Set up for slaves with distributed clock

[NicolasAubin]

Hi,

would it be possible to have some guidance or reference on how to set up slaves which have a Distributed Clock function.

Other question, would it be possible to parametrize to de-activate the DC function on a specific slave as well?

Thank you,

Nicolas

[NicolasAubin]

Hi,

just wondering how to set up the assign_activate: 0x0300 # DC Synch register

What are the possible values and what mode are they associated to?

I understand that if you want to start the slave without the DC behavior, you have to either not use the assign_activate or set it to 0x00.

Thank you for your help,

Nicolas

[NicolasAubin]

Hi,
posting as I made some progress on how to set up distributed clock for slave. You need the XML ESI file and look for your slave specifically as some ESI xml file from manufacturers may contain several device. Then you search the OpMode tag.

For the Beckhoff AM81xx, you would get:

<Dc>
	<OpMode>
		<Name>DC</Name>
		<Desc>DC-Synchron</Desc>
		<AssignActivate>#x700</AssignActivate>
		<CycleTimeSync0 Factor="0">62500</CycleTimeSync0>
		<ShiftTimeSync0 Input="0">0</ShiftTimeSync0>
		<CycleTimeSync1 Factor="-1">0</CycleTimeSync1>
		<ShiftTimeSync1>0</ShiftTimeSync1>
	</OpMode>
	<OpMode>
		<Name>Synchron</Name>
		<Desc>SM-Synchron</Desc>
		<AssignActivate>#x0</AssignActivate>
		<CycleTimeSync0 Factor="0">0</CycleTimeSync0>
		<ShiftTimeSync0>0</ShiftTimeSync0>
		<CycleTimeSync1 Factor="1">0</CycleTimeSync1>
		<ShiftTimeSync1>0</ShiftTimeSync1>
	</OpMode>
</Dc>

You can identify that several modes are available: Dc for distributed clock and SM synchron.

I had to do the following modification (based on some support from IgH Etherlab EtherCAT) in this repo code that I cloned locally to test. You need to modify the

ethercat_inteface/src/ec_master.cpp

  // check and setup dc

  // Original code of this repo
  // if (slave->assign_activate_dc_sync()) {
  //   struct timespec t;
  //   clock_gettime(CLOCK_MONOTONIC, &t);
  //   ecrt_master_application_time(master_, EC_NEWTIMEVAL2NANO(t));
  //   ecrt_slave_config_dc(
  //     slave_info.config,
  //     slave->assign_activate_dc_sync(),
  //     interval_,
  //     interval_ - (t.tv_nsec % (interval_)),
  //     0,
  //     0);
  // }

  //New part to consider AMI81xx drive distributed clock values
  if (slave->assign_activate_dc_sync()) {
    struct timespec t;
    clock_gettime(CLOCK_MONOTONIC, &t);
    ecrt_master_application_time(master_, EC_NEWTIMEVAL2NANO(t));
    ecrt_slave_config_dc(
      slave_info.config,
      slave->assign_activate_dc_sync(), //AssignActivate
      62500, //SYNC0 cycle time
      0, //SYNC0 shift time
      interval_ - 62500, //SYNC1 cycle time
      0); //SYNC1 shift time
  }

I then rebuild my local repo

rosdep install --ignore-src --from-paths . -y -r
colcon build --cmake-args -DCMAKE_BUILD_TYPE=Release --symlink-install
source install/setup.bash

And I can then get it to work with the assign_activate to the value identified in the ESI xml file for the device (here 0x700).

# Configuration file for Beckhoff AMI8000 DS402
 vendor_id: 0x00000002
 product_id: 0x0003010f
 assign_activate: 0x700  #  DC Sync0+Sync1, need modification to driver to work,  does not work as is, slave goes into PREOP + error

 rpdo:  # RxPDO receive PDO mapping (ouput from master)
   - index: 0x1600
     channels:
        - {index: 0x6040, sub_index: 0x00, type: uint16, command_interface: control_word} #Do not put default to autorize automatic transition
   - index: 0x1606
     channels:
      - {index: 0x607a, sub_index: 0x00, type: int32, command_interface: positioncmd, default: .NaN} #Works, drive the drive to 0 | sdos is int32, xml is uint32, int32 show errors?


 tpdo:  # TxPDO transmit PDO mapping (input to master)
   - index: 0x1a00
     channels:
       - {index: 0x6064, sub_index: 0x00, type: int32, state_interface: position}       #Position, uint32 allows sart up but create message issue: the position shows periodicly values of other tpdo. It should be int32!
   - index: 0x1a01
     channels:
       - {index: 0x6041, sub_index: 0x00, type: uint16, state_inferface: status_word}   #StatusWord
   - index: 0x1a06
     channels:
       - {index: 0x60f4, sub_index: 0x00, type: int32, state_inferface: followingerror} #Following Error, uint32 works but it should be int32!


 sm:  # Sync Manager
  - {index: 0, type: output, pdo: ~, watchdog: disable}
  - {index: 1, type: input, pdo: ~, watchdog: disable}
  - {index: 2, type: output, pdo: rpdo, watchdog: disable}
  - {index: 3, type: input, pdo: tpdo, watchdog: disable}

So I can get now the slave into OP with the DC mode running.

It can be verified via the ethercat config -v, which shows some distributed clock information.

Alias: 0
Position: 0
Vendor Id: 0x00000002
Product code: 0x0003010f
Attached slave: 0 (OP)
Watchdog divider: (Default)
Watchdog intervals: (Default)
SM2, Dir: Output, Watchdog: Disable
  PDO 0x1600
    PDO entry 0x6040:00, 16 bit
  PDO 0x1606
    PDO entry 0x607a:00, 32 bit
SM3, Dir: Input, Watchdog: Disable
  PDO 0x1a00
    PDO entry 0x6064:00, 32 bit
  PDO 0x1a01
    PDO entry 0x6041:00, 16 bit
  PDO 0x1a06
    PDO entry 0x60f4:00, 32 bit
SDO configuration:
  None.
IDN configuration:
  None.
Feature flags:
  None.
DC configuration:
  AssignActivate: 0x0700
         Cycle [ns]   Shift [ns]
  SYNC0        62500            0
  SYNC1      3937500            0

In my case this modification is hardcoded, so only valid for my specific slave and if there were two slaves with different DC parameters on the same bus this will not work.

I believe the ethercat driver ros2 repo code may benefit from having the following parameters:
//SYNC0 cycle time
//SYNC0 shift time
//SYNC1 cycle time
//SYNC1 shift time
that could be set inside the slave config file

Thank you,

Nicolas