This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

EtherCAT Master and Slave

Other Parts Discussed in Thread: AM4377

I have a question about using the AM4377 in a redundancy mode using EtherCAT. We will be employing this processor with EtherCAT in a new system. It will use multiple modules, some processors and others IO. All will be connected via EtherCAT. We require redundancy between the processor modules. If the Master fails the Standby processor needs to take over. We have already developed and perfected this redundant system but using high speed multi-drop serial comms as the bussed architecture.

1) So, can anyone suggest a solution that allows a Standby processor to begin generating Ethernet (EtherCAT) frames if the Master processor fails in firmware or hardware? We already have a solution that maintains a mirror image of the database between the two processor modules so that in our existing system, when one processor fails, the other can take over.

2) There are a number of scenarios here. Our system is running Linux and our system application. This is what is most prone to failure. Is it true then in an EtherCAT connected system using the AM4377, that if the main core systems (Linux) fail, that the EtherCAT implementation in the PRU ICSS will keep running and generating frames and relaying data coming upstream back to downstream slave devices (for instance)?

Okay, possibly my understanding of Master and Slave EtherCAT relationships is incorrect. I read somewhere that you can have muiltiple EtherCAT Masters on the same EtherCAT 'circuit'. All Master and Slave functionality will be implemented in the AM4377 devices. External PC's are not involved in this application, other than to communicate with the processor modules via standard Ethernet IP, over separate ports.

Thanks for your help in advance.

Regards, Steve.

  • Hi Stephen,

    I will ask the Industrial team to comment on your use case.
  • Hi Steve,
    I assume for this discussion we need to distinguish between redundant system and a redundant network connection. EtherCAT does have means for redundancy in the network. Basically you use a ring and if one direction fails (due to broken device or cable failure) the master can switch over to the other direction. This feature is probably best discussed with EtherCAT technology group, Beckhoff or any master stack vendor. On the PRU ESC side we support redundancy and just follow the standard.

    But system redundancy is more complex I assume and I am not aware of standards. It is also independent of the network comms I assume.
    So for 1) I assume you need two processor systems where each can act as master. Just replace your current serial comms with EtherCAT.

    2) no, you can not assume the PRU ICSS to continue to act as a working EtherCAT device or master in case ARM side fails. We are only implementing layer 2 of EtherCAT on the PRU ICSS. So best case you will get packets forwarded in both directions but even that is not guaranteed to my mind.

    Best regards,
  • Hi Frank,
    Thanks for your reply. Yes, we will be using the redundant ring topology for our network. My main question was the possible application of two EtherCAT masters onto the one network, where (obviously) only one of them is active as a master at any one time. The serial multi-drop system we have easily lends itself to multiple masters as all the bus transceivers are RS485 multidrop, so the backup master can easily take over the bus if the original master stops transmitting. However I don't believe this is possible with EtherCAT. I suppose it may be possible if the backup master (which is downstream of the normal master) takes over but loses the network ring backup. I still have the issue of the backup master, which in it's standby mode must be operating as a slave, now needs to change over to master mode and begin generating EtherCAT frames.
    Best Regards
    Steve
  • Steve,
    I don't see an issue here with EtherCAT (or Ethernet in general). Any device on the Ethernet network can send and receive. So definition of master/slave is just a logical one. Software may change any device from master to slave and back. It is a matter of your system to correctly control this. I suggest you discuss the issue of multiple masters with ETG and ask them if there is a defined protocol or something to help with this. The question for sure would be if changing the master has any effect on the slaves and their operation. If the masters are not fully in sync ( or redundant units) this could be an issue.
    As our EtherCAT solution currently focuses on the slave ESC implementation this is really outside the scope of our deliverables.
    Regards,
  • Frank,

    Thank you for that. Yes, you are quite right, my questions are outside your deliverables. Thanks for your answers in any case. I shall get onto the ETG and pose the question there. I'm sure we will have a lot more questions regarding the AM437x when we start our design work in another month or so.

    Best regards

    Steve