disconnect error
25 Mar 2024 07:43 #296689
by bkt
disconnect error was created by bkt
after a long investigation, see my ethercat system sometime fails ... checking dmesg I discover these messages ...
[242992.697621] igb 0000:02:00.0 enp2s0: igb: enp2s0 NIC Link is Down
after these working domanin changes ...
[243208.231620] EtherCAT 0: Domain 0: Working counter changed to 0/20.
[243208.335623] EtherCAT WARNING: Datagram 00000000869a65ae (domain0-0-main) was SKIPPED 1 time.
[243209.235628] EtherCAT 0: Domain 0: Working counter changed to 20/20.
[243310.748313] EtherCAT 0: Domain 0: Working counter changed to 0/20.
And i have warning and datagram looses until a joint have follow error after these system sometime return to right domain (seeems after a sort of negotiation ... but I'm not an expert) and only see lcnc OFF, sometime Lcnc crasch .... so question is:
autonegotiation of some nic on chain (thinck some servo and ek1100 work as nic) can be the problem?
these is actual situation on that enp2s0 ( think is ethercat line):
:~$ ethtool -a enp2s0
Pause parameters for enp2s0:
Autonegotiate: on
RX: on
TX: on
Put autonegotiation in OFF state can solove the situation? ... my ethercat cable is 4 wire type after ek1100 because other servos have only ethercat 4 wire ....
regards
[242992.697621] igb 0000:02:00.0 enp2s0: igb: enp2s0 NIC Link is Down
after these working domanin changes ...
[243208.231620] EtherCAT 0: Domain 0: Working counter changed to 0/20.
[243208.335623] EtherCAT WARNING: Datagram 00000000869a65ae (domain0-0-main) was SKIPPED 1 time.
[243209.235628] EtherCAT 0: Domain 0: Working counter changed to 20/20.
[243310.748313] EtherCAT 0: Domain 0: Working counter changed to 0/20.
And i have warning and datagram looses until a joint have follow error after these system sometime return to right domain (seeems after a sort of negotiation ... but I'm not an expert) and only see lcnc OFF, sometime Lcnc crasch .... so question is:
autonegotiation of some nic on chain (thinck some servo and ek1100 work as nic) can be the problem?
these is actual situation on that enp2s0 ( think is ethercat line):
:~$ ethtool -a enp2s0
Pause parameters for enp2s0:
Autonegotiate: on
RX: on
TX: on
Put autonegotiation in OFF state can solove the situation? ... my ethercat cable is 4 wire type after ek1100 because other servos have only ethercat 4 wire ....
regards
Please Log in or Create an account to join the conversation.
25 Mar 2024 08:50 #296692
by rodw
Replied by rodw on topic disconnect error
Ethernet only use 4 wires. other wires can be used for phones as it uses different pairs but more often today for Power over Ethernet (PoE)
The following user(s) said Thank You: bkt
Please Log in or Create an account to join the conversation.
- scottlaird
- Offline
- Premium Member
Less
More
- Posts: 136
- Thank you received: 155
25 Mar 2024 15:20 #296728
by scottlaird
Replied by scottlaird on topic disconnect error
100 Mbps Ethernet is only 4 wires; Gigabit (and higher) use all 8, and some specialized variants only use 2 (100base-T1, for example).
In any case, it's unlikely that autonegotiation is your problem. Frankly, forcing autonegotiation with Linux has caused more problems for me than it's ever fixed, but YMMV.
My first guess would be a cabling or interference problem. Beckhoff recommends shielded cables to avoid interference. I'm not convinced that this is that important, but it's also fairly cheap.
Try looking at Linux's network counters by running `ip -s link`. You should see 0 (or very close to 0) frames in the errors/dropped/missed/carrier/collsns fields on your EtherCAT link(s). If any of these are large, and *especially* if they're growing while you're using the device, then you have a network problem, and swapping cables should be the first step.
In any case, it's unlikely that autonegotiation is your problem. Frankly, forcing autonegotiation with Linux has caused more problems for me than it's ever fixed, but YMMV.
My first guess would be a cabling or interference problem. Beckhoff recommends shielded cables to avoid interference. I'm not convinced that this is that important, but it's also fairly cheap.
Try looking at Linux's network counters by running `ip -s link`. You should see 0 (or very close to 0) frames in the errors/dropped/missed/carrier/collsns fields on your EtherCAT link(s). If any of these are large, and *especially* if they're growing while you're using the device, then you have a network problem, and swapping cables should be the first step.
The following user(s) said Thank You: bkt
Please Log in or Create an account to join the conversation.
- scottlaird
- Offline
- Premium Member
Less
More
- Posts: 136
- Thank you received: 155
25 Mar 2024 15:32 #296729
by scottlaird
Replied by scottlaird on topic disconnect error
On the interference front, some devices (like VFDs) can produce a huge amount of RF noise, and are known for causing problems with nearby devices. I've seen vendor recommendations around putting VFDs outside of your main CNC electronics enclosure just for this reason.
Twisted-pair Ethernet (any of the RJ45 variants, basically) is usually pretty good at dealing with interference, because it uses differential signaling across a pair of twisted wires. Generally, any noise will impact both wires equally, and differential signaling really only cares about the *difference* between the level on two wires, which should cancel out the noise.
There can be issues with ground loops and shielded Ethernet cables, if your two ends don't share a ground. This can happen with unshielded cables as well, but Ethernet PHYs have *some* protection against this, while the shields don't. I'd be shocked if that's an issue within a normal CNC cabinet, though. If it *is* an issue, then I won't be the one getting the shock...
Twisted-pair Ethernet (any of the RJ45 variants, basically) is usually pretty good at dealing with interference, because it uses differential signaling across a pair of twisted wires. Generally, any noise will impact both wires equally, and differential signaling really only cares about the *difference* between the level on two wires, which should cancel out the noise.
There can be issues with ground loops and shielded Ethernet cables, if your two ends don't share a ground. This can happen with unshielded cables as well, but Ethernet PHYs have *some* protection against this, while the shields don't. I'd be shocked if that's an issue within a normal CNC cabinet, though. If it *is* an issue, then I won't be the one getting the shock...
Please Log in or Create an account to join the conversation.
25 Mar 2024 16:15 - 25 Mar 2024 16:26 #296736
by bkt
Replied by bkt on topic disconnect error
@rodw -- These is know ... but because after many and many check (I've put ferrite ring in my underware now ... sorry for the French) ... not understand where problem is, I question my knowledge first so that comparison with third parties constitutes a solid starting point, for my mental sanity as first and for second because I have to check everything again ...
@ scottlaird never touch self-negotiation first ... but because not know how work ek1100 and asda2-E .... I think as last possibility can be a problem .... as I write first .... I add a lot of ferrite rings in every start and end of every twisted cable (max lenght 1,5mt), and for all other connection use only patch cable CAT6-A ... exept a long cable of 5mt (5mt not 50mt) not twisted but shielded that is CAT5-A .... not connect the shield to mobile part of machinery, and other side of shield is connect to rj45 and put it to ek1100 ... not sure ground is connect to ground of ek1100 .... these is a point. Other point running `ip -s link` not try, because it's new for me.
@ scottlaird never touch self-negotiation first ... but because not know how work ek1100 and asda2-E .... I think as last possibility can be a problem .... as I write first .... I add a lot of ferrite rings in every start and end of every twisted cable (max lenght 1,5mt), and for all other connection use only patch cable CAT6-A ... exept a long cable of 5mt (5mt not 50mt) not twisted but shielded that is CAT5-A .... not connect the shield to mobile part of machinery, and other side of shield is connect to rj45 and put it to ek1100 ... not sure ground is connect to ground of ek1100 .... these is a point. Other point running `ip -s link` not try, because it's new for me.
Last edit: 25 Mar 2024 16:26 by bkt.
Please Log in or Create an account to join the conversation.
- scottlaird
- Offline
- Premium Member
Less
More
- Posts: 136
- Thank you received: 155
25 Mar 2024 17:32 - 25 Mar 2024 17:33 #296745
by scottlaird
Replied by scottlaird on topic disconnect error
You can also run `ethtool IFNAME``, where IFNAME is your Ethercat interface name (eth0, etc). Here's a chunk of what I see talking to an EK1100:
It advertises 10/100, half or full. We decided on 100/full. Not exciting.
Link partner advertised link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
Link partner advertised pause frame use: No
Link partner advertised auto-negotiation: Yes
Link partner advertised FEC modes: Not reported
Speed: 100Mb/s
Duplex: Full
Auto-negotiation: on
It advertises 10/100, half or full. We decided on 100/full. Not exciting.
Last edit: 25 Mar 2024 17:33 by scottlaird.
The following user(s) said Thank You: bkt
Please Log in or Create an account to join the conversation.
06 Apr 2024 10:56 #297678
by bkt
Replied by bkt on topic disconnect error
these problem was Cat5E cable. I hope is not because re-install that 5mt cable was an hell, because structure of machine ... but after change seems all solve.
thanks
thanks
Please Log in or Create an account to join the conversation.
Time to create page: 0.107 seconds