EPass and EWarn reset in TMS320F28377D

Maulik,

            A "collision" should never happen in a CAN bus. What could happen is "contention" which is resolved by bit-wise arbitration. Remember CAN is termed CS/MA/CA (Carrier Sense/Multiple Access/Collision Avoidance ---- not Collision Detect , like Ethernet) . Are the slaves transmitting with the same MSGID? If the slaves are transmitting with different MSGIDs, I don't see how a situation like what you have described could happen.

Dear Hareesh,

Thanks for reply.

I had used collision instead of contention, that was my mistake. Thanks for clarifying about bit-wise arbitration.

Yes, Slave is transmitting with different MSGID. But i got error(Epass, EWarn and RX msg error counter full). How can i overcome with this situation? Which kind of take care should i take?

Note:  When there is only 1 slave(one to one communication) then there is no problem everything is healthy but when 2nd slave is connected (Network) at that time problem is started.

Regards,

Maulik

Maulik,

Contention should not cause any error. The protocol is designed to resolve contention gracefully. Are you sure the slaves are not putting out the same MSGID on the bus? While this may not cause a problem during arbitration phase, it will generate error when the bits in the data-bytes start differing. There is some other problem, I think.

Have you tried a lower bit-rate? Say, 50 kbps, for example?

Are you using an external clock source or the INTOSC for 28069? If you are using INTOSC, do you perform temperature compensation of INTOSC?

What is the end-to-end bus length?

Has the bus been terminated correctly (with 120-ohms) at either ends (only)? (The bus must be terminated only at either ends and with a 120-ohm resistor. IOW, no more than two terminator resistors may be present on the bus, unless split termination is followed, in which case there will be two resistors on either ends).

Have you tried to reduce the bus length?

Dear Hareesh,

1. No, slave is not putting data with same MSGID on bus. MSGID is different.
2. Problem exist with reduced speed.
3. I'm using External clock source for 28069. (20Mhz and afterward it will boost through internal PLL to 90Mhz)
4. I'm using optical fiber for communication.
5. Is there any need for termination when i use optical instead of wired connection. what do you think?
6. In case of optical fiber, I think bus length doesn't matter.

Special Note: Im sending NMT command to all slave -> all slave receive it and reply PDO at a time with different MSGID.
Is there any problem with this process? Because when i am sending NMT at that time only problem occure. And one more thing, Problem does not exist when there is single slave on network.

Regards,
Maulik Timbadiya

Maulik,

            It is important to first determine if this is a physical layer (hence H/W) issue or a software issue. I would like to eliminate CANopen from the picture and determine whether the physical layer is good. i.e. if the 3 nodes in your system (1 x 28377 & 2 x 28069) are able to communicate with each other without any problem. Try "one node to another node" and "one node to the other two nodes (broadcast)". Once we are convinced that the physical layer is perfectly fine, we can turn our focus elsewhere.

Optical fiber is a somewhat unusual choice for the physical layer. Admittedly, the CAN protocol specification does not mandate a specific physical layer. That being said, I am curious how the bus has been implemented. Specifically, how the ACK signal is sent to the transmitter. Yes, termination applies only to "regular" wires.

I find some contradiction in your post:

You say "I am sending NMT command to all slave -> all slave receive it and reply PDO at a time with different MSGID". This seems to indicate that there is no problem in the physical layer and all 3 nodes are able to communicate with each other.

But, in the next line, you say "Because when i am sending NMT at that time only problem occurs".

Dear Hareesh,

No issue with physical layer. Because, Network working properly in normal condition. Problem happens only when i issue NMT command for "system ready check" and "fault reset". For example, when i issue NMT command to all node to reset fault (Not can fault, This is panel external fault not related to CAN) . both node respond at same time with different MSGID (With updated status of fault) . I can see in the waveform that network work properly as per timing set and all. But when NMT is issue at that time both slave fall down into continuous retransmission after some time Master fall down into BUS OFF state. And Due to Auto Bus On feature network working properly after sometime.

Note: This is not happening all the time when NMT command is issue but Sometimes only. Approx. 1 out of 10 trial.

How can i solve this issue. Any specific software care should i take? Please suggest.

Regards,

Maulik

Dear Hareesh,

Before One year: I have added small delay to the slave as per id alloted and observed it is working fine.

(Already done the same thing as u said)

But now a day i have a question: as per CAN specification, if data transmit at same time may not cause any problem then why this is happening?

If both slave started to transmit at same time then the lower id always win arbitration as per protocol.

1. I have confirmed that both slave transmitting with different MSGID.

2. In my case when this kind problem arias, bus will go in to continuous retransmission. After random time it will be covered back. That time may be 1 to 3 min.

Thanks and regards,

Maulik

Maulik,

            You are correct that two nodes (with different MSGIDs) may start transmission at precisely the same time and that the node with the numerically lower MSGID will win arbitration. The key phrase here is " with different MSGIDs ". You have already confirmed that the two slaves have different MSGIDs. If that is indeed the case, then what you describe should not be happening.

Hard to debug this further without more knowledge of the system. Could you work with your local TI FAE to share the schematics of the CAN portion of your system? I also need to look at the scope capture of the CANTX pin of both slaves.