CAN bus installation best practices

01Plan before you touch the harness

A vehicle network is a shared resource. Every ECU on a bus depends on the same two wires, the same termination and the same bit timing. Change any of these, even slightly, and the result can surface weeks later as an intermittent warning lamp on an unrelated function. Good practice starts at the desk.

• Service information first. Get the wiring diagrams and topology for the exact model, year and equipment level; one badge can hide several architectures.
• Know the bus type. High-speed CAN to ISO 11898-2, fault-tolerant CAN to ISO 11898-3, single-wire CAN to SAE J2411 and SAE J1939 follow different electrical rules. See Bitrates and bus types.
• Stabilise the supply. Use a battery support unit whenever the ignition stays on for long; voltage dips cause spurious fault codes later blamed on the new device.
• Record the as-found state. Save the fault memory, note the termination reading and photograph the harness. You cannot prove you left a vehicle healthy if you never measured it.

Transport mode and delivery state

Many new vehicles leave the factory in a transport mode that disables convenience functions and reduces quiescent consumption; it is normally cleared at pre-delivery inspection. Installed before that step, systems behave differently, sleep timing changes and a sleep-current reading means nothing for the vehicle in service. Commission only once the vehicle is in its final customer configuration.

02Choosing the tap point

The tap point is where your device joins the bus electrically. Treat it as an engineering decision with electrical, mechanical, safety and serviceability criteria, and follow the vehicle-specific installation instructions wherever they exist.

• Electrical: a point on the backbone or at a node connector, close enough to the device that the stub stays short.
• Mechanical: dry and protected, away from heat, moving parts, hinges, seat rails and pedals. Footwells and sills, where water collects under carpets, are poor choices.
• Safety: clear of airbag, pretensioner and high-voltage harnesses. Never tap a wire in an airbag circuit, usually marked by yellow sleeving or connectors, and never run alongside orange high-voltage cables.
• Serviceability: reachable for later inspection without dismantling the cabin, and documented in the vehicle file.

Why the OBD connector is rarely the right tap point

The diagnostic connector looks convenient: CAN-H on pin 6, CAN-L on pin 14, grounds on pins 4 and 5, battery positive on pin 16. On current vehicles, however, those pins usually belong to a dedicated diagnostic bus behind a central gateway, not to the networks that carry the vehicle's operational data. It is also exposed in the footwell and needed by every workshop; a permanent device there blocks the port, is easily kicked loose and can collide with diagnostic sessions. Treat the OBD-II connector as a service port, not as a supply or data point; OBD-II and secure gateways explains why.

Stub length and topology

High-speed CAN is a line topology: one backbone, terminated at both ends with 120 Ω, with short stubs to each node. Every stub is an unterminated branch that reflects part of each edge; ISO 11898-2 is built around stubs of about 0.3 m at 1 Mbit/s on a 40 m bus. Most passenger-car buses run at 500 kbit/s, which leaves a little more margin; CAN FD networks with a 2 or 5 Mbit/s data phase leave less. The stub includes the wiring inside the device up to its transceiver, so a 1 m pigtail plus 50 cm of tap wire is already a 1.5 m stub.

03Keeping the twisted pair intact

CAN is differential. Every bit is the voltage difference between CAN-H and CAN-L, and the twisted pair makes interference couple equally into both wires, so the difference stays clean. Untwist it and you have two antennas. Vehicle harnesses typically use a lay length of around 20–25 mm per twist, and many manufacturers' repair instructions limit any untwisted section at a joint to roughly 40–50 mm.

• Never cut the backbone to insert a device in series. A CAN node is a parallel connection; routing the bus through a device makes the whole network depend on your connector.
• Never extend CAN wires to reach a convenient spot. Move the device, not the bus. Extensions lengthen the backbone or the stub and are often made with untwisted wire.
• Keep polarity. CAN-H to CAN-H, CAN-L to CAN-L. A swapped pair at one joint inverts the signal for everything beyond it.
• Twist your own leads to the harness standard and route them away from ignition coils, inverters and high-current cables.

Solder, crimp or connector: choosing the joint

The joint is the one thing you add to the vehicle's own wiring, so it must be at least as reliable as the harness around it. Vibration, temperature cycling and moisture kill joints, not the first day in service.

Whatever the method, stagger the CAN-H and CAN-L joints so they do not chafe against each other, keep the untwisted length minimal, and restore the harness to its original routing and fixings.

04Power and ground: +30, +15 and 31

Many installation faults that look like network faults are really power faults. The terminal designations of DIN 72552 remain the common language of European workshops:

Identifying the circuits with a meter

Use a high-impedance digital multimeter (10 MΩ input), never an incandescent test lamp. A test lamp draws enough current to damage electronic outputs and to wake or confuse control units, and it must never touch a CAN wire.

1. 01
   Measure with the ignition off

   A terminal 30 circuit reads battery voltage, roughly 12.4–12.8 V on a rested 12 V lead-acid battery. A terminal 15 circuit reads close to 0 V.

2. 02
   Switch the ignition on

   Terminal 15 should rise to within a few tenths of a volt of battery voltage. If it rises only part of the way, you are on a signal line or a monitored output: do not load it.

3. 03
   Crank the engine

   Note which circuits drop while the starter turns. A device that must survive the start belongs on terminal 30.

4. 04
   Switch off and time the fall

   Terminal 15 should fall to near 0 V within seconds. A line lingering at several volts is being held up, often by an earlier accessory feeding back.

5. 05
   Prove the ground under load

   Measure the voltage drop from the ground point to the battery negative post with a load running; a common workshop limit is about 0.1 V.

The +15 load problem

On older vehicles terminal 15 was a heavy wire from the ignition switch. Today it is often generated by a body control module through an electronic high-side switch with current monitoring, or it exists only as a state on the network. Loading such a circuit causes three classic problems:

• Overload and fault codes. The driver sees more current than designed, logs a fault, limits the current or switches off.
• Back-feeding. A device with its own terminal 30 supply can push a few volts back into the terminal 15 line through its input stage. Other units then see a ‘half ignition’, stay partly awake or store implausible ignition states.
• Delayed switch-off. Input capacitance holds the line up after ignition off, confusing modules that watch for the falling edge.

The correct pattern is simple: power the device from a fused terminal 30 supply and use terminal 15 only as a high-impedance sense input. If a load genuinely must follow the ignition, drive a relay coil from the ignition signal, once you have confirmed the circuit can carry the coil current, and switch the load from a separately fused terminal 30 circuit. Where ignition is a network state rather than a wire, follow the manufacturer's body-builder guidance instead of borrowing a random switched feed.

Fusing

A fuse protects the wire, not the device: size it for the smallest conductor and place it as close to the source as possible. Give each add-on its own fuse, and never share one with a safety system, a powertrain control unit or a body control module output. With a fuse tap, find which terminal of the slot is live with the original fuse removed and put the tap's supply leg on that side; the wrong way round, the original fuse ends up carrying both circuits.

Ground loops and shared ground

CAN tolerates ground offset because receivers look at the difference between the wires, but only within the transceivers' common-mode range, classically −2 V to +7 V for ISO 11898-2 high-speed devices. A ground loop that is harmless when parked can produce a large offset while the starter turns, and it injects noise into every single-ended output, such as a pulse line to a taximeter or tachograph.

Worked example. A starter draws 200 A. With 1 mΩ in the return path between engine block and body, the body sits 0.2 V away from engine ground during cranking, which is tolerable. Let corrosion raise that to 5 mΩ and the offset becomes 1 V, plus starter ripple, precisely when the vehicle is starting.

• Ground the device at the same reference as the units it communicates with, or at the manufacturer's designated ground point.
• One ground path per device; if an attached instrument has its own ground, share one point instead of closing a loop.
• Never ground to painted panels, seat frames, hinges or brackets mounted on plastic.
• Verify with a voltage-drop test under load: an ohmmeter cannot see a joint that fails under current.

05Sleep current and battery drain

A parked vehicle is not switched off; it is asleep. After ignition off and locking, network management lets the control units agree to stop communicating, the transceivers enter low-power mode and the vehicle settles at a quiescent sleep current, typically a few tens of milliamps for a passenger car. Getting there takes a few minutes on some vehicles and half an hour or more on others, with periodic wake-ups for connectivity, battery monitoring or, on electric vehicles, 12 V top-up.

An add-on device's own quiescent current adds to the vehicle's. More seriously, a device that transmits on the bus, or keeps a control unit active through terminal 15, can stop the network from sleeping at all; the vehicle then draws amperes, not milliamps.

The last row is why sleep behaviour must be checked after every installation: the network's ability to fall asleep matters far more than the device's own consumption.

Measuring sleep current correctly

1. 01
   Prepare the vehicle

   Close the doors, hold the bonnet and tailgate switches closed and keep the key out of range.

2. 02
   Connect without breaking the supply

   Use a DC clamp meter with milliamp resolution or an ammeter with a bypass switch; disconnecting the battery wakes the network.

3. 03
   Lock, wait and log

   Wait for the complete sleep sequence given in the service information and record current against time; one reading reveals nothing about periodic wake-ups.

4. 04
   Compare with the as-found value

   The difference before and after installation is the device's real contribution and should match the supplier's documented quiescent current. New wake-up spikes point to the installation.

06Electric and hybrid vehicles: READY is not ignition

Electric and hybrid vehicles separate states that a combustion vehicle merges. ‘Ignition on’ powers the electronics; READY means the high-voltage system is live and the vehicle will move when the accelerator is pressed, silently. The network can also be fully awake with nobody inside: while charging, pre-conditioning, running departure timers or topping up the 12 V battery through the DC/DC converter. CAN in electric and hybrid vehicles covers the architecture.

• Never route add-on wiring along orange high-voltage cables; only qualified persons work near them.
• Terminal 15 does not tell you whether the vehicle can move; READY does.
• Expect a busy bus during charging. A device that assumes ‘no ignition, no traffic’ will misbehave.

07Commissioning: test with all four wheels on the ground

On most current vehicles the road speed carried on the network is calculated by the braking control unit (ABS/ESC) from all four wheel-speed sensors, with plausibility checks between them. On a two-post lift with only the driven axle turning, two wheels move and two stand still; the vehicle may report a reduced or invalid speed, light ABS and stability-control warnings or enter a fallback mode. A single-axle roller dynamometer does the same unless the vehicle is in its dedicated dynamometer mode. A function test on a lift therefore proves nothing.

1. 01
   Re-measure termination

   Network asleep or battery disconnected: the reading between CAN-H and CAN-L must match the as-found value, normally about 60 Ω. About 40 Ω means a terminating resistor has been added.

2. 02
   Check bus voltages

   Ignition on: CAN-H slightly above 2.5 V, CAN-L slightly below, adding up to roughly 5 V.

3. 03
   Read the fault memory again

   New communication faults point to the installation.

4. 04
   Road-test on all four wheels

   Drive through stops, turns and a range of speeds and confirm the device output from standstill upwards. For a regulated instrument, follow its own installation and inspection procedure.

5. 05
   Verify sleep and document

   Repeat the sleep-current check after the road test, then record readings, joint locations, fuse positions and photographs in the vehicle file.

Taximeters and tachographs follow their own calibration rules; see Taximeter speed signals and Tachographs and the 2026 rules. When something misbehaves, turn to Field diagnostics.