Another 7i96 Error finishing read

Hello!

I've had a plasma cutter set up and running for over a year now. It recently started throwing the finishing read/write errors. I did just do a bit of a rebuild of motor/sensor wiring, but everything was external to the control box. I'm able to replicate the read/write errors with none of the motors plugged in, so I'm assuming this is unrelated to my recent change, and just an unfortunate coincidence.

I've searched around and seen several instances of these errors showing up, but I can't seem to find a solution. From my understanding the other instances were when a new computer was being used. Nothing on my setup has changed in terms of the computer and control board, so not sure why it started throwing the errors. I've attached a screenshot of the errors for reference.

Another difference with my case seems to be that the errors don't always come up right away. It'll throw the errors anywhere from immediately upon opening linuxcnc to sometimes several minutes into a job.

I've attached a screenshot of the output for `halcmd show param *.tmax` which shows 4,571,563 for hm2_7i96.0.read.tmax.

Here's what I've tried so far, with no luck: 

Ethernet cable

• Tried with a different ethernet cable.
  • Tested the ethernet cable with a cable tester.
    
    Servo Period
    I changed the servo period to 2,000,000 and P value to 500 as recommended in several threads.
    
    Power Supply
    I saw mention that this error could appear with an inadequate 5v supply. I have the PC and 7i96 in a large PC case, so I'm using the 5v from the PC power supply. While debugging I noticed that the PC power cable wasn't making great electrical contact, which lead me to think the power supply might be the culprit.
    • Used a different power cable that had better electrical contact
      • Used a 5v supply from an iPad charger and cut up USB cable
        • Replaced computer power supply (still with iPad USB power to 7i96)
          • Used 5v from the replaced power supply (noted mention that USB cables might not be adequate)
            
            Configuring Intel Network Chip
            The computer being used is an older intel. I don't know the specs, but the motherboard has an intel silkscreen. (I can get the specs if necessary). I can't find the thread anymore, but @PCW mentioned that intel chips need some "IRQ Coalesce" setting disabled. Tried adding the following lines to my `/etc/network/interface` as suggested in the docs, but I get a network error when booting up the computer (Failed to start Raise network interfaces.).

            auto eth1
            iface eth1 inet static
            address 192.168.1.1
            hardware-irq-coalesce-rx-usecs 0
             I think that's everything I've tried so far. Any help is greatly appreciated!
            
            Chad

There may be a typo, the text shows "hardware-irq-coalesce-rx-usecs 0" but the image frome the terminal shows "hardware-irl-coalesce-rx-usecs 0"

Good catch! That was my second time making that change, just so that I could get an image of the error. I went in and corrected the typo, but still get the same error on startup.

Looks like your Ethernet name is enp2s0, not eth1

Also, you can verify if the coalesce settings took (or are even applicable)
with:

ethtool -c enp2s0

(from the command line)

Looks like your Ethernet name is enp2s0, not eth1

Also, you can verify if the coalesce settings took (or are even applicable)
with:

ethtool -c enp2s0

(from the command line)

 Changing eth1 to enp2s0 seemed to have an affect. My `/etc/network/interfaces` now looks as follows:
 

source /etc/network/interfaces.d/*

# The loopback network interface
auto lo
iface lo inet loopback

# The primary network interface
allow-hotplug enp2s0
auto enp2s0
iface enp2s0 inet static
address 192.168.1.1
hardware-irq-coalesce-rx-usecs 0

I tried running `ethtool -c enp2s0` but apparently `ethtool` isn't installed on my machine.

I'm unable to replicate the issue after making this change, hurray! If I change my network config back to what it was before (pre disabling irq coalesce), I can replicate the issue within a few minutes. So it appears this has resolved the issue.

Since this is an intermittent issue, and because the shop this is normally setup in is an hour away from home, I'd like to do some further digging if possible before sending it back to the shop. Especially considered this error just started happening after a year or more of it running fine.

After linux cnc has run for a few minutes, the read.tmax is `714326`. Is this high? I ran a few commands to get the specs for the machine, and it's an Intel Core 2 Duo @ 2.80 GHz.

I've attached a couple more screenshots. Any further help to debug/double check that this has been resolved would be greatly appreciated!

Chad

714326 (about 250 usec ) is a lot better than the
4571563 ( about 1600 usec) that is was previously...

@PCW definitely! Do you think it's of concern though? Seems odd to me that it just randomly started throwing errors after a year or more of no issues.

PCW could you explain how to calculate the period based on TMAX? i think 1 divided by cpu clock multiplied by tmax. is that so? i have an i5 2.3ghz tmax 149000 cutter and an i7 4 ghz 179000 plasma both with 5i23 one with 7i40 and one with 7i30 closed loop. they work very well it was just to clarify how it is calculated... I think it can be of use to others.

 

On x86, the time is TMAX/CPU _CLOCK, on ARM, it is in nS

As PCW said above:
tmax in ms = TMAX / TSC_kHz
Where tsc_kHz can be retrieved with