Mesa 6i24 - high stepgen speed, axis go crazy

Today I tried to setup my CNC mill with Mesa 6i24 card. I used 5i24's firmware SVST2_8 with a custom breakout board.

I have configured all axis using PnCConf. I also tested them in PnCConf. When I used speed smaller than 2000 mm/min, everything worked fine. However when I went over 2000 mm/min, the axe "went crazy" - it didn't stop with the given acceleration; whole decceleration took about 3-4 seconds and it sometimes changed a direction. I thought the problem is in the drivers or steppers; I thought I could be using higher step rate than the driver is capable of, but I wasn't. Then I noticed the speed_rpm net reflected the current axe speed I was observing - this leads me to thinking the problem is inside the system (wrong speed command is issued).

Have you ever experienced similar behavior? What could be wrong? How to diagnose this problem?

A couple possibilities
1. Pncconf has a bug with stepgen tuning. Try changing the P terms for all axis to 1000
2. Make sure the 6I24 is jumpered for 5I24 emulation

Thank you for the answer. The card is in compatibility mode.

I am not right now by the machine, however I don't remember any PID setting during the stepgen setup. Where should I look it, please?

in the INI file under each [AXIS] section

Chris M

I wonder if you could zip and post your 5i24 XML file so I could test with it. Thanks

All my firmwares that I have compiled from source (including XML files) and PncConf configuration file can be found here: dl.dropboxusercontent.com/u/39138503/mill.zip

So if I understand correctly - this is a known bug in PncConf? This behavior is "normal" and its solution is manually change P constant in ini file? So there is no way how to test axe in PncConf?

cmorley: If you want my XML files to try then in PncConf, you also have to apply this PR to add support for 5i24 and 6i24: github.com/LinuxCNC/linuxcnc/pull/46

Pretty sure the latest version of pncconf has this built in, but it may not have been done by the 2.7.4 release

Also sounds like the following error limit setting should be tweaked for the stepgen configs
so any tracking problems give a instant following error.

(wide ferror settings make sense for initial servo configs to allow tuning but are not needed for stepgen configs)

yaqwsx wrote:

All my firmwares that I have compiled from source (including XML files) and PncConf configuration file can be found here: dl.dropboxusercontent.com/u/39138503/mill.zip

So if I understand correctly - this is a known bug in PncConf? This behavior is "normal" and its solution is manually change P constant in ini file? So there is no way how to test axe in PncConf?

cmorley: If you want my XML files to try then in PncConf, you also have to apply this PR to add support for 5i24 and 6i24: github.com/LinuxCNC/linuxcnc/pull/46

Yes that is what i would like to test. Thanks
Chris M

PCW wrote:

Pretty sure the latest version of pncconf has this built in, but it may not have been done by the 2.7.4 release

Also sounds like the following error limit setting should be tweaked for the stepgen configs
so any tracking problems give a instant following error.

(wide ferror settings make sense for initial servo configs to allow tuning but are not needed for stepgen configs)

Yes pncconf sets wide ferror for servo tuning I never thought to change them for PID steppers.
What would be reasonable settings?

Chris M

yaqwsx wrote:

So if I understand correctly - this is a known bug in PncConf? This behavior is "normal" and its solution is manually change P constant in ini file? So there is no way how to test axe in PncConf?

The fix was added after 2.7.4 release but you can edit the PID settings in pncconf - the axis test will honour them

Chris M