Advanced Search

$ reboot world --recompile /universe/all --ty

The "tmax" command is like this:

peter@EliteDesk-800-G1-USDT:~$ halcmd show param *.tmax
Parameters:
Owner   Type  Dir         Value  Name
    33  s32   RW              0  hm2_7i76e.0.read-request.tmax
    33  s32   RW         760736  hm2_7i76e.0.read.tmax
    33  s32   RW          71032  hm2_7i76e.0.write.tmax
    23  s32   RW          18936  motion-command-handler.tmax
    23  s32   RW          48472  motion-controller.tmax
    27  s32   RW          18176  pid.0.do-pid-calcs.tmax
    27  s32   RW          18588  pid.1.do-pid-calcs.tmax
    27  s32   RW          20580  pid.2.do-pid-calcs.tmax
    27  s32   RW          22780  pid.3.do-pid-calcs.tmax
    24  s32   RW         795540  servo-thread.tmax


Have you disabled all power management (Turbo modes, EIST , C states > C1 etc) in your PCs BIOS?

This is most likely an issue with the host PC or its setup

What does a ping test show?

ping -c 4 -i .2 10.10.10.10
 and then
sudo chrt 99 ping -i .001  -q 10.10.10.10 

let the last command run for a few minutes and then hit control-C
to print out the ping statistics (substitute your cards IP address for the 10.10.10.10 if it differs)

also for a latency test, you should disable the base thread:

latency-test 1ms
 or
latency-histogram --nobase --sbinsize 1000
 

add https:// to the links. Be mindful of the splitting /.../ in the urls on this forum. Always better to right-click on urls and copying the actual link.

forum.linuxcnc.org/10-advanced-configura...tency?start=0#288870

I've tryied to compile a new kernel, but with now success. First of all, executing sudo apt upgrade resulted in an error: "dpkg: error processing package linux-image-rt-amd64 (--configure)" because: "Package linux-image-6.1.0-17-rt-amd64 is not configured yet",
after that, git clone -b v6.6 --depth 1 git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git linux-6.7 falted out saying that the repository doesn't exist.

I'm now trying LinuxCNC2.9.2 installed directly from the image on the website, update ASAP
 

I have two freezers, but they where turned off this past week, before this wave of errors happened

10x smaller is being overly kind, I'd say 20~30x rather. In desperate times comes desperate measures, hence the kernel trial in my case. So far so good, and yes particularly on the parport laser and 3ax mill I have running here.

Still recommend a try, who knows. I've seen enhancement on the ethernet based controllers too... not getting those infamous random 'unexpected delay on task 0' (whatever the reason for those is/was) anymore.

Q; what's on your mains circuit apart from the controller/motors and al? Not having a 3000watts radiator periodically switching on/off like I had a week ago(lol)? That caused me quite a bit of troubleshooting too...

It could make a difference for parport machines, while for mesa cards it the margins are smaller. The biggest problem for me is regarding the uspace thingy. If i recall correctly, some time ago, before many updates and debian version, the max jitter was 10x smaller

Quick shot in the dark, tried another kernel before thinking of going the re-install-all route? Beware - compiling takes quite a bit of time and may turned out to be a waste of time. Worth a shot anyway (I had luck with it). Others; cold'er temp and dry environment == buzzy and zappy. 

forum.linuxcnc.org/10-advanced-configura...tency?start=0#288870

UPDATE: i've run a latency test:
Max intervals:

• Servo thread: 1133566 ns
• Base thread: 191345 ns
  
  Max jitter:
  • Servo thread: 134850 ns
  • Base thread: 166345 ns

Kernel-release: 6.1.0-17-rt-amd64
Kernel-version: #1 SMP PREEMPT_RT Debian 6.1.69-1 (2023-12-30)

As per probe basic installation procedure, i've installed linuxcnc-uspace, than i found the following lines on linuxcnc.org/downloads/

“linuxcnc-uspace” is compatible with the preempt-rt realtime kernel, which is close to standard Linux. This version runs in userspace like most other applications and will run with any preempt-rt kernel.“linuxcnc-uspace” will also run on a standard Linux kernel for use as a simulator, but it should not be used to control hardware.

On kcjengr.github.io/probe_basic/quick_start.html :

sudo apt install linuxcnc-uspace linuxcnc-uspace-dev mesaflash

What i would like to avoid is to have to remake an entire disk as installing linuxcnc this way leads to terrible performance

Here are two pictures of the control panel, it's a mess, i have to tidy it up a bit, i know 

Hallo, die verkabelung scheint korrekt zu sein,
Grüße

Do you have a picture of your setup and wiring? Maybe we can directly see some emi issues. If it's finaly the problem I don't know. No matter what it is I would take care a bit of emi.

Hello everybody,
I've been experiencing the dreaded "error finishing read" for the last three days, here are the commonly requested things in this situation:

• Debian 12 Bookworm installed following the procedure listed on the probe basic website: kcjengr.github.io/probe_basic/quick_start.html
• Linuxcnc 2.9.0~pre1+git20230208.f1270d6ed7-1
• Ping to 7i76e in the 0.15/0.28 ms, CR3, CR4, CR5, CR6 count up as in normal operation
• halcmd show param *.tmax doesn't show anything, but looking at hm2_7i76e.0.read.tmax with halmeter shows values in the upper 12.000.000
• CR1 LED on after error
• Power supply output 24.5/24.7V with ripples of +- 0.1V
• Machine config @ github.com/EragonPower/LinuxCNCMachineConfiguration
• Computer: AK34 Pro w/ Intel Celeron J3455 quad core@1.5GHz 6GB LDDR4, Dual Realtek Ethernet RL8111/8168/8411
• The error happens with and without hyperthreading
• Safe boot and fast boot off
• Tryied to change ethernet cable from a standard server CAT5 to a CAT6 but nothing changed
  
  Something that i've noticed is that these error waves come more frequent in some days than others, for example a month ago it only happend once in a week, yesterday it happend a few minutes after homing, today it happend as soon as Probe basic started.
  Frankly i don't know what to do, the only thing that comes in mind is that on the 7i76e manual suggest to ground the mesa using the high-left hole near the ethernet port.
  I'll try to ground it, but i'm not so sure about it considering the EMI interference of the VFD ecc, but maybe it could resolve something as yesterday the error appared as soon as i started the spindle.
  
  It all looks like a witch hunt for me, maybe someone has a better idea. Thanks in advace to everybody that looks into this
  

Sounds much nicer. Sadly once you go over to the ethercat darkside you have to manually write your hal files wich takes me back to when there was no support in pncconf for the 7i76e (So I have mostly done my configs by hand anyway.

For spindle synchronized motion you need to set the encoder scale
to the number of encoder counts per revolution (and you need index to work)