UVW Axis prediction bug?

Kulitorum · 15 Jun 2022, 10:14

There is a bug!

Board: Duet 3 MB6HC (MB6HC)
DSF Version: 3.3-rc3
Firmware: RepRapFirmware for Duet 3 MB6HC 3.3RC3 (2021-05-26)

We are using the U axis to rotate the nozzle of our printer, and it seems quite clear that the firmware does not take the u-axis into account when calculating G1 moves.

For example, if they U axis is the slower for the move, the firmware will not properly predict the needed slowdown, and rather run the other axies as if the U would not have a problem, and only when we get to where the U axis cannot keep up, all the other axies stalls, and does very uneven moves.

This has always been the case.

Is the UVW axies implemented differently in the firmware? - Can you take a look at it, please?

Thank you.

Michael, COBOD

jay_s_uk · 15 Jun 2022, 11:10

@kulitorum please update to 3.4.1 and report whether this is still an issue.

JoergS5 · 15 Jun 2022, 11:29

@kulitorum by default UVW are handled as linear axes and ABCD as rotational. You can change it with the M584 R parameter.
https://duet3d.dozuki.com/Wiki/M584

Kulitorum · 15 Jun 2022, 11:35

Interesting development here: We renamed the Axis to A and the problem went away......

We are updating now.

dc42 · 15 Jun 2022, 16:31

@kulitorum the UVW axes are implemented just like XYZ. Their maximum speeds should be honoured at all times.

Also, because UVW default to being linear, the U motion will be considered when interpreting the F parameter, which is almost certainly not what you want to do. I suggest that you either create U as a rotational axis in the M584 command, or use axis letter A instead (as you are now trying).

Kulitorum · 15 Jun 2022, 16:59

@dc42 I know that's the theory. None the less, if we use the A axis everything is smooth, if we use the U axis it's not.

This is true in both cases if we use rotational or linear configuration. Using A is Ok, Using U is not.

PS: I'd still prefer to use U, so I don't have to re-explain this to all my customers and redo all the documentation, manuals etc.

fcwilt · 15 Jun 2022, 17:24

@kulitorum said in UVW Axis prediction bug?:

@dc42 I know that's the theory. None the less, if we use the A axis everything is smooth, if we use the U axis it's not.

This is true in both cases if we use rotational or linear configuration. Using A is Ok, Using U is not.

PS: I'd still prefer to use U, so I don't have to re-explain this to all my customers and redo all the documentation, manuals etc.

What do your config lines for that look like when trying to use U as rotational?

Thanks.

Frederick

dc42 · 15 Jun 2022, 19:03

@kulitorum said in UVW Axis prediction bug?:

This is true in both cases if we use rotational or linear configuration. Using A is Ok, Using U is not.

I find that hard to understand. The only difference between how axis letters U and A are treated in firmware is that U defaults to linear while A default to rotational.

You can run M584 without parameters to see how the axes are configured. In the response, rotational axes will have (r) after them.

I note that you are running firmware 3.3RC3. Please note, RC version of firmware are unsupported an may contain bugs that are fixed in later versions. Please consider upgrading to firmware 3.4.1.

Kulitorum · 16 Jun 2022, 07:18

@dc42 We are updating the firmware today, but this is not a new thing, it has been the case since forever. (4 years) - so I doubt updating will make any difference.

We spend 3 days trying to figure out why we have these uneven movements, and changing the axis "name" to A fixed it.

We will examine further, but it does feel like the Axies are somehow running different code.

Kulitorum · 16 Jun 2022, 16:10

@dc42 Hello again.

We setup a test where we ran the same gcode with the nozzle rotation configured for the U axis and for the A Axis.

Thinking that the shaking would be clear, I mounted a gopro on the printhead - looking down - and put some gloves on the floor for reference. The most obvious result is that the whole print happens a lot faster when the nozzle is running as the A-axis, but looking at the printer in real life, you also notice some shaking and hard accelerations on the U-axis version.

I have uploaded an (unlisted) video here: https://youtu.be/abVquTFmLH8

I will mail the config files and gcode files (only the first layer) to you.

My conclusion is that the axies cannot be running the same code, something is not right.

Please have a look.

Michael

Phaedrux · 16 Jun 2022, 16:52

Thanks for the video. It will be interesting to see the config files.

Kulitorum · 17 Jun 2022, 08:08

@phaedrux I mailed it to DC42