Does PncConf load the firmware on a mesa 7i96s?

When I try to test an axis on this 7i96s, the red FPGA/INIT light comes on.   Do I need to run mesaflash to install the firmware, or does PncConf do it?

Where can I find the firmware and explanations of what their features?

The installed firmware is appropriate for most uses
unless you have an added parallel daughtercard.
So you don't need mesaflash.

A red init light indicates a watchdog bite

Are you running LinuxCNC?

I'm running the "Test/Tune Axis" routine right after you set the Stepper Scale.

And PncConf seems to have 3 firmwares to choose from.
7i96s_7i74 - I just assumed this one was for a setup with a daughtercard, which I don't have.
7i96s_d
7i96s_dpl

Is there a difference between the second two? I can't find any documentation.

The dpl configuration is for Plasma systems and modifies the
encoder input to include 3 "A" only counter inputs.

The _7i74 configuration includes support for 8 additional
SSerial channels on an added 7I74 daughtercard.

If you look at the firmware distribution archive, in addition to the
xxxx.bin firmware files, there are xxxx.pin text files that show the
pinout of the associated firmware files.

If you get a red light during test/tune, it's likely you have a real time issue
on your host. In this case its better to run LinuxCNC as it has better diagnostics than pncconf.

Also a simple ping test would give an idea why you are getting a watchdog
bite:

(assuming the 7I96S card is set to 10.10.10.10)

sudo chrt 99 ping -i .001 -q 10.10.10.10

(run for a minute or so and hit control C to stop and look at stats)

Well, that's better. I had to do a search, but found out that you have to go back to Mesa's sight where I bought the card and download the "Support Software". It'll be a zipped file that contains a directory of firmware builds.

Back out to the workshop. Run the ping test and then load LinuxCNC to see what we can see.

I think I'm making progress. If I got the motor to move before, it was just a violent shaking in place. Now, while I have the violent shaking, I can get the shaft to move in the desired direction. I think what did it was setting the Step On-Time, Step Space, Direction Hold and Direction Setup to their max values.

I ran the ping command, and got min/avg/max/stdev of .141/.208/2.270/.030

And I did get an error in LinuxCNC.
hm2/hm2_7i96s.0:error finishing read!

Would that be a communication error over the ethernet? I put together my own 2 inch ethernet cable. So maybe it is suspect. Anyway, I think I have two distinct issues at this point. A suspect ethernet cable, and a timing issue with the CL86T driver.

Have you run a latency test on the machine?

say:
latency-histogram --nobase --sbinsize 1000

2.27 ms max Ethernet ping time is pretty bad.
The "error finishing read" happens when too many
(default 10 in a row) packets are timed out, meaning the return
packet from a read -request took longer than 80% of the
servo thread period.

Oscillations sound like too high a P term for the servo thread period
What is your servo thread period and what is the P term set to?

P should be 1/servo thread period so 1000 for a 1ms (1000000 ns)
thread period

Yeah. I think it's a bad ethernet connection. I'm going to work on that.

But, the P term? is that Actual Servo Period set on the Base Information page of PncConf? I have it set to the default 1,000,000 (commas added here for clarity)


Should I increase or decrease it? Does it matter that these are closed-loop steppers?

Oh, and BTW, latency for the OPi5 I'm using has tested extremely low.   

Have you run the latency test?

That would verify that there is not a problem in your particular installation

The 2.27 ms is a killer

Here what I get on my HP Elite 800G1:

--- 10.10.10.10 ping statistics ---
207403 packets transmitted, 207403 received, 0% packet loss, time 207402ms
rtt min/avg/max/mdev = 0.066/0.104/0.143/0.001 ms

(Linux 6.4)

Wait. I found the P term. It is on the motor configuration page. Still set to the default 1000.