Replacing SWI TRAK CNC 2 Controller

Hey guys, my Dad has a 90s era knee mill with an SWI TRAK CNC 2 controller that I want to replace.
Everything works fine, but the controller is extremely limited.
This machine is already a retrofit since it was completely manual at first, then a SWI kit was installed aftermarket. All of the handles are intact so it can be used as a manual machine.
It is currently only used as a manual machine with DRO because dad can't remember how to program the controller, and frankly it is an extremely frustrating and limiting process by modern standards.
There is no spindle speed control, tool changer, coolant pump, or automatic oiler. I'll probably end up adding a coolant pump and/or air cooling because I think dad's system of standing there with a squirt bottle is a bit ridiculous.
My intent is to make a complete replacement for the TRAK CNC 2. I want to be able to unplug all the cables, unmount it from the arm, and attach my new controller in it's place.
This will only be used for hobbyist level projects and gaining experience machining with CAM. I doubt it could hold the extremely close tolerances required for job shops, and lack of features would make complex jobs uneconomical.

I've scoped out the signals from the controller, and it uses locked-antiphase differential PWM at 2.5V and 15.5kHz referenced to machine ground. If you look at the servo drive inputs there are 5 terminals: two are jumped with a wire (since there are no limit switches...), one is left disconnected, and the other two are the differential PWM signals.
The box on the side of the machine holds the servo drives and AC transformer. It looks just like pictured here .
There are encoders on the ends of the X and Y servos, as well as mounted on the axis where a wheel rolls on a bar mounted to the table or saddle. There is only one encoder for Z and is not on the servo.
On the encoder connectors there are 9 terminals, 4 are unique to each, 3 connect to chassis ground, and the other two are common so probably supply voltage.
I believe these encoders are non-indexed dual channel (A A- B B-) 2000 quadrature with 500 count per revolution. Same as these .
Currently the E-Stop cuts AC power to the Transformer/Servo drives as well as a circuit in the controller. I'll probably use the same setup unless someone has an argument against it.

I'm hoping to use a Raspi 4 with it's own heatsinks and fan inside a large sealed aluminum enclosure that also houses the MESA cards and user interface. I'm more concerned about dust getting into the enclosure than heat at this point.
PCW suggested using a 7i85s in another thread for someone with similar servo drives, but it only has 4 encoder inputs while I have 5 encoders. I could just not hook up the X and Y motor encoders, but I was hoping to use the 2 encoder per axis technique here.
So instead of a 7i85s I'm thinking about using a 7i52s, which looks very similar in specs to the 7i85s except it can handle 6 encoders/drives instead of 4 and uses a 50pin connection instead of 25pin.
Since the 7i52s needs a 50pin connection I'm looking at the 7i93 to interface via ethernet to the Raspi. I like the idea of ethernet because I could connect it to a PC instead for whatever reason. 7C81 looks really nice for a Raspi, but then I would be locked to using a Raspi and could not use 50pin daughter cards.
Currently AC power comes into the pendant, through a toggle switch, then through a fuse, then through the E-Stop switch, then back out of the pendant to the box with transformer and servo drives.
There is an EMI filter and a separate fuse between the toggle switch and controller.
My Current plan is to use a Schaffner FN9264B-10-06 which is a combination IEC power entry, power switch, and medical grade EMI filter. Would this be a reasonable replacement, or would a separate EMI filter for just the controller be more desirable? My concern here is if the transformer or servo drives introduce their own EMI onto the line that would be after the filter and not isolated from the controller.
I also plan to use a Meanwell LRS-? power supply that has built in short circuit / overload / over voltage protection, so I'm thinking a fuse is not really needed.

I already have a 10" HDMI touchscreen ready to mount, but I want the most used buttons to be actual buttons rather than touch-only.
My list includes cycle start, feed hold, and of course an E-Stop. What kind of control would be good for coolant? On/Off? Selector switch? Both?
Should I include a keypad of some kind? I'm thinking it would be a number pad with letters G M X Y and Z. I'll also have a mini wireless keyboard stashed away if I really need it.
I plan to have a 3-position selector to switch between Manual/DRO only, MDI for simple routines, and Auto for G-Code generated on a PC. This would also switch which interface is on the touchscreen and in the case of manual mode, completely turn off the transformer and servo drives.
I also want an actual knob for feed rate override. I'm leaning toward using a small optical encoder connected directly to the Raspi since it isn't timing-sensitive, and the Raspi has no analog input for a potentiometer.
Should there be a rapid override? I can see why it would be useful if the operator and programmer are different people, but I can't see a case where I would need it personally.
I don't see jogging controls as a necessity since I can just grab the handle to move an axis. Thoughts?
There is currently no feedback mechanism for spindle RPM, and I'm undecided on if it is needed or not.

My apologies for the giant forum post, and thanks in advance for any feedback.
@techrat @huntert98 @vsidon Have any configuration files that you can share?

Sounds like fun.

"I also plan to use a Meanwell LRS-? power supply that has built in short circuit / overload / over voltage protection, so I'm thinking a fuse is not really needed."

Just a heads up, we originally went with a hefty Meanwell triple voltage +5, +12 -12VDC PS that had to have a certain amount of load to make it function. Which we didn't have.

The "fix" was to put a dummy load resistor on the outputs to draw enough of a load to get it to operate properly. Nope, not doing that. So back it went for a different brand.

Nothing against Meanwell either, we have a Meanwell for the +15, -15 VDC part of the system, and it works great. No "pre-load" issues at all on this one.

Thanks for the warning, looks like they have done some updates so there is no longer a minimum load requirement. I'll just have to be careful when ordering that I'm getting the new revision.
I had a ? because I'm not certain what my load will be until everything else is locked down in the design.

It sounds like Meanwell is listening. That's a good thing. They've always have been a favorite of mine.

And remember, it's only "Meanwell" as long as they work, it's "Meantwell" once they die.

(I'll be here all week, be sure to tip your waitress.)

Zenduin,

I'm working on this same thing - happy to see you've somewhat trod before me. Any updates on this project?

I am looking at using a 6i24 + 7i52s ... and as I'm scoping the signal wires I'm finding what you're seeing, but the peak-peak voltage on the PWM signals is about 3.3V. Are you hooking the 7i52s directly to these lines, or are you stepping down the voltage at all with any sort of level shifter?

This link might be helpfulhttps://forum.linuxcnc.org/30-cnc-machines/29358-proto-trak-mx3-retrofit-which-boards-mesa?start=0