To use 24V fans, you should left V_FAN and VIN shorted with jumper, as photo.

@Danal, thanks for your answer, it's almost full. Interesting about belts. The same technique for boards - use star topology, don't route wide loops for single ended signals (ground is a return path for single ended signals), be careful with long high impedance lines and so on. This is EMC and it's cool.
I want not only to eliminate the problem, but to find the reason, the most EMC-sensitive line.
So, my first question from previous post is still actual. I'll try some solutions and post the result here. Maybe my 5 cents will be useful for colleagues, maybe not, but the result is interesting.

@dc42 thanks for your answer.
Are layer comments the same for every slicer?
Could you answer two more questions, please:

1. How DWC determines layer change?
2. If I'll place any G-command (one or block of commands) with summary Z, DWC will take it, divide per layer height, then show me correct info? You wrote, that it looks for Z moves , so, it looks like "last Z" minus "pre-last Z" = "layer height" for DWC, then "last Z" / "layer height" = "layer quantity"?

Hi everyone!
No ideas about this? Maybe mr. @dc42 could give any comments?
Mostly, I'm interested in how DWC determines layer quantity.

Dear community and administration!

I'm using only Cura slicer and there is a bug, when selected "Special Modes" / "Print Sequence" = "Once at a Time".
Two screenshots: DWC and Cura.
You can see 5 details, in first will be printed bottom row from right to left (the same details with max height eq. 5 mm, 25 layers), then (last) upper frame (height is 3 mm, 15 layers).

DWC reports me, that there is only 15 layers and 3mm height, as for last detail. So, while printing, I've got "Layer: 25 of 15" (because first detail has 25 layers) and last value of "Current Layer Time" will count up to the end of file (real last layer).

I've tried to edit g-code file (edit first ";LAYER_COUNT:" and delete others, renumber all layers at ";LAYER:"), but no - looks like DWC determine number of layers in other way - change of Z value.

I think, that DWC g-code parser should enumerate not only positive Z change, but all Z changes. What do you think?

PS I've searched for the same problem here, but looks like nobody use "Once at a Time".

Hi everyone!
18kOhm pull-up resistor from W5500 nRST to Vdd and 10nF pull-down capacitor to GND - all the day without problems.

@Danal, thanks for your answer, it's almost full. Interesting about belts. The same technique for boards - use star topology, don't route wide loops for single ended signals (ground is a return path for single ended signals), be careful with long high impedance lines and so on. This is EMC and it's cool.
I want not only to eliminate the problem, but to find the reason, the most EMC-sensitive line.
So, my first question from previous post is still actual. I'll try some solutions and post the result here. Maybe my 5 cents will be useful for colleagues, maybe not, but the result is interesting.

@DocTrucker, I've decided to make some tests before connecting power supply minus with mains earth and chassis.

1. When web interface has fallen, I've restarted it with M552 S0 => M552 S1 (by USB-console), and it went back. Did SPI had restarted with this sequence and W5500 nRST had triggered? I think that electrostatic discharge affects nRST because trace to the SAM4 is very long, and it uses an internal uC pull-up resistor. This is not always good for EMC, IMHO. Thinking about placing 10k resistor from nRST to +3.3V with a small cap to the GND.
   What if web interface will fall during print process and I'll send M552 S0 => M552 S1? Will it restart web immediately?
2. There is C_GND through-hole, that is connected with GND using parallel RC. Thought about to connect C_GND with chassis.

Hi, friends!
I'm using Duet 2 Ethernet (main board v.1.04, ethernet board v.1.05) for about one year. It connects directly to the secondary LAN of my PC with short cable, static IP for both PC and board. Board is fixed to the plexiglass cover of my printer, not shielded, not grounded. Power supply is original Meanwell 24V 480W - it's grounded with printer's chassis and well connected with mains earth.
Web control interface very often falls with "Connection Lost". Reloading page brings me to the connection timeout. Also I can't ping it (unreachable), LAN's LEDs - orange is always on, green is blinking when pinging it. Looks, like Wiznet chip is alive.
Several months ago, because of this problem, I've soldered Ethernet board to main with IDC connector pins, but problems stayed. Fall of web control doesn't affect printing process - printer continues to print, but the only way to bring interface back is to reset main board.
What I've noticed is that problem happens when I do something close to the printer. I don't touch it but static discharges may be a problem - now air is drier and looks like the problem appears more often.
Maybe solution is to ground main board, but there is no ground connector and connect minus to ground may be not the best idea. Your thoughts, please.

To use 24V fans, you should left V_FAN and VIN shorted with jumper, as photo.