Mesa 7i96s a little jitter

You can also just re-enter (up arrow)

ethtool -C eno1 rx-usecs 0

and ethtool will complain that it's already set...

Hi all, so I have tried the mesa with the added command line of ethtool -C eno1 rx-usecs 0. The ping reports an avg of 0.128ms vs 0.305ms so good improvement. When I run halscope now things look better however not as good as that which pcw generated. I have 2 questions. I seem to have to run the ethtool command everytime the pc is restarted - is there a way to make the change permanent?
Second question - to disable Irq coalesce I plan to follow the guide and edit the etc network interface file - should I be following it exactly? (Seen as I have already set up using the ‘network manager way’). Sorry about the questions but Linux is completely foreign to me.
Thankyou!

Thanks for the reply Tommy. Theres a session and startup with application autostart - what do I need to add there? I edited the network interfaces file with this # source /etc/network/interfaces.d/*
And this
hardware-irq-coalesce-rx-usecs 0
now everytime I restart I get an additional connection icon at the top right of the screen! They all have the same name and all show active when I connect the mesa. Any ideas?? I knew I’d get in trouble here - sigh…

Yeah it still works - il figure the multiple network manager applet opening later. I tried adding the static ip etc from the other tutorial and that rendered things dead so am back to where I started but with the improved ping response. Attached scope is of things now - some large spikes which is disappointing… any thoughts?
thanks

After reading some more I’m wondering if I should have or whether I do have the dpll timer enabled. Could this explain the spikes I’m seeing in the dpll phase error?

The big spikes are expected (those are from host latency)

To get an idea how bad they really are, you need to set the scale much larger
(say 200 per div or 0.2 ms per division)

Ok thanks pcw I will do that. When I compare your scope plot (attached) from here forum.linuxcnc.org/38-general-linuxcnc-q...7i76e-solved?start=0
to mine (also attached) mine seems to have magnitude 10 worse joint follow error (after conversion to mm) and at least double the dpll phase error plus the spikes. Your joint follow error also appears to be zero at constant velocity where as mine is not. I’m trying to get it similar to yours as that’s what it should look like correct?

Yes, it looks like you are not using the DPLL.

Can you post your hal/ini files?