Upgraded to 2.7.13 -- W T F --

tecno wrote: OK, I follow what you say.

So is it time to play with new numbers on P = 1000 on each axis and SERVO_PERIOD ?

1000000 or 1.000.000 without dots, that is one million,
On your reply you used one hundred thousand and three hundred thousand, that will not work properly.

tommylight wrote:

tecno wrote: OK, I follow what you say.

So is it time to play with new numbers on P = 1000 on each axis and SERVO_PERIOD ?

1000000 or 1.000.000 without dots, that is one million,
On your reply you used one hundred thousand and three hundred thousand, that will not work properly.

That was a typo in my reply

Here a new test this morning

First part is with 1000000 setting and from line 100 the setting is 2000000



this picture with 2000000 and only keyboard jogging and it fails after a while.

This is all quite weird as the differences between 2.7.12 and 2.7.13 are not that significant and not related to Ethernet communications at all. I think something else must have changed when you updated, perhaps specific to your PC

Can you run a latency test on you PC to make sure something bad has not happened to real time...

something like

latency-histogram --nobase

and run it for 1/2 hour or so then use latency-histograms screenshot option to get the image and post it here

Also if you have an Intel MAC you need to set irq coalescing off (if you did not automate this setting you would lose it when you re-booted)

Hi Peter,

Will do that after my dinner.

Also if you have an Intel MAC you need to set irq coalescing off = sorry I do not understand what you say here, advice pls.

Only change on 2.7.12 was adding Touch and the update / upgrade

Do you have a intel MAC? (check with dmesg | grep eth)

The hm2_eth manual page describes setting the RX IRQ coalescing option

man hm2_eth