Remora (SKR 1.4 RPi) – massive SPI noise, “bad payload” communication fails

Hi all,I have a working Remora configuration running with Raspberry Pi (LinuxCNC 2.9), SKR 1.4 board (LPC1768), and SPI connection – but only using an old Remora component, because the recent ones do not support LPC1768 anymore (as Scott confirmed). The system works perfectly “on the bench” with just the boards and motors, but as soon as I connect it to my real machine, the SPI link becomes super unstable (“bad SPI payload / estop won’t reset” errors).Symptoms:

• Dupont wires or short IDC ribbon cable used for SPI (RPi CM4 → SKR1.4, <30cm).
• Works fine with just the controller and stepper drivers.
• The moment I connect “real loads” (motors, hotwire, power MOSFETs, or even touch the machine frame), SPI constantly fails.
• E-stop triggers, payload errors jump in LinuxCNC log.
• Even switching on the heater causes instant SPI dropouts.
• With ribbon cable it’s even worse (won’t leave estop).

What I tried:

• Multiple cables, power supplies, SD cards, re-flashing, common ground.
• SPI config is what works for old component – I can’t change SPI freq etc, I get “unknown command” if I try.
• System is totally stable on the bench, not with machine.

Limitations:

• New Remora components don't work – the SKR1.4/LPC1768 is legacy, so I must use the oldest one.
• This old Remora doesn’t let me change SPI speed, delay, etc.
• Hardware fix suggestions welcome, but software/firmware is stuck “as is” for now.

Has anyone solved this?

• Is it possible to get reliable SPI with legacy SKR1.4 + RPi, in a noisy CNC environment?
• Any proven hardware methods? (twisted pairs, shielded cable, ground, ferrites, extra pullups, RC on SPI lines…?)
• Is it possible to patch the Remora legacy component to allow lower SPI clock?
• Any old firmware/component version which allows SPI configuration and works with LPC1768?

Any advice would help – right now the setup works great “bare”, but physical machine (motors/heaters/vibrations) kills SPI instantly…Thanks!

I know that with the mesa 7c81 boards it is recommended to have a length of less than ~2.5" for the SPI cable length. So it just maybe the cable length is too long.
Dupont cables aren't ideal for SPI transmission at speed.
So ideas to try.
Ribbon cable with every second conductor a ground terminated at one end only, of course you will need a dedicated return ground. This may involved a custom bit of PCB at each end.
Shorter cable length for sure.
There's some other fixes I was going to suggest, but since the remora bards are "all in one" they would not apply.
But I guess a quality PSU, think Meanwell, might be worth considering.
My first port of call would be to reduce the SPI cable to the absolute minimum, preferably to under 2.5" max.
One thing of note is that I have the RPi5 can tolerate a longer SPI cable that the RPi4, the caveat being this is based on experience with a home made 7c81 clone.
I'm apologise if the references to the 7c81 is annoying, but it is entirely based on my experience with SPI as a communication medium. I honestly hope that some of the above suggestions will give you good for thought.

Ribbon cable with every second conductor a ground terminated at one end only, of course you will need a dedicated return

I would suggest the same with the grounds connected at both ends.

Ground bumping and  ground impedance for external HF currents from external noise sources
is a real issue for high speed single ended signaling, so the more ground connections between
the RPI and the UC, the better...