Dead in the water on 7i92 config

How do I plot following error?

With halscope, plotting axis.N.f-error (N being the axis of interest)

Well, I was searching through the docs on .ini again-

I used the values in the 7i92step.zip file given earlier:
FERROR = .0002
MIN_FERROR = .0001

This is substantially different than the linuxcnc ini doc defaults:
MIN_FERROR = 0.010
FERROR = 1.0

I don't really understand what the meaning of what the number is. Yes I have the manual at linuxcnc.org/docs/2.6/html/config/ini_config.html , but I don't get what an "error" is without encoder feedback, or a perspective on what number is appropriate.

Well setting FERROR=1 & MIN_FERROR=0.010 made the Following Errors disappear. But I don't know if that's a fix or just a hack masking something else?

You can certainly set the following error limits so wide that it's impossible to ever get a following error.
This will definitely mask many types of setup issues

You can also set the limits so tight that you get following errors caused by
real time jitter in the hosts position sampling time. (this is much improved if the DPLL is used)

The ideal values are ones that are small enough to detect errors that compromise your machine tolerance
but large enough that you do not get nuisance following error trips

Note that the limits in the example file should be OK, here is a plot of X following error with a fairly well tuned system:

freeby.mesanet.com/1khzstepgen.png

Note the the scale for the following error is 5 uinch/division so peak following error is about 2 uinch
this is a 480 IPM 80 IPS^ setup with a 1 KHz servo thread

This is why I suggested plotting the error, and posting your hal/ini files so I or others on the forum can possibly spot setup issues.

Well, here it is- FERROR peaks at about 0.07.

Axis 3 is sync'ed to Axis 0 (X)

I was still getting Following Error on Joint 0 sometimes when switching from Jog Mode to MDI. When it did that, it could NOT be reset. Toggle EStop/Toggle Machine Power would only repeat the Following Error and no motion as soon as a move was commanded. I had to quit LinuxCNC and restart it.

Can't really tell but sort of looks like a real time error or Ethernet hardware problem to me

Can you tell me what OS (the result of uname -a), Ethernet hardware
and LinuxCNC version you are running?

Have you run a latency-test or latency-histogram? what were the results?

Ah it's looking pretty large, in the 14,000,000 territory. I also went through and found that it's an Intel 82801 chipset- which is known to cause problems because of its system interrupts.

Are you sure you are running a Preempt-RT kernel?
Frequent multi-ms latencies sounds more like a non-RTOS kernel to me

uname -a

linux localhost 3.4-9-rtai-686-pae #1 smp PREEMPT Debian 3.4.55-4 linuxcnc i686 Gnu/Linux