I also ordered some Ethercat boards to play around with it, for now I just use regular Ethernet. I will suspend working on it until after chinese new year since I need connectors.
I also hope I can get hands on more Mitsubishi/possibly other machinery in Taiwan.

Chaining up multiple boards with Ethercat seems to be interesting, especially for the peripheral part.

The Zynq is definitely super powerful, I was able to get away with a microcontroller and accurate timings.
But the MDS-B series only needs 3.55ms framing, not 0.222, jitter is around 20ns (while the measured jitter on the Meldas M65 controller is 40-60 ns .. of course all relatively due to my non certified measurement equipment).
My plan is to make it pretty much modular with Ethercat chaining up various modules for certain tasks.
I want to mix my system with different motors than Mitsubishi for 5 axis support.

I have an old DE2 FPGA board (with ethernet), and a couple of LatticeSemi LFE3/5 boards (also with Ethernet) here which I almost wanted to use for it. I think I will just stick with various microcontrollers and small FPGAs if needed to keep it simple in the future.

Finally, got to test drive the pullmax_optima component with closed loop PID, only p set to 5. Left and right side sync nicely, but with loud groaning noise. This machine runs smooth and quiet when I run it open loop. I think what I'm seeing in halscope is a combination of untuned PID and non-linear hydraulics.

chan. 1 pwmgen.04.value = controls left ram servo valve (driven by j3.vel_pid.output)
chan. 2 pwmgen.05.value = controls right ram servo valve (driven by j4.vel_pid.output)
chan. 3 pwmgen.06.value = controls tonnage proportional valve; ultimately comes from a slider on my test gui.
chan. 4,5,6,9 are the discrete spool valve coil power outputs (cr-200 to cr-203)
chan. 7 j3-motor-pos-cmd ultimately comes from a slider on my test gui. This is the input to the pos_pid.command
chan. 8 j3-motor-vel-cmd is signal connecting pos_pid.output => vel_pid.command

I was hoping velocity PID will allow me to skip the low frequency dither these valves like; but now I'm thinking my lack of that dither is what is fighting my PID. I will try lowering the p value first in case 5 is actually too high, next step is adding that low frequency dither, probably via siggen or a custom component.

It should be possible with hm2_modbus,
(as long as the devices sharing a link have the same baud rate)
mesa-modbus only supports one device (per host UART)
 

No, if you set all PID values to 0 in the .ini file, run linuxcnc
and set LinuxCNC into the machine-on state, all PWM values will be 0
and the PWM enables should all be true (at least for the joints enabled in the hal file)

“Also insure the all PWM enables are true, and all PWM values are 0”
Should I define this in HAL?

Great, that's got the sserial and pktuart ports appearing and working, thank you for the quick response.

We have 3x identical modbus cards to communicate with. Using mb2hal it was possible to address each one but I can't see how to do that using mesa_modbus - is there a way to talk to different addresses over a single pktuart port? This may be a question for Andy Pugh...

Is this possible in hm2-modbus?

OK that looks good.

The next test is to set all PID pins to 0 (P,I,D,FF0,FF1,FF2) in the ini file
and start LinuxCNC, then enable motion, at this point, PWM should be enabled on all axis
but the PWM value should be 0 on all axis. Again measure all the Analog outputs.

Also insure the all PWM enables are true, and all PWM values are 0

 (test done with nothing connected to the 7I97T except 5V power and Ethernet)
 

Have a look in syslog "sudo dmesg" and if you find the word EoE that is a problem.
Typically that interferes with the linuxcnc lcec component, which can not handle EoE.

There is a way to read out the eeprom content from the drive, flip a bit and write the eeprom content back.
However, it is much better to compile the ethercat master yourself with eoe disabled.
It has been recommended before in the forum, please search (I don't have it readily available).

See how far that takes you. 

Axis0 11mv
Axis1 9mv
Axis2 10mv
Axis3 9mv
Axis4 10mv
Axis5 9mv

The 6 axis ranges from 9-11mv

Failed communication connections are a software issue.


If you power the 7I97T (and not run LinuxCNC), do you read
near 0V on all analog outputs? (that is, less than 15 mV)