LinuxCNC-RIO - RealtimeIO for LinuxCNC based on FPGA (ICE40 / ECP5)

Strange things those control boards, tell ya. Interfacing digital is no problem for me, but I'm having a really hard time understanding why most folks use optos yet share the same ground as the motion board. No use imho...

I'll look into this external/digital isolation when my head functions again (happens once or twice max per month that lol). Probably going to need that  i s o l a t i o n 'as per the legend' , just dunno for what exactly.

Again, sorry for the hijack. I'm out. I'll update on progress ad schematics though. ~hacktheworld!

RIO related!!

Found a bug. They're really hard to find so I'm pretty happy about that

Should be "Angular", not rotary, in the axis joint/stepper plugin.

Enjoy your day, Olivier. All the best man.

Mecanix wrote:

Strange things those control boards, tell ya. Interfacing digital is no problem for me, but I'm having a really hard time understanding why most folks use optos yet share the same ground as the motion board. No use imho...

I'll look into this external/digital isolation when my head functions again (happens once or twice max per month that lol). Probably going to need that  i s o l a t i o n 'as per the legend' , just dunno for what exactly.

I feel guilty, I think I had separated GND for the tangoboard, but not for the others.
But if there's only one voltage source, it's also difficult.

Optos give you a better feeling

Mecanix wrote:

Again, sorry for the hijack. I'm out. I'll update on progress ad schematics though. ~hacktheworld!

has everything to do with rio, all good !

Should be "Angular", not rotary, in the axis joint/stepper plugin.

is fixed, many thanks !!!


all the best for you too

Crazy stuff. Linuxcnc's next level? AXIS Protocol... hardcoded lol


 

?????
This is also only UDP via an RMII device and without a soft core

we have everything, but not yet for all FPGAs and toolchains:
github.com/multigcs/riocore/tree/dev/riocore/plugins/rmii
github.com/multigcs/riocore/tree/dev/riocore/plugins/arty_mii

How did I missed that out. omg

I would like to bypass the UDP and IP layers, similar to Ethercat with its own lean protocol.
but then only a direct connection is possible, but a very fast one with minimal delay

Also possible with W5500 on Mac level

github.com/njh/W5500MacRaw/blob/main/sendeth/sendeth.c

Optos driving optos, people copying other designs with no real knowledge. Or “it’s what someone on cnczone recommended”.
re UDP, it’s simple to setup, proven by years of implementation and works.
I think what would be great, is that the FPGA reports it’s capabilities to the driver so the driver doesn’t require rebuilding. Tho it would make things way more complicated to implement.
Regarding isolated drivers that topic did come up a little while ago maybe 12 ish months, so it is something that has discussed.
Oh yeah common grounds with optos, that’s something I just can’t fathom. Maybe there is a reason in some applications, but for the most common application, providing isolation, I don’t get it.

These could be an idea for output
www.icbase.com/File/PDF/ONS/ONS56621412.pdf

meister wrote:

Also possible with W5500 on Mac level
 

No Thank You Very Much. Interesting none the less. 

What we currently have is STUPID FAST and high performance already. Unless a mysterious someone has a requirement for an hypersonic motion machine-tool, nobody here (in our star system anyway) will be able to ceiling the perf of RIO as-is on any parts/processes level. 

Not sure what you did but after updating the git today I'm clocking this x2 speed of what it was a few days ago. Used to be 2.5ms, now down to little more than a millisecond. Impressive stuff, Olivier.