All the pulley mounts are custom designed and printed for my sand table, and have gone through about 6 versions, and will probably go through a couple more before I’m done. There’s no reason you couldn’t make those parts from metal if you have access to a mill. The pulleys are also 3D printed and use stacked F625 bearings that are normally used in electric motors. Compared to 3D printer pulleys, they have bigger balls (size matters!) and should last a lot longer. They definitely run quieter. 3D printer pulleys with tiny bearings failed in about a year of occasional use in my table.
The belts are tensioned by pulling the motors in the t-slot until the belts are tight, then tightening the motor mounting bolts. In a corexy mechanism the belt tension is used to set the X and Y axes square to each other (though they should start that way before the belts get installed). Installing the first belt will pull the axes out of square, and installing the second belt will pull them back into square.
I have used microswitch endstops but now use optical endstops because they are silent. The Duet board’s motor drivers (if you use them) can theoretically detect motor stalls, so you can home that way, but it will probably be a little noisy when the magnet carriage and Y axis start banging into the physical ends of the axes. The servomotors have their own built-in drivers, so when connecting to the Duet board all you’re doing is taking the step/direction/enable signals from the Duet. You might be able to use the alarm output from the servomotor drivers to tell the Duet when the physical end of the axis is hit, but the motors are very strong and it is going to be noisy. You probably wouldn’t want to do that with the servomotors- if you manage to stall them mechanically they will probably cause the power supply(s) to shut down -if you’re lucky. If you’re not lucky they’ll break something mechanical in the process. A lot of people seem to want to build 3D printers without endstop switches for some reason. Most who try the endstopless homing go back to using endstop switches of one sort or another. They are easy to wire, very cheap, very reliable, and in a 3D printer where it matters, very precise.
I have scoured the web for every bit of info I can find on the servomotors and have been unable to come up with anything in English about tuning these specific motors. The English language manual and tuning software are all but useless. It looks like a couple guys in Germany may have figured them out, and I’ll be sending them some email about it, but the good news is that they work fine and quietly with the factory settings so there’s probably no need to tune them for a sand table.
That mechanism you linked will work, but it uses belts- they’re just hard to see. The problem with that type of design is that you have to run a cable to the motor that drives the magnet carriage, and that is on a moving part of the mechanism. That means a long drag chain has to be installed, and it usually means much lower reliability because the cables in the drag chain end up flexing too much and eventually break. The corexy mechanism is ideal for a sand table because both motors are stationary.
Wheeled carriages are always an option for the mechanism in a corexy machine, but they take up a lot of space. I don’t know if they’re quieter than bushings (like I use) when run fast, but at low speeds they are quiet. I don’t think rubber wheels would be a good idea- I think they’ll develop flat spots when the table sits unused and they’ll start making noise when you run the table.