Debugging sporadic read errors from Mesa 7i95

More
6 years 6 days ago #187171 by piranha32
Hello,

I get sporadic read errors from my Mesa card, and I would like ask for an advice how to debug them.
The card most of the time works fine, but when I leave linuxcnc running, after some time, from several hours to a day or two, communication stops, and I get an error similar to this:

hm2/hm2_7i95.0: error finishing read! iter=79413542

Besides showing the error and killing communication, everything else seems to work ok (LinuxCNC does not crash). During the this time system is completely idle, with screen blanked.
Card is connected directly to the ethernet port on the computer, everything is sitting on my desk in office-like conditions (no heavy EMI), all peripherals are disconnected from the board.
The card is powered from the same 5V power supply as the computer.

The system runs on AtomicPi board with Atom x5-Z8350, OS is Debian 10.6 with kernel 4.19.0-11-rt-amd64 #1 SMP PREEMPT RT, installed from live LinuxCNC image. LinuxCNC version is 2.9.0-pre0 installed from Github source.

Thank you,
p

Please Log in to join the conversation.

  • BeagleBrainz
  • New Member
  • New Member
More
6 years 6 days ago #187178 by BeagleBrainz
Replied by BeagleBrainz on topic Debugging sporadic read errors from Mesa 7i95
What are the Amperage rating of the PSU for the Atomic Pi ?

Please Log in to join the conversation.

More
6 years 6 days ago #187214 by PCW
Can you run this command:

halcmd show all *.tmax

with LinuxCNC running and post the results along with the CPU speed?

Please Log in to join the conversation.

  • BeagleBrainz
  • New Member
  • New Member
More
6 years 6 days ago #187216 by BeagleBrainz
Replied by BeagleBrainz on topic Debugging sporadic read errors from Mesa 7i95
I know from the redit forums that many people aren't using the recommended PSU for the Atomic Pi. The sellers recommend a 5v 4A supply. If the OP has a borderline PSU would this be a contributing factor to the errors he is seeing ? He does mention both the Atomic Pi & 7i95 are powered by the same supply.

Or would it be that the board is a little underpowered CPU wise ?

Please Log in to join the conversation.

More
6 years 6 days ago #187220 by PCW
It could be power supply related. It could also be
power management that has not been disabled,
or the CPU is just slow (a 2 ms Servo thread might solve the issue)

Please Log in to join the conversation.

More
6 years 6 days ago #187221 by piranha32

BeagleBrainz wrote: What are the Amperage rating of the PSU for the Atomic Pi ?

3A. This is more than what I received in the "full devel kit" (2.5A)

Please Log in to join the conversation.

More
6 years 6 days ago #187222 by piranha32

BeagleBrainz wrote: I know from the redit forums that many people aren't using the recommended PSU for the Atomic Pi. The sellers recommend a 5v 4A supply. If the OP has a borderline PSU would this be a contributing factor to the errors he is seeing ? He does mention both the Atomic Pi & 7i95 are powered by the same supply.
Or would it be that the board is a little underpowered CPU wise ?

The PSU is more powerful than what I received in the AtomicPi devel kit. It is a bit strange, because everything worked fine under load, with all motors and peripherals running
I'll power the board from a separate PS, and see what will happen.

Please Log in to join the conversation.

More
6 years 6 days ago #187226 by piranha32

PCW wrote: Can you run this command:
halcmd show all *.tmax
with LinuxCNC running and post the results along with the CPU speed?


CPU runs at 1.68GHz. Motion runs without problems, system doesn't feel underpowered.
Here is the output from "halcmd show all *.tmax"
Code:
$ ~/build/linuxcnc-dev/scripts/rip-environment halcmd show all *.tmax Loaded HAL Components: ID Type Name PID State Component Pins: Owner Type Dir Value Name Pin Aliases: Alias Original Name Signals: Type Value Name (linked to) Parameters: Owner Type Dir Value Name 57 s32 RW 23940 comp_x.tmax 57 s32 RW 21528 comp_y.tmax 57 s32 RW 23364 comp_z.tmax 23 s32 RW 25236 ddt_x.tmax 23 s32 RW 21852 ddt_xv.tmax 23 s32 RW 28818 ddt_y.tmax 23 s32 RW 23076 ddt_yv.tmax 23 s32 RW 24354 ddt_z.tmax 23 s32 RW 26928 ddt_zv.tmax 65 s32 RW 0 hm2_7i95.0.read-request.tmax 65 s32 RW 1216764 hm2_7i95.0.read.tmax 65 s32 RW 170118 hm2_7i95.0.write.tmax 34 s32 RW 36342 limit_speed.tmax 19 s32 RW 26946 motion-command-handler.tmax 19 s32 RW 160434 motion-controller.tmax 40 s32 RW 23040 near_speed.tmax 54 s32 RW 24426 or2_0.tmax 43 s32 RW 25740 rpm_rps.tmax 20 s32 RW 1265814 servo-thread.tmax 31 s32 RW 34200 sim_spindle.tmax 37 s32 RW 27306 spindle_mass.tmax 26 s32 RW 48546 vel_xy.tmax 26 s32 RW 24624 vel_xyz.tmax Parameter Aliases: Alias Original Name Exported Functions: Owner CodeAddr Arg FP Users Name Realtime Threads: Period FP Name ( Time, Max-Time )

Please Log in to join the conversation.

More
6 years 6 days ago #187227 by piranha32

PCW wrote: It could be power supply related. It could also be
power management that has not been disabled,
or the CPU is just slow (a 2 ms Servo thread might solve the issue)


Installing OS from the live LinuxCNC image I did not expect any power saving mechanisms, and cpufreqd is not even installed, but I'll try to force performance governor and see if it will help.

Please Log in to join the conversation.

  • BeagleBrainz
  • New Member
  • New Member
More
6 years 6 days ago #187230 by BeagleBrainz
Replied by BeagleBrainz on topic Debugging sporadic read errors from Mesa 7i95
I've always used a min of 4amp supply for the Atomic Pi. Many of the brick type PSUs hve what may be called "exaggerated" specs, so I've always tended to go a little higher when using them.

Please Log in to join the conversation.

Time to create page: 0.056 seconds
Powered by Kunena Forum