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

Cant do this anymore bye all wrote:

Finally got it running without latency errors when coming out of estop. Finally get to use the RPi5.
Tho I can not rule out my messy SPI wiring from the FPGA to the WT32-ETH01.

Still have to have a try with a W5500 module.

Not sure how this effects RIO, but with the remora w5500 one, the wires were crucial to be as short as possible, but also something ground loop was important even with short wires. jiggling the cables effected it a lot more than I expected.

When I was working with the Mesa firmware when using SPI wiring from the Pi to the dev board needed to be good. On the design of the board for SPI interface I designed to have a gnd between every signal. Since I’m using a different dev board I’m being a bit, okay really, lazy.
Considering a slight redesign of the board that brings out 3.3v and 5v on the connector that has the spi connector. That of course will be future Rob’s decision.

This is project is awesome, I just ordered some WT32-ETH01's, keep up the good work!

A big thank you from me!

have add a little plugin-overview:

github.com/multigcs/riocore/blob/main/PLUGINS.md

It’s a shame that the various Ethernet modules about that use the same PHY as the arty, DP84xxx, only bring out one tx & one rx signal.
I’ve been slowly looking the W5500 code to see if I could see anything that jumps out that points to connection loss.
I’ve tried using a simple board similar to I’ve used when experimenting with a 7c81 copy, no dupont jumpers acting like antennas and reducing the divider to reduce the spi clock speed all to no avail. I’ve got 2 of the waveshare type boards and one of the Wiznet ones that are basically just the size of the RJ45 connector. Same results with both.
I don’t know why I’m so intrigued in getting it working, the UDP to SPI bridge works fine with my RPi 5 but not so with the RPi4.

the only thing i could find is that the error rate depends on the SPI speed, maybe try using 'pullup' resistors to 'stretch/stabilize' the signal lines