Flsun Q5 delta printer, poor dimensional accuracy

[Dog]

Hi

I have recently installed duet 2 wifi on my Flsun Q5 delta printer, and I got it working ok. Print quality is nice, but the dimensional accuracy is way off. I have printed 75mm wide hexagon and the side to side measurements are 74,55 on X axis 74,89 on Y and 73,25 on Z (XYZ axis as in perpendicular to respective tower)

I think my measurements on delta rods are accurate, and i have tried to play with different dimensions in the firmware, but nothing seems to work, and everytime i do the delta calibration G32 the result seems to go towards the pattern of small inaccuracies on X and Y and large inaccuracy on Z.

M665 L215.0 R106.3 H215.0 B105.0
M666 X0 Y0 Z0

Currently running this and getting this gfrom G32

M665 L215.000:215.000:215.000 R106.077 H217.644 B105.0 X0.126 Y0.204 Z0.000
M666 X0.452 Y0.101 Z-0.553 A-0.41 B0.15

I tried setting different lengths on the separate axis rods, that seemed to have somewhat positive impact, but if i then ran G32 the problem reset back to the starting point more or less

I did add the measurements to config.g rather than the override created by G32 because otherwise the changes made would reset to the defaults provided to config.g when running G32. I also restarted the board between changes and verified that they were enabled when printing another calibration piece

https://www.thingiverse.com/thing:1274733#files The tool i used

I tried the excel calc method, but didnt get the result i wanted, if anyone has any input on what to try next please feel free to suggest anything Im not shooting for .01 accuracy, but atleast .2 or so would at least be acceptable

[jay_s_uk]

@dog can you post your bed file?

[Dog]

@jay_s_uk
bed.g
Here, im new to RRF so had no idea that this could have an effect on this issue.

[jay_s_uk]

@dog the S value on the last G30 controls what calculations are done with G32. See here https://docs.duet3d.com/User_manual/Tuning/Delta_calibration#setting-up-the-bedg-file
I however think that S8 is suitable so I wouldn't be too worried.
How are you measuring your rod?
i also find that i get better accuracy probing the bed on a smaller circle. say have the bed diameter.
i normally use this to generate a bed file https://escher3d.com/pages/wizards/wizardbed.php

[Dog]

@jay_s_uk I dont quite get what you mean by this:
"i also find that i get better accuracy probing the bed on a smaller circle. say have the bed diameter."
Wouldnt a circle with a diameter of the bed be as big as it can probe rather than being smaller? I measured my rods from the center of the bolt that goes through the balljoint to the other similar fixing point.

I'll try the S6 method as it is descrtibed as most reliable for most printers. I'll also go for a smaller circle say R70 or so. I'll inform later about the results i get with these

[jay_s_uk]

@dog sorry, meant half the bed diameter. my delta is 380mm so rather than probe at the extremes of the diameter to calibrate, i find i get better results probing at ~ half the diameter (at 140mm)

How accurate are your measurements? What are you using?

[Dog]

@jay_s_uk Digital vernier. Pretty accurate, but ofc there is some difficulty to estimate center of a hex bolt.

E: First results from same probing diameter, but with S6: No difference, running a test print with diameter of 50 (half the bed ish). Report on that later

[Dog]

No real difference with smaller probing area. Measured the rods again for a bit more accurate measurement, trying the S9 method, since i have had the best success by altering rod lengths even tho they all seem to be quite similar in size (max deviation of 0.1mm, if even that)

[Dog]

Finally got some positive results to share

Decided to go with altering the rod lengths, because that seemed to work and I kinda understood how it affects the print when it comes to dimensions. (measured length / desired length * rod length = new rod length ish) basically if the print is too small, shorten the rod, if too big, lengthen it.

these are the measurements I ended up with

in config.g:
M665 L216.750:215.000:210.100 R106.126 H217.714 B105.0 X0.229 Y0.223 Z0.000
M666 X0.601 Y-0.021 Z-0.580 A-0.53 B0.12

after delta calibration, probing 16 points with S4 method;
M665 L216.750:215.000:210.100 R105.838 H217.678 B105.0 X0.229 Y0.223 Z0.000
M666 X0.340 Y-0.123 Z-0.217 A-0.53 B0.12

I printed a hexagon with flat sides perpendicular to the towers. desired width was 75, and the results i got now are X:74,95 Y:74,95 Z:74,85 These could be extrusion issues not related to movement so decided to end here. Printed 30 ish hexagons chasing accuracy...

Decided to print 5 20x20 cubes, one on the middle, and 4 on the extremities of the bed, all spaced 40mm from the center both up and away (40:40) (-40:-40) coordinates and so on. Some measure bang on 20 on one side but 20,30 on the other while the middle one is good on 20,05 ish on all sides. The deviation between the cubes printed on the extremities differ slightly. I think this deviation on the cubes is result of changing rod lengths rather than offsets etc, but currently im fine with these results. Does anyone know how the offsets or tower angles affect the print directly? I know the current rod lenths are not correct real world measurements, and would like to try a more legitimate method of calibrating. Maybe then i could even get accurate parts all over the bed

[oliof]

@dog http://boim.com/DeltaUtil/CalDoc/Calibration.html gives you a nice explanations how different inaccuracies in a delta print affect dimensional accuracy.

[lord binky]

@dog said in Flsun Q5 delta printer, poor dimensional accuracy:

Decided to go with altering the rod lengths, because that seemed to work and I kinda understood how it affects the print when it comes to dimensions. (measured length / desired length * rod length = new rod length ish) basically if the print is too small, shorten the rod, if too big, lengthen it.

I've used a piece of carbon paper over a piece of paper and did a heated nozzle tap (takes less force to mark) to do basically the same thing. It can speed up the process without doing a full print and having to account for anything other than just movement simply measuring from the outside of one dot to the inside of the other.