input cannot ride through the charging transients of a 48 V system without an external converter or a different input stage. Details are covered in Powering an Industrial PC in a Vehicle.

Functional Safety Boundaries

Dual CAN separates traffic; it does not create a safety function. Emergency stop and other safety functions need a safety chain designed and assessed to the applicable standard — for example IEC 61508 or ISO 13849 — with safety-rated components or protocols such as CANopen Safety (EN 50325-5). A general-purpose industrial PC with two CAN ports is not SIL- or PL-rated unless explicitly certified.

EMC Protection: ESD, EFT, Surge & Industrial Transients

Galvanic isolation and transient protection solve different problems. The barrier blocks ground-potential differences; TVS diodes, common-mode chokes, layout and shielding decide whether the port survives ESD, EFT and surge.

TVS Diodes and Common-Mode Chokes

Place TVS protection at the connector, with a short return path to the correct reference — the bus-side ground on an isolated port. Choose a working voltage above normal CAN levels and the expected fault voltage, and a clamping voltage below the transceiver's absolute maximum. TVS capacitance loads the bus and can unbalance CAN_H and CAN_L, so use low-capacitance, matched parts for CAN FD. Common-mode chokes reduce emissions and improve immunity, but some parts introduce ringing or asymmetry at high data rates; select chokes specified for CAN FD.

Isolation vs Surge Protection — Different Functions

An isolation rating describes the barrier. It is not the CAN port's surge immunity level, and neither is equivalent to a product certification. A product's EMC declaration applies to the tested configuration, ports and cabling, so a configuration change — an added isolated option, different cable, different grounding — may need re-verification.

IEC 61000-4-2 / -4-4 / -4-5 Testing

Performance criteria are usually reported as A (normal operation within specification), B (temporary degradation with self-recovery) or C (loss of function requiring intervention). For a CAN port, request the criterion observed on the bus — for example, whether error frames or a bus-off occurred during the test.

Automotive ISO 7637-2 Considerations

ISO 7637-2 covers electrical transients conducted along vehicle supply lines, so it applies to the industrial PC's power input, not to the CAN pair. Transients coupled onto other lines are addressed by ISO 7637-3, and CAN transceiver EMC is evaluated at component level under IEC 62228-3. Keep the supply-line and data-line evidence separate in test plans and reports.

Linux SocketCAN, CAN FD Drivers & Troubleshooting

A CAN port is ready for production on Linux when it appears as a native SocketCAN interface, accepts the required bit rate and data bit rate, reports error states correctly and recovers in a controlled way. Verify each point on the actual BSP image.

Native SocketCAN vs USB-CAN Vendor Drivers

SocketCAN exposes CAN controllers as network interfaces (can0, can1), configured with ip and tested with can-utils. Applications use standard sockets, so the same code runs on any SocketCAN driver. Adapters that depend on a vendor library work only where that library is supported and need separate validation. Several USB adapters have mainline SocketCAN drivers; the controller and driver behind each port should still be identified.

Checking CAN FD Support in Linux

CAN FD needs support at every layer: controller silicon, kernel driver, device tree or bus bridge configuration, and transceiver. Configuring fd on fails on a controller without CAN FD support, and the kernel requires the data bit rate to be at least the arbitration bit rate. After configuration, ip -details link show can0 shows the FD flag and both bit-timing sets.

# Show the current state, bit timing and error counters
ip -details -statistics link show can0

# Configure CAN FD (only if controller and driver support it)
sudo ip link set can0 down
sudo ip link set can0 type can bitrate 500000 dbitrate 2000000 fd on restart-ms 100
sudo ip link set can0 up

# Monitor traffic, then re-check state and counters
candump can0
ip -details -statistics link show can0

restart-ms 100 is a demonstration value. Choose automatic recovery according to the system's fault-handling strategy — see below.

Using candump, cansend, canbusload and canfdtest

Diagnosing Bus-Off, Error Counters and Recovery

The -details -statistics output reports the controller state (ERROR-ACTIVE, ERROR-WARNING, ERROR-PASSIVE, BUS-OFF), the TX/RX error counters (berr-counter) and event counts for restarts, bus errors, lost arbitration, warnings, passive transitions and bus-off. With restart-ms 0 there is no automatic recovery, and a bus-off interface is restarted manually with ip link set can0 type can restart. Automatic restart restores communication after a transient fault, but on a permanently shorted or mis-configured bus it repeats error bursts. Applications that depend on the bus should supervise message timeouts rather than rely only on driver recovery.

What to Verify in an Industrial PC BSP

CAN XL, CANsec & 10BASE-T1S: Future Industrial Networks

For most industrial projects today, CAN FD covers the bandwidth and ecosystem requirements. Evaluate CAN XL, CANsec or 10BASE-T1S when a requirement CAN FD cannot meet — larger payloads, data-link-layer security or Ethernet-native integration — and when silicon, software stacks and tools exist for the target platform.

CAN XL vs CAN FD

CAN XL is standardized in ISO 11898-1:2024 and extends the data field to 1–2048 bytes, adjustable in single-byte steps. Data-phase rates of 10 Mbit/s and above require CAN SIC XL transceivers. Nodes that implement CAN XL can transmit and receive all frame formats defined in ISO 11898-1:2024.

CANsec and Network Security

CANsec is a data-link-layer security protocol for CAN XL developed by CiA in CiA 613-2. A bit in the CAN XL control field marks secured frames, and a CANsec header carries cipher control information, a secure channel identifier and a freshness value. Key management is handled separately. Before planning on CANsec, confirm the current status of the specification and the availability of controllers and stacks that implement it.

CAN vs 10BASE-T1S for Robotics and Automation

10BASE-T1S is defined in IEEE 802.3cg-2019. It carries standard Ethernet frames at 10 Mbit/s, half-duplex, over a single twisted pair in a multidrop topology, and uses Physical Layer Collision Avoidance (PLCA) to schedule transmit opportunities. It is a short-reach PHY; onsemi documentation cites 25 m for multidrop segments.

When to Keep CAN and When to Evaluate Alternatives

• Keep CAN FD when the network carries short, prioritized control messages, the device ecosystem (drives, BMS, sensors) is CAN-based, and CANopen or J1939 tooling is already in place.
• Evaluate CAN XL when payloads exceed 64 bytes or when tunnelling higher-layer protocols over CAN is planned, and silicon is available for every node.
• Evaluate 10BASE-T1S when edge devices should join an IP/Ethernet architecture directly and segments are short.

Choosing a BITECH Industrial PC with CAN / CAN FD

Select a platform by verified interface facts — CAN type per channel, whether isolation is standard, optional or ODM and what its rating means, and the OS and driver path — rather than by the number of CAN connectors.

Native CAN vs Optional Isolated CAN

On several BITECH platforms isolation is a configuration choice. Specify it for drive networks, long or cross-domain cabling, vehicle power and high-EMI floors; omit it for clean cabinets with a unified ground to save cost and loop delay. Two CAN ports do not imply two isolated ports.

CAN 2.0B vs CAN FD Hardware Requirements

Confirm CAN FD per channel. A platform with CAN FD listed on one model and CAN 2.0A/B on another needs the exact configuration checked, and CAN FD in hardware does not by itself mean a native SocketCAN CAN FD driver on your image.

Linux, Windows and Driver Qualification

Ask for the driver path for each OS: native SocketCAN on Linux, and the vendor API or driver on Windows. Run the validation checklist above on the delivered BSP version.

OEM/ODM CAN Interface Customization

For volume projects, CAN channel count, isolation, connector type and BSP can be adjusted at board level. Send the requirement list in the RFQ checklist below.

Industrial CAN Bus Engineering FAQ & RFQ Checklist

RFQ Checklist for CAN Interface Projects

• Application and target environment (cabinet, vehicle, outdoor, ESS site)
• Classical CAN, CAN FD, CANopen, J1939 or proprietary protocol
• Number of CAN ports; independent controllers required?
• Isolation requirement and its basis (working voltage, withstand test, separate grounds)
• Bit rate and data bit rate, maximum bus length, topology and node count
• Linux or Windows; kernel, driver and API requirements
• Supply voltage, ignition input, EMC levels and temperature range
• Expected quantity, project phase and validation evidence needed