How to Select a CAN or CAN FD Transceiver for Automotive and Industrial Designs

Why the transceiver choice decides bus reliability #

The CAN controller lives inside the MCU, but it only speaks to the wiring through a transceiver. That small interface chip converts the controller's single-ended TX/RX into the differential CANH/CANL pair, sets the common-mode range the bus can tolerate, and defines how the node behaves during arbitration, error frames and sleep. Pick the wrong one and the rest of a perfectly good design still fails EMC, drops into bus-off under load, or quietly corrupts frames on a noisy harness.

For automotive body and powertrain, industrial PLC backplanes and battery-management systems, the transceiver is therefore a first-class selection, not a copy-paste from the reference schematic. The good news is that the decision collapses into a handful of well-defined dimensions.

High-speed CAN, CAN FD and fault-tolerant: know which bus you are on #

Three physical-layer families cover almost everything you will meet:

• High-speed CAN (ISO 11898-2) runs up to 1 Mbit/s on a 120 Ω terminated bus. This is the default for powertrain, chassis and most industrial control.
• CAN FD keeps the 1 Mbit/s arbitration phase but raises the data phase to several Mbit/s, so a frame carries far more payload in the same bus time. It needs a transceiver that is explicitly FD-tolerant — the faster data phase changes the bit shape and the transmitter must not choke on the stuff-bit counter. A legacy 1 Mbit/s-only part will not reliably support FD nodes.
• Fault-tolerant CAN (ISO 11898-3, low-speed) tops out near 125 kbit/s but tolerates a short to battery or ground on either line, which is why body-comfort and door modules still use it.

If your network mixes FD and classic nodes, the transceiver must support the highest mode any node on that segment uses. A single non-FD part on an FD segment forces the whole segment back to classic timing.

Isolated or not: the decision that changes your BOM #

The second dimension is whether the transceiver needs galvanic isolation from the bus.

A non-isolated part such as the NXP TJA1042 sits on the same ground as the MCU and is the right call for nodes that share a clean local ground — most body modules and many cost-sensitive industrial sensors. It is small, cheap and introduces negligible propagation delay.

An isolated part becomes necessary when ground potentials can differ between nodes (long cable runs, separate power domains, floating battery packs) or when safety and noise immunity demand it. Isolation also blocks the common-mode transients that otherwise walk straight into the MCU. Two implementation styles exist:

• Integrated solutions such as the ADI ADM3053 embed an isolated DC/DC converter (isoPower) so you get signal and power isolation in one package — fewer board parts, but you must respect the layout and decoupling rules the datasheet prescribes for the isolated supply.
• External-isolator solutions such as the TI ISO1050 use a separate isolated supply for the bus side, which gives more freedom in choosing the isolated rail but adds a regulator and transformer or isolated DC/DC to the BOM.

Isolation is not free: it adds propagation delay, which eats into the bit-time budget at high data rates, so confirm the loop timing still meets your bit-rate and cable-length targets.

Power, standby and silent modes #

Most transceivers run from a 5 V bus-side supply, but the logic interface (VIO) is increasingly 3.3 V to match modern MCUs directly. Confirm the VIO range before wiring TX/RX to an MCU that only has 3.3 V GPIO.

Low-power nodes — a door module asleep for weeks, a sensor on a battery-management string — need a real standby or silent mode with a wake path. Check the standby current, whether the part wakes on bus activity (via the INH or WAKE pin driving an external regulator), and whether the MCU can force the transmitter off during a fault so a stuck-dominant bug does not block the whole bus.

EMC and bus protection are part of the selection #

A transceiver alone does not make a compliant node. The common-mode range (typically ±12 V, wider on some automotive parts) sets how much offset the receiver tolerates; slope-control or slew-rate-limited versions reduce radiated emissions on slower buses. On the board, the usual protections still apply: 120 Ω termination at both ends of the segment, a split-termination network for the recessive bias, a common-mode choke for noisy environments, and TVS/clamp devices sized to the bus fault expectation.

Specify the transceiver's EMC and fault features against the actual cable length, bit rate and harness environment, not against a generic "CAN" label.

Selection checklist before you release the BOM #

1. Confirm the bus type: classic high-speed, CAN FD, or fault-tolerant low-speed — and pick a transceiver rated for the highest mode on that segment. 2. Decide isolation from the ground topology and EMC target, not from habit. 3. Match VIO (3.3 V or 5 V) to the MCU GPIO, and confirm standby/wake behaviour for low-power nodes. 4. Check package, AEC-Q100 qualification and the date code against the product's lifecycle window — long-life automotive programmes especially need traceable, current-date-code stock. 5. Verify the part is actually released and not in PCN/PDN limbo before committing it to a production BOM.

Representative published parts #

For a quick start, three published LimChip lines illustrate the split above:

• TJA1042T/3/CM,118 — NXP high-speed CAN transceiver for classic 1 Mbit/s buses, non-isolated, with standby mode.
• ADM3053BRWZ — ADI isolated CAN transceiver with integrated isoPower DC/DC, for nodes that need signal and power isolation in one package.
• ISO1050DUBR — TI isolated CAN transceiver using an external isolated supply, a flexible alternative when you already have an isolated rail.

If your segment is CAN FD, confirm FD support on the exact ordering code before specifying — not every part in a transceiver family is FD-capable.

Conclusion #

Selecting a CAN or CAN FD transceiver comes down to four questions: which bus mode the segment actually runs, whether the ground topology demands isolation, what logic voltage and sleep behaviour the node needs, and whether the package and qualification fit the product's lifecycle. Answer those and the analogue interface stops being a surprise source of EMC and bus-off failures. When you are ready to lock the BOM, confirm the exact ordering code, date code and lifecycle status against current stock rather than assuming the line is still active.