Hello. Sorry for my silence.
I have broken my glass bed. Waiting for the new one to do new tests.
And the new one will be this bed https://flex3drive.com/product/printbite/?attribute_size=Ø+250mm&attribute_colour=Clear&attribute_graphic=Plain&attribute_texture=Matt

@Danal Thank you for your help. Unfortunately I don't know my rods length and my caliper is too small.
They are not original Hayden Huntley ones but they are the same.
I agree with you, autocalibration is not perfect and the rod length has too much difference after each autocalibration. And I guess that the rod length is the most critical settings for good calibration.

I tried to measure the rod length of each one using this method.
For each tower:

• I fix my caliper on the tower. It's perpendicular to the tower.
• The end of the caliper is touching the head.
• I move the head 50 mm along the axis of the tower to the tower

Tower Y exemple
G1 X25.98 Y-15
G1 X69.282 Y-40

• Then I change the setting of the rod length and measure again until the head move the right distance.

Here is an exemple for the Y tower

Not sure than this method is the best one and if it's accurate enough. But I was surprised when I read from one autocalibration quite the same result.

As you can see, printing using this settings shows that the 120° angle are wrong. 100x100 square is not really square and 100mm round piece is not really round as well.

What about manually changing M665 X,Y and Z settings ?

@Phaedrux Thank you. I know it and I did my best by following this methods many time. The main difficulties is to find what you really have to fix according the issue you face. I have checked everything many time, and for sure touching the hardware add sometime new issue.

Tested with Chrome, no disconnection at all

@Phaedrux Thank you. Perhaps it's due to the browser but It doesn't happen before the upgrade with the same browser.
I will check with another browser and let you know.

Hello.
I have the same issue since I upgrade to 2.05.1.
I run before 2.04 beta

When I switch to another tab of my browser and then go back few second later to the Duet one, the UI is disconnected, and then connected back automatically

Connection interrupted, attempting to reconnect...
HTTP request timed out

Hello.
I need your knowledge and experience about autocalibration on delta printer.

Below are my steps of autocalibration.

Initial settings

M665 L288.6:288.6:288.6 R170 H261.6 B114.0 X0 Y0 Z0.000
M666 X0 Y0 Z0 A0 B0

1st autocalibration

G32
Calibrated 7 factors using 13 points, deviation before 0.367 after 0.010

M665 L291.540:291.540:291.540 R171.633 H261.667 B114.0 X-0.351 Y-0.271 Z0.000
M666 X0.958 Y-1.234 Z0.276 A0.00 B0.00

2nd autocalibration

G32
Calibrated 7 factors using 13 points, deviation before 0.009 after 0.009

M665 L291.684:291.684:291.684 R171.682 H261.665 B114.0 X-0.350 Y-0.278 Z0.000
M666 X0.963 Y-1.236 Z0.273 A0.00 B0.00

3rd

G32
Calibrated 7 factors using 13 points, deviation before 0.012 after 0.010

M665 L291.670:291.670:291.670 R171.679 H261.660 B114.0 X-0.337 Y-0.253 Z0.000
M666 X0.953 Y-1.229 Z0.276 A0.00 B0.00

Last one

G32
Calibrated 7 factors using 13 points, deviation before 0.010 after 0.009

M665 L291.669:291.669:291.669 R171.687 H261.659 B114.0 X-0.327 Y-0.234 Z0.000
M666 X0.933 Y-1.218 Z0.285 A0.00 B0.00

And then Mesh Compensation

21 points probed, min error -0.015, max error 0.065, mean 0.011, deviation 0.026

According to me it looks good, it could be better but it should run well.

So I printed this STL file to test the angle between each tower.

Angle calibration 0.6.stl

I have 2 questions:
1/ What is your opinion about the result of autocalibration and mesh compensation ?
2/ If you zoom in, you can see that the angle correction is not perfect and can affect the squareness of big prints. Autocalibration seems not to be able to do better. Is there a way to have perfect correction ?

Thank you.

@peirof Yes sure, Before doing an autocompensation I always did autocalibration. My first goal is to have a good autocalibration.

@dc42 As I'm not able to solve the effector tilt effect on the edge of the bed, I set in the beg.g a radius of 70 mm instead of 140 using 13 points (This is the radius of the calibration tool above)

G30 P0 X0.00 Y70.00 Z-99999 H0
G30 P1 X35.00 Y60.62 Z-99999 H0
G30 P2 X60.62 Y35.00 Z-99999 H0
G30 P3 X70.00 Y0.00 Z-99999 H0
G30 P4 X60.62 Y-35.00 Z-99999 H0
G30 P5 X35.00 Y-60.62 Z-99999 H0
G30 P6 X0.00 Y-70.00 Z-99999 H0
G30 P7 X-35.00 Y-60.62 Z-99999 H0
G30 P8 X-60.62 Y-35.00 Z-99999 H0
G30 P9 X-70.00 Y-0.00 Z-99999 H0
G30 P10 X-60.62 Y35.00 Z-99999 H0
G30 P11 X-35.00 Y60.62 Z-99999 H0
G30 P12 X0 Y0 Z-99999 S7

Here is the result of 2 autocalibrations

14:32:51 M666
Endstop adjustments X-0.31 Y0.60 Z-0.29, tilt X0.62% Y0.12%
14:32:48 M665
Diagonal 287.262, delta radius 169.759, homed height 227.619, bed radius 114.0, X -0.516°, Y -0.286°, Z 0.000°
14:32:27 G32
Calibrated 7 factors using 13 points, deviation before 0.009 after 0.006
14:31:53 M666
Endstop adjustments X-0.30 Y0.57 Z-0.28, tilt X0.62% Y0.12%
14:31:49 M665
Diagonal 286.994, delta radius 169.712, homed height 227.628, bed radius 114.0, X -0.554°, Y -0.313°, Z 0.000°
14:31:31 G32
Calibrated 7 factors using 13 points, deviation before 0.244 after 0.004

And the result of Autocompensation

When I print the calibration tool. Angles between Y/X and Y/Z are almost perfect. But the distance are still wrong. X 49.86 Y 50.50 Z 50.23

Step by step checks

Towers must be parallel to each other
My Delta has metal corners. I have checked spacing with a caliper at bottom, top and half way. the max difference between 2 towers is less than 0.2mm

Bed should be at right angles to the towers in both X and Y directions
I have checked with a square, the bed is perpendicular to the 3 towers

All 6 rods must be the same length, measured between bearing centres
Checked, see above.

The spacing between the bearings at the top of a pair of rods must be exactly the same as the spacing between the bearings at the bottom
Corrected, see above

The joints must have enough angle of movement to allow the nozzle to reach all of the printable area
I use magnetic balls, not an issue.

The two bearings on each carriage must be at the same height
Should be by design, I don't know how to check.

Carriages must not be rotated or warped about the Z axis
It should not be my case. Check the video with the bulls-eye level. I suppose If one of the carriage has this issue, the level should move during mouvement along the Z axe.

There must be no play in the joints
I use magnetic balls

I have printed this calibration tool https://www.thingiverse.com/thing:745523

M665 L289.536 R170.677 H227.388 B114.0 X-0.504 Y-0.516 Z0.000

It seems that autocalibration failed to correct the Y tower position. And it seems that I have a rod length error as well, but all lengths are equal

The result is better, mostly for the first layer but it could be improved further.
What are the other reasons which can tilt the effector ?

I have printed a large rectangular object which should measure 120x170mm, but in fact the print is smaller and not rectangular, the shape is trapezoid
Only the height is correct.