CAN Bus Guide

Both ODrive Pro and ODrive S1 use a custom protocol called CANSimple. This currently supports CAN 2.0b and will support CAN-FD soon, at 12mbps for Pro, and 8mbps for S1. For more information, please refer to the CAN protocol page.

Note

This guide is intended for beginners to set up CAN on the ODrive and on their host device. We will be working with a Linux system as our host device, using Python, but the examples can be applied to any other systems.

Hardware Setup

ODrive assumes the physical layer is a standard differential twisted pair in a linear bus configuration with 120 ohm termination resistance at each end. The CANH (High) and CANL (Low) pins are used for CAN communication. Connect CANH to CANH and CANL to CANL (found on the datasheet pinout (Pro, S1)) for all devices on the bus, and ensure you have a good ground between each node.

  • Semi-isolated CAN FD network

    • CAN transceiver is powered by the onboard power

    • CAN GND should be connected to the system star ground point

Important

Make sure to connect CAN GND to DC- at the system ground star point.

CAN picture

Note

Many devices make use of a DE-9 (commonly mistakenly called DB-9) connector for CAN. The typical pinout is CANL on pin 2 and CANH on pin 7.

Important

If your ODrive is the “last” (furthest) device on the bus, you can use the on-board 120 Ohm termination resistor by switching the DIP switch to “CAN 120R”. Otherwise, add an external resistor. The nominal bus resistance should be 60 ohms between CANH and CANL.

Setting up the Host

On most host computers you need external hardware to communicate over CAN. For most applications, we recommend the following USB CAN dongles:

These work out of the box on Linux systems and you only need to enable it.

For Raspberry Pi and similar boards, there are CAN hats available, for example:

These require additional setup steps, see below.

Enable CAN Hat

  1. Open the /boot/config.txt with

    sudo nano /boot/config.txt
    
  2. At the bottom of the file, add the following:

    link to product

    dtparam=spi=on
    dtoverlay=spi1-3cs
    dtoverlay=mcp251xfd,spi0-0,oscillator=40000000,interrupt=25 # configure can0 interface
    dtoverlay=mcp251xfd,spi1-0,oscillator=40000000,interrupt=24 # configure can1 interface
    
  3. Reboot the device for these changes to take effect.

sudo reboot

Note

For more details regarding this process, checkout this in depth tutorial and this forum post

Enable CAN interface

Before you can use the CAN interface, it needs to be enabled with:

sudo ip link set can0 up type can bitrate 250000

This is required after every reboot. If this fails with RTNETLINK answers: Device or resource busy, the interface might already be up.

Important

Make sure this value matches the ODrive’s baud_rate, by default the ODrive uses 250 kbps (250000)

Setting up the ODrive (via USB)

The ODrive’s CAN interface is enabled by default. All you have to do is to assign a unique node ID to each ODrive on the bus so that they don’t conflict. For some applications you may also want to enable periodic feedback messages.

Under Configuration > Interfaces:

../_images/gui-can-config.png

Setting up the ODrive (CAN only)

This section describes how to set up an unconfigured ODrive in environments where only CAN communication is available, specifically for systems that do not have USB access.

Hint

If this is your first time using an ODrive, we recommend that you use the Web GUI first to familiarize yourself with the device and get a motor spinning. Since the GUI does not have CAN support yet, this requires a USB connection. Once it spins over USB you can follow this section to (re-)configure additional ODrives where USB is not available.

The process consists of three parts:

  1. Bootstrapping CAN communication via CAN

  2. Writing a previously configured set of configuration

  3. Running calibration procedures

Bootstrapping CAN communication via CAN

New in Firmware 0.6.9

To bootstrap CAN communication, use can_enumerate.py to assign each ODrive to its function.

In the simplest use case, it will scan the CAN bus for any ODrives and assign a random node ID to each one. For each discovered ODrive, it will print the serial number and node ID that it picked.

python3 /path/to/can_enumerate.py --channel can0 --save-config

If you want control over which ODrive gets which ID, you can run the script in interactive mode. To do so, pass pairs of label=node_id to the script (where label can be any user defined designation). For example, if your ODrive has two wheels and you want to assign node ID 0 to the left wheel and node ID 1 to the right wheel:

python3 /path/to/can_enumerate.py --channel can0 left=0 right=1 --save-config

The script will then randomly pick one ODrive on the bus and blink its status LED. It will then ask the user which ODrive has a blinking LED and thereby know which node ID to assign to it. This is repeated for all ODrives.

For more options, see can_enumerate.py --help.

Writing configuration

Once the ODrives are addressed, you can write all the other configuration variables.

  1. Prepare a file called config.json that has the following form:

{
  "config.dc_bus_overvoltage_trip_level": 40,
  "config.dc_max_negative_current": -10,
  "config.brake_resistor0.enable": true,
  "config.brake_resistor0.resistance": 2,
  "axis0.config.motor.motor_type": 0,
  ...
}

If you used the GUI for prototyping (see note above), you can export the set of configuration variables by going to Configuration > Apply & Calibrate, copying the Python snippet from there, and doing a bit of manual reformatting into JSON.

  1. Download flat_endpoints.json corresponding to your ODrive and firmware version from this page.

  2. Run can_restore_config.py. For example:

python3 /path/to/can_restore_config.py --channel can0 --node-id 0 --endpoints-json /path/to/flat_endpoints.json --config config.json --save-config

For more options, see can_restore_config.py --help.

Running calibration

Finally, to run the calibration sequence, you can use can_calibrate.py. For example:

python3 /path/to/can_calibrate.py --channel can0 --node-id 0 --save-config

For more options, see can_calibrate.py --help.

Your ODrive is now ready to spin a motor! See Code Examples for next steps.

Verifying Communcation

Once a node ID is assigned to an ODrive (see above), it will by default send a heartbeat message at 10Hz. We can confirm our ODrive communication is working by dumping the incoming CAN messages in the terminal:

sudo apt-get install can-utils # one-time install
candump can0 -xct z -n 10

This will read the first 10 messages from the ODrive and stop. If you’d like to see all messages, remove the -n 10 part (hit CTRL+C to exit). The other flags (x, c, t) are adding extra information, colouring, and a timestamp, respectively.

candump can0 -xct z -n 10
(000.000000)  can0  RX - -  001   [8]  00 00 00 00 01 00 00 00
(000.001995)  can0  RX - -  021   [8]  00 00 00 00 08 00 00 00
(000.099978)  can0  RX - -  001   [8]  00 00 00 00 01 00 00 00
(000.101963)  can0  RX - -  021   [8]  00 00 00 00 08 00 00 00
(000.199988)  can0  RX - -  001   [8]  00 00 00 00 01 00 00 00
(000.201980)  can0  RX - -  021   [8]  00 00 00 00 08 00 00 00
(000.299986)  can0  RX - -  001   [8]  00 00 00 00 01 00 00 00
(000.301976)  can0  RX - -  021   [8]  00 00 00 00 08 00 00 00
(000.399986)  can0  RX - -  001   [8]  00 00 00 00 01 00 00 00
(000.401972)  can0  RX - -  021   [8]  00 00 00 00 08 00 00 00

Alternatively, if you have python can installed (pip3 install python-can), you can use the can.viewer script:

python3 -m can.viewer -c "can0" -i "socketcan" which will give you a nice readout.

Code Examples

This section uses Python to illustrate how CAN communication with the ODrive works. You can download and run the full examples as-is to get started, and then go from there based on your needs. The examples are kept simple so that if you intend to use something other than Python, you can easily transfer them to your target language and platform.

Intro: Read Heartbeat

  1. Make sure all dependancies are install

    pip3 install python-can
    
  2. Create a CAN bus object and flush old messages

    import can
    
    bus = can.interface.Bus("can0", bustype="socketcan")
    
    # Flush CAN RX buffer so there are no more old pending messages
    while not (bus.recv(timeout=0) is None): pass
    
  3. Define the heartbeat message ID

    node_id = 0 # must match `<odrv>.axis0.config.can.node_id`. The default is 0.
    cmd_id = 0x01 # heartbeat command ID
    message_id = (node_id << 5 | cmd_id)
    

    Note

    CANSimple separates the CAN message ID into two parts: An axis ID and a command ID. Please refer to the CANSimple page for more information.

  4. Wait for the heartbeat, and print results

    import struct
    
    for msg in bus:
      if msg.arbitration_id == message_id
          error, state, result, traj_done = struct.unpack('<IBBB', bytes(msg.data[:7]))
          break
    print(error, state, result, traj_done)
    

Closed Loop Control

Important

Make sure your ODrive is fully connected and configured for closed loop control before continuing (this will require a USB connection). If you haven’t already, please see the Getting Started guide or the GUI Wizard to complete this step. Once finished, the USB connect is no longer required.

Full Python Example Here.

  1. Make sure all dependancies are installed

    pip3 install python-can
    
  2. Create a CAN bus object and flush old messages

    import can
    
    bus = can.interface.Bus("can0", bustype="socketcan")
    
    # Flush CAN RX buffer so there are no more old pending messages
    while not (bus.recv(timeout=0) is None): pass
    
  3. Put axis into closed loop control state

    import struct
    
    bus.send(can.Message(
        arbitration_id=(node_id << 5 | 0x07), # 0x07: Set_Axis_State
        data=struct.pack('<I', 8), # 8: AxisState.CLOSED_LOOP_CONTROL
        is_extended_id=False
    ))
    
  4. Wait for axis to enter closed loop control by scanning heartbeat messages

    for msg in bus:
        if msg.arbitration_id == (node_id << 5 | 0x01): # 0x01: Heartbeat
            error, state, result, traj_done = struct.unpack('<IBBB', bytes(msg.data[:7]))
            if state == 8: # 8: AxisState.CLOSED_LOOP_CONTROL
                break
    
  5. Set velocity to 1.0 turns/s. See also Set_Input_Vel.

    bus.send(can.Message(
        arbitration_id=(node_id << 5 | 0x0d), # 0x0d: Set_Input_Vel
        data=struct.pack('<ff', 1.0, 0.0), # 1.0: velocity, 0.0: torque feedforward
        is_extended_id=False
    ))
    
  6. Print encoder feedback. See also Get_Encoder_Estimates.

    Note that the ODrive sends the Get_Encoder_Estimates message periodically by default without explicit requests. To receive other feedback in the same way, see Cyclic Messages.

    for msg in bus:
        if msg.arbitration_id == (node_id << 5 | 0x09): # 0x09: Get_Encoder_Estimates
            pos, vel = struct.unpack('<ff', bytes(msg.data))
            print(f"pos: {pos:.3f} [turns], vel: {vel:.3f} [turns/s]")
    

Arbitrary Parameter Access

If the predefined CAN messages do not cover your needs, this example describes how to read or write any parameter that is acessible from the GUI or odrivetool.

Every parameter is identified by a number. This number can change across firmware and hardware versions. There a json file available for this name-to-id mapping for every release of the firmware, for every supported hardware.

Full Python Example Here.

  1. Download flat_endpoints.json from this page.

  2. Load flat_endpoints.json

    import json
    with open('flat_endpoints.json', 'r') as f:
        endpoint_data = json.load(f)
        endpoints = endpoint_data['endpoints']
    
  3. Define some constants we’re going to use further down:

    OPCODE_READ = 0x00
    OPCODE_WRITE = 0x01
    
    # See https://docs.python.org/3/library/struct.html#format-characters
    format_lookup = {
        'bool': '?',
        'uint8': 'B', 'int8': 'b',
        'uint16': 'H', 'int16': 'h',
        'uint32': 'I', 'int32': 'i',
        'uint64': 'Q', 'int64': 'q',
        'float': 'f'
    }
    
    node_id = 0 # must match the configured node_id on your ODrive (default 0)
    
  4. Initialize the bus object as shown in Intro: Read Heartbeat.

  5. Verify that the ODrive’s hardware and firmware are what we expect.

    This is an important safety measure because the endpoints ID in flat_endpoints.json can change with every firmware and hardware version.

    # Flush CAN RX buffer so there are no more old pending messages
    while not (bus.recv(timeout=0) is None): pass
    
    # Send read command
    bus.send(can.Message(
        arbitration_id=(node_id << 5 | 0x00), # 0x00: Get_Version
        data=b'',
        is_extended_id=False
    ))
    
    # Await reply
    for msg in bus:
        if msg.arbitration_id == (node_id << 5 | 0x00): # 0x00: Get_Version
            break
    
    _, hw_product_line, hw_version, hw_variant, fw_major, fw_minor, fw_revision, fw_unreleased = struct.unpack('<BBBBBBBB', msg.data)
    
    # If one of these asserts fail, you're probably not using the right flat_endpoints.json file
    assert endpoint_data['fw_version'] == f"{fw_major}.{fw_minor}.{fw_revision}"
    assert endpoint_data['hw_version'] == f"{hw_product_line}.{hw_version}.{hw_variant}"
    
  6. Write parameter

    import struct
    
    path = 'axis0.controller.config.vel_integrator_limit'
    value_to_write = 1.234
    
    # Convert path to endpoint ID
    endpoint_id = endpoints[path]['id']
    endpoint_type = endpoints[path]['type']
    
    # Send write command
    bus.send(can.Message(
        arbitration_id=(node_id << 5 | 0x04), # 0x04: RxSdo
        data=struct.pack('<BHB' + format_lookup[endpoint_type], OPCODE_WRITE, endpoint_id, 0, value_to_write),
        is_extended_id=False
    ))
    

    Raw messages, assuming node_id = 0, endpoint_id = 0x0182, value_to_write = 1.234:

    0x004  00 82 01 01 b6 f3 9d 3f
    
  7. Read parameter

    import struct
    
    path = 'axis0.controller.config.vel_integrator_limit'
    
    # Convert path to endpoint ID
    endpoint_id = endpoints[path]['id']
    endpoint_type = endpoints[path]['type']
    
    # Flush CAN RX buffer so there are no more old pending messages
    while not (bus.recv(timeout=0) is None): pass
    
    # Send read command
    bus.send(can.Message(
        arbitration_id=(node_id << 5 | 0x04), # 0x04: RxSdo
        data=struct.pack('<BHB', OPCODE_READ, endpoint_id, 0),
        is_extended_id=False
    ))
    
    # Await reply
    for msg in bus:
        if msg.arbitration_id == (node_id << 5 | 0x05): # 0x05: TxSdo
            break
    
    # Unpack and print reply
    _, _, _, return_value = struct.unpack_from('<BHB' + format_lookup[endpoint_type], msg.data)
    print(f"received: {return_value}")
    

    Raw messages, assuming node_id = 0, endpoint_id = 0x0182, return_value = inf:

    0x004  00 82 01 00
    0x005  00 82 01 00 00 00 80 7f
    
  8. Function call

    import struct
    
    path = "save_configuration"
    
    # Convert path to endpoint ID
    endpoint_id = endpoints[path]['id']
    
    bus.send(can.Message(
        arbitration_id=(node_id << 5 | 0x04), # 0x04: RxSdo
        data=struct.pack('<BHB', OPCODE_WRITE, endpoint_id, 0),
        is_extended_id=False
    ))
    

    Raw messages, assuming node_id = 0, endpoint_id = 0x0253:

    0x004  01 53 02 00
    

Troubleshooting

  • RTNETLINK answers: Device or resource busy

    The CAN interface may already be started. To be sure, you can disable it with sudo ip link set can0 up.

  • can.CanError: Error receiving: [Errno 100] Network is down

    Start the CAN interface as shown in Enable CAN interface.