refClockSyncCycles="-1", I got Messages.

I found lot of messages
"Failed to get reference clock time: Input/output error"
in LinuxCNC Console 
when I set "refClockSyncCycles="-1" to EtherCat Master configuration.

But if I set refClockSyncCycles="1", I could not see such messages.
I attached
refClockSyncCyclesMinus1.linuxcnc.txt (LinuxCNC Console Log)
refClockSyncCyclesMinus1.dmesg.txt (dmesg Log)
when I execute with  "refClockSyncCycles="-1"

And also, for reference, I attached 
refClockSyncCycles1.linuxcnc.txt (LinuxCNC Console Log)
refClockSyncCycles1.dmesg.txt (dmesg Log)
when I execute with  "refClockSyncCycles="1"
 

github.com/linuxcnc-ethercat/linuxcnc-ethercat/pull/461

There is a bug in the linuxcnc-erhercat if you just set the negative option

The problem is most likely related to the issue with  LWR and LRD/LRW modes that was discussed earlier.
You must at least get that fixed before you can rely on that error message.

automata has a solution. You cannot run lcec with those drives. automata attached
the code for another component to use with the Yaskawa drive.
attached as C file with a Makefile. Download and compile the code.
Use it as a component in hal.
I know it can be a handful to do this, I would foresee some experimentation to get this right.
Edit: I looked in the code and that isn't a ready-to-use component.
There is a lot of work needed to get this to work.

Frankly tar_san, I understood you have the drives on lone for evaluation, right?
Do yourself a favor and return them and get some other servo drives.
I can not spend that much time in getting this going. Hopefully someone else can.

Hi Hakan,
I have been able to wedge the double domain stuff into lcec_main.c and got a single yaskawa drive working with lcec_generic
For now I have made the double domain an option for the master. But I still need to check how to make it work only for yaskawa slave without affecting the other non-yaskawa slaves.
I'll try to wiggle that option into only the needed syncmanager or the required pdos.
Unlike fdarling, I did not need to disable fsoe part. So that part is good.
Will post more updates towards the end of the week.
-automata

Thanks everyone. 

Now I can make Yaskawa Sigma X  OP state and jog move from LinuxCNC stably.  ( Just did EoE disabled. )
Slave : SGDXS-xxxxA0x EtherCAT(CoE) SERVOPACK Rev02.09

According to the specification on Web, ( www.e-mechatronics.com/product/servo/sgm...c/xs_ether/spec.html)

LRD, LWR and LRW   seem to be supported. (attached image)
 
​​​But Sorry, I'm not familier with this problem.

Which configuration seems to be better in this situation?

refClockSyncCycles="-1" , with messages "Failed to get reference clock time: Input/output error"
refClockSyncCycles="1" no messages

Should I check some error from log? 



 

+1 if you have good latency

Ok, well that went over my head. Was so focused on the LWR issue.

You need to get rid of most of the "UNMATCHED" messages in the syslog.
"UNMATCHED " are usually caused by bad latency or very commonly by using RealTek network adapters.
If you have a Realtek network adapter, you should install the r8152-dkms or r8168-dkms driver via apt.

Can you show the output of "lspci -k"? Only the network interfaces are of interest.

Thank you so much.
I had been using N100 PC (or such Atom CPU )  for CNC machine, but this time, I installed LinuxCNC to Dell small PC ( 4 core i5 CPU ) just because of having smaller latency and more CPU power.

I have attached lspci - k output. 

I use N100 miniPC with good success. Anyway.

Yes, you have a Realtek network adapter.
 

Please install the r8168-dkms driver
"sudo apt install r8168-dkms"
 

Did you build ehtercat yourself?
My two cents, don't use the official build, make one yourself with disable EoE disabled, and use good shielded cables.