Communication solution on my own CNC Machine

You'll need to run a realtime OS your boards, as you are essentially using "software step" generation at the end point. Which is no different from using "software step" generation on the Linuxcnc machine. The operation system will add some latency, which when viewed via a scope will tend not to be a stable (regular) pulse train. Latency isn't really about "time or out right speed" but the ability to do things at regular intervals without a lot of variation.

There are some projects on the forum that use STM microntrollers, I suggest you check them out.

Cant do this anymore bye all wrote:

You'll need to run a realtime OS your boards, as you are essentially using "software step" generation at the end point. Which is no different from using "software step" generation on the Linuxcnc machine. The operation system will add some latency, which when viewed via a scope will tend not to be a stable (regular) pulse train. Latency isn't really about "time or out right speed" but the ability to do things at regular intervals without a lot of variation.

There are some projects on the forum that use STM microntrollers, I suggest you check them out.
 

Can you give examples of those projects in the forum? What kind of real-time communication is there between Linuxcnc and stm cards? Ethernet?

Cant do this anymore bye all wrote:

It's 512Mbit = 64MByte.

That is miserable !!!
In that case, FPGA's are much better. Or even micro controllers.

 
Do you think that 64MB RAM will not be enough to open the Ethernet packets coming from the main Linux computer and send them to the servo drivers via IOs? Or are you hesitant about real-time communication?

Is it possible to build this structure in Linuxcnc? Maybe it is cumbersome for axis management. But it will be a very practical solution for hand wheel or ATC management with ethernet interface.

Great writeup. Lots of enthusiasm and anticipation, good start!

To my knowledge anything can be done in the automation sphere, including making your own proprietary motion controller(s) & HMI. Many have done it for their robots and other machine-tools. It's no secret to anyone that much of the modules required to get you started are already-done by those gate array manufacturers and serviced to their developers via "IP cores". (stepgens, pulse waves, encode/decode, communication protocols, filters, mipi/dsi, sram, dma, everything's there). Have a look at the ip cores and hdl example sources from Lattice's Radiant, AMD's Vivado, Intel's Quartus, Gowinsemi Designer, etc etc... option list is endless if you add in the second tier manufacturers of those semiconductors...

And so can a mesa-like solution be achieved? of course it can. Question is, though; is your goal to be productive, as in making parts? or be a night-time developer who waste time reinventing what already works? I.e MESA, Litexcnc, RIO, Remora... to name a few

(Side note) I've always been under the assumption that those raspi were engineered to play video games on a 7" lcd and for making youtube unboxing vids (aka marketing toy). Not particularly aimed at the automation signal gen & processing market, at least not at the data integrity level one you require and deserve. That being said, you might want to re-think that strategy!

happy-happy chip makin'