Parport to 7i92

Hello
I am running for several years a parallel port/gecko G540 linuxcnc for my home built little cnc and it was fine.
Motherboard is GIGABYTE GA-D525TUD.
From time to time, I had the well known popup window « Unexpected realtime delay...  » but I was able to machine my parts.
So lately I ordered a Mesa 7i92 and switched to V2.8 to have the real time part done by the card and to get rid of this warning message.
But, with the 2.8 kernel, latency is horrible (see screenshots in the attached files).
In PNCconf, on the Axis configuration page when testing motors, I have motor stalling randomly, even at lower mm/min than with V2.7.

Is it due to the poor latency performance or should I look for errors in my configuration files ? (HAL and INI for both version are also in the attached files)
Thanks
 

1 I don't think the pncconf axis testing is going to work, you need to launch LinuxCNC to test

2. The D525 is a bit marginal for Ethernet at a 1 ms thread period so I would change:

setp hm2_7i92.0.dpll.01.timer-us -50
to
setp hm2_7i92.0.dpll.01.timer-us -200

and if you get real time warnings, increase the thread period
(to say 2 ms)

Note that for latency testing when using Mesa cards you don't
need a base thread so you will get more accurate results with

latency-histogram --nobase --sbinsize 1000

@ tommylight
uname -a
Linux Faiseuse 4.19.0-17-rt-amd64 #1 SMP PREEMPT RT Debian 4.19.194-3 (2021-07-18) x86_64 GNU/Linux


2.8.1 iso then I updated

Thank you for your quick answer

@PCW
1) With my current configuration, in linuxCNC, it stalls quickly when I jog at full speed. in pncconf it takes a bit longer with the +- 50mm test.

2) I will make the changes and test.

Many thanks for your help

Hello,

I made the tests within LinixCNC using your sugestion
* setp hm2_7i92.0.dpll.01.timer-us -200
* thread period : 2 ms
and it's much better I can now reach 1900 mm/min

Still have a few things to tweak, but I should be able to mill soon

Many thanks for your hep

Hello
Back to my conversion
I now have some errors I don't find any information about :
Randomly I have this full list of error poping up. Error queuing read.... Error queuing write..

And an other error: after homing, I can jog with no problems. But most of the time, when I start the spindle, the limit switch from x or/and y trip (also in hal scope). I will get an oscilloscope to check this.
My spindle and vfd are on rs485. Limits switch are o P1-17 to P1-22

As tommylight is suggesting, the errors point to real time
issues on the host

Also check 7I92 ping times
ping 10.10.10.10 (or 192.168.1.121)

IRQ coalescing is off if your motherboard has
an Intel MAC (this is described in the hm2_eth manual page)

@tommylight
I will make another latency histogram since I update linux packages yesterday, the August one may be outdated. I know my latency is not so good with the kernel that comes with 2.8.x Following your advices, I set the servo thread to 2ms to get ride of « Unexpected realtime delay...  »

@PCM
I'll make those checks as soon as I'll go back to my workshop.

But I didn't have those kind of issues with my previous setup : my limit switches (all hal sensors) were connected on the G540 behind the 7i92.
X min  + X max switches on INPUT 1
Y min  + Y max switches on INPUT 2
Z min  + Z max switches on INPUT 3
I soldered diodes 1N4007 and proper pull-ups on a prototyping board and it was working fine. But it was messy.
But since I added the 7i92, I have plenty of IO. So I added an 8 channels opto board for P1 and now I'm on P1-17 to P1-22, each sensor on his own input. Much cleaner.

But with I start the spindle, on a oscilloscope in "single shot mode", I can see (just like in hal scope) P1 inputs randomly triggered.

And look at the 5V on the 7i92 power connector in the attached. (+5V is coming from the PC USB)
I suspect my yl620a VFD is heavily polluting my setup.
Don't you think it may also crash the 7i92 with such a power supply and lead to queuing errors?