Rotary delta calibration not implemented?

Phaedrux

I'm not exactly sure how the -Z offset on tool2 is going to have an effect. I'm not even really sure what the reasoning or intension is there.

You're using an IR sensor though, so you should have an XY offset configured for it in G31

G31 X0 Y0 Z33.385 P25

And the Z trigger height should be calibrated.

https://duet3d.dozuki.com/Wiki/Test_and_calibrate_the_Z_probe

I would try simplifying your tool definitions to
; Tools
M563 P0 D0 H1 F0 ; define tool 0
G10 P0 X0 Y0 Z0 ; set tool 0 axis offsets
G10 P0 R0 S0 ; set initial tool 0 active and standby temperatures to 0C

And then select T0 at the end of config.g instead of T2

Then try G32 again.

Orang_

https://github.com/Duet3D/RepRapFirmware/blob/dev/src/Movement/Kinematics/RotaryDeltaKinematics.cpp

221 bool RotaryDeltaKinematics::DoAutoCalibration(size_t numFactors, const RandomProbePointSet& probePoints, const StringRef& reply)
222 {
223 	return true;	// auto calibration not implemented yet
224 }
225 
226 // Write the parameters that are set by auto calibration to a file, returning 227 true if success
228 bool RotaryDeltaKinematics::WriteCalibrationParameters(FileStore *f) const
229 {
230	return true;	// auto calibration not implemented yet
231 }

Orang_

I understand correctly that there is no calibration for rotary delta?

dc42

@Orang_ said in Rotary delta calibration not implemented?:

I understand correctly that there is no calibration for rotary delta?

That is correct. The problem is that there are potentially many calibration variables for a rotary delta. For example, there are two pivot locations (like the tower position corrections for a linear delta), three upper arm lengths, three lower arm lengths, three pivot height corrections, three homing position corrections, and the delta radius. That's 18 parameters. It's unlikely the by probing a flat bed and measuring the height errors, it would be possible to calibrate all of these. However, if some of these are known to be accurate (for example, if all the arms are manufactured to precise lengths) then it may be possible to auto-calibrate the remaining parameters.

I can try adding auto calibration when I have time, but it might not give very good results.

Orang_

@dc42 said in Rotary delta calibration not implemented?:

That is correct. The problem is that there are potentially many calibration variables for a rotary delta. For example, there are two pivot locations (like the tower position corrections for a linear delta), three upper arm lengths, three lower arm lengths, three pivot height corrections, three homing position corrections, and the delta radius. That's 18 parameters. It's unlikely the by probing a flat bed and measuring the height errors, it would be possible to calibrate all of these. However, if some of these are known to be accurate (for example, if all the arms are manufactured to precise lengths) then it may be possible to auto-calibrate the remaining parameters.
I can try adding auto calibration when I have time, but it might not give very good results.

Thank you, I really like your firmware, if there are any changes with the calibration, please let me know, I would be happy to test it. Also there is not a big group of people who own similar printers I think they will also support it.

jay_s_uk

I'll just make @gloomyandy aware of this thread

PCR

@Orang_ on another note. Is there a good project for building one? Saw this one this week: https://www.youtube.com/watch?v=Tj1KAbq99dc

Orang_

@PCR I have one of these https://www.youtube.com/watch?v=COUdIsoDgMw&ab_channel=AlexKorvinWorkshop

Orang_

@Orang_ https://www.thingiverse.com/thing:3489886

Orang_

@jay_s_uk but smoothie doesn't work quite right.

jay_s_uk

@Orang_ who said anything about smoothieware?

alankilian

@jay_s_uk I think he's saying smoothie (which runs on that thingiverse-link printer) doesn't work, so he's looking at Duet firmware.

Orang_

The project specified in the link works on smoothie. but I reconfigured it to rrf

jay_s_uk

@Orang_ ah, ok. no problem

dc42

@Orang_ said in Rotary delta calibration not implemented?:

@dc42 said in Rotary delta calibration not implemented?:

That is correct. The problem is that there are potentially many calibration variables for a rotary delta. For example, there are two pivot locations (like the tower position corrections for a linear delta), three upper arm lengths, three lower arm lengths, three pivot height corrections, three homing position corrections, and the delta radius. That's 18 parameters. It's unlikely the by probing a flat bed and measuring the height errors, it would be possible to calibrate all of these. However, if some of these are known to be accurate (for example, if all the arms are manufactured to precise lengths) then it may be possible to auto-calibrate the remaining parameters.
I can try adding auto calibration when I have time, but it might not give very good results.

Thank you, I really like your firmware, if there are any changes with the calibration, please let me know, I would be happy to test it. Also there is not a big group of people who own similar printers I think they will also support it.

Do you know which parameters of your machine are likely to be accurate, and which need to be calibrated?

Another issue may be available memory. The RAM needed to do the calculation goes up as the square of the number of parameters to be calibrated. Nine is OK because that's the maximum supported on linear delta printers.

Orang_

@dc42 said in Rotary delta calibration not implemented?:

Do you know which parameters of your machine are likely to be accurate, and which need to be calibrated?
Another issue may be available memory. The RAM needed to do the calculation goes up as the square of the number of parameters to be calibrated. Nine is OK because that's the maximum supported on linear delta printers.

M669 K10 U169.85 L320 R55.94 A-45:52 H432 X-60,Y60,Z180
U, L, - maybe accurate enough to measure;
R,H - within a small margin of error
A - it is difficult to measure without special tools
x y z - depends on the construction of the printer

Orang_

@dc42
"Another issue may be available memory. The RAM needed to do the calculation goes up as the square of the number of parameters to be calibrated. Nine is OK because that's the maximum supported on linear delta printers."

is it possible in the future to add support for spi or i2c external ram?
or is it easier to buy a board with more memory?

dc42

@Orang_ said in Rotary delta calibration not implemented?:

is it possible in the future to add support for spi or i2c external ram?

No.

or is it easier to buy a board with more memory?

Yes, such as the Duet 3 MB6HC or Duet 3 Mini.

Orang_

@dc42
Why doesn't the G30 work in any way? I can't properly measure multiple points(

dc42

@Orang_ said in Rotary delta calibration not implemented?:

@dc42
Why doesn't the G30 work in any way? I can't properly measure multiple points(

Because the auto calibration code has to be written to suit the kinematics, even though the basic principle will be the same as for a linear delta. We may also need a more sophisticated way of configuring which parameters are to be calibrated, rather than just having a # of parameters as we do for linear delta.

My guess is that the homing switch positions and delta radius are the most important to calibrate. Do you agree?