15mm Calibration Cube - not quite "cube-uler"

Last night I decided upgrade to the 1.09d-dc42 firmware and the latest 040815 software so that I could play around with the 7 point calibration. I changed the final G30 command from S7 to S6 so as not to change the diagnoal rod length, and after the a few iterations the values settled down and I updated config.g

Next up I printed out a 15mm cube, and measured it with a sheet of glass, an engineering set square and some electronic calipers.

The faces were all at 90 degrees to each other, and the layers were all nicely straight on top of each other - good news.

However the dimensions were not quite a cube, the "X" width was 15.04mm , the "Y" Width was 15.96mm and the "Z" height was 14.78mm.

Am I correct in thinking that I can use the following Gcode to compensate?

M556 S78 X0 Y0 Z0

If so what is the best way of calculating what values to put in? Would printing these from the Omerod instructions be suitable?

Alternatively Is there a better way of calibrating to make sure I can print a perfect cube.

Any suggestions would be greatly appreciated.

You can't use that gcode to compensate, that is for compensating for the faces of the cube not being at right angles, for example because the towers all lean in the same direction.

What layer height are you using? The Z height will be rounded by the slicer to a multiple of your layer height. You can look at the Layers preview in slic3r 1.2.9 to see whether it really is trying to print exactly 50mm high.

The extra 0.92mm size in the Y direction looks odd to me. One possible explanation is that the probe trigger height varies with position on the bed, and you haven't fully captured this in the G31 corrections in your bed.g. file. So the firmware is moving the tower positions to compensate for a geometry error that isn't there. What happens if you use just 4-factor calibration, with the tower offset corrections set to zero?

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
dc42
What layer height are you using? The Z height will be rounded by the slicer to a multiple of your layer height. You can look at the Layers preview in slic3r 1.2.9 to see whether it really is trying to print exactly 50mm high.

The extra 0.92mm size in the Y direction looks odd to me. One possible explanation is that the probe trigger height varies with position on the bed, and you haven't fully captured this in the G31 corrections in your bed.g. file. So the firmware is moving the tower positions to compensate for a geometry error that isn't there. What happens if you use just 4-factor calibration, with the tower offset corrections set to zero?

The layer height is set to 0.2mm. , Thanks for the suggestions, I'll reset the config.g and go try the 4-factor cal before reprinting after work.

Do you reckon it mighe be worth rotatating the bed by 120 degrees and see if the error changes axis?

I'd check if your bed is flat. This sounds as it's it's calibrating against a non-planar surface.

If the top an bottom of the cube is flat then your Z error is either down to using the wrong steps per mm in the config or your first layer(s) are being rammed into the bed. The latest SD image has a macro called setz.g which can be used to raise the Z origin to account for errors due to the probing method.

Rotating the bed is pointless as if there's a problem, you're just moving the error elsewhere. Indeed, I would specifically try to put the bed back in the same rotation each time to prevent you invalidating your calibration.

The bed 7 (with S6) came back as

Endstop adjustments X0.99 Y1.68 Z-2.67
Diagonal 160.00, delta radius 84.20, homed height 181.35, bed radius 75.0, X 6.48°, Y 1.35°

Endstop adjustments X0.92 Y1.70 Z-2.62
Diagonal 160.00, delta radius 84.05, homed height 181.29, bed radius 75.0, X 6.21°, Y 1.05°

Endstop adjustments X0.92 Y1.64 Z-2.56
Diagonal 160.00, delta radius 83.98, homed height 181.28, bed radius 75.0, X 5.96°, Y 1.21°

Endstop adjustments X0.91 Y1.67 Z-2.58
Diagonal 160.00, delta radius 84.03, homed height 181.29, bed radius 75.0, X 6.37°, Y 0.89°

After resetting the config.c the 4 factor calibrations the 15mm cube came out as x 15.26 , y 15.97, z 14.81.
and the reported values were

Endstop adjustments X1.01 Y1.31 Z-2.32
Diagonal 160.00, delta radius 83.62, homed height 181.18, bed radius 75.0, X 0.00°, Y 0.00°

The bed looks flat, the buildtak is clean as well.

Hum, something wrong here:
The X column angle is very important. Such an angle shall be very visible without any instrument. Considering the manufacturing method, its is possible that all columns are twisted similarly, but the calibration cannot detect it, however the fact that only one column is reclined so much is highly improbable.

The radius is quite important and indicates that the machine is considering a non flat area (the theoretical radius is 81).

There are two reasons why calibration considered a non flat area :
1/ It is really not flat (mine had 0.25 mm bend). Going to a really flat bed have for me reduced the apparent radius by approx 1mm.
2/ The measurement offset is much higher in the center than on the edge. This give the same effect as a lower center.
This have been described in another topic. However, I found that with well tensioned belts and no mechanical play, the difference is not that large. You can measure it manually by moving the head first to the contact of the bed (use paper sheet), then triggering the probe.
I have seen that a correction of 0.05 mm between the center and the edge reduced the apparent radius by 0.2~0.3 mm.

both corrections (flat bed and proper probe offset) reduced my part oversize from 3% to ~0.7 %. The apparent radius for me is now 82.1 mm, so still a bit too much, but a significant improvement.

You probably have a mechanical problem and shall check all your mechanics :
- Supports have not slided along the rods
- belts are well tensioned (but that tend to slide supports on rods...)
- Idler carriage is well vertical and the belt don't rub too much on the idler carriage (see other topic), that may drive to skipped steps
- Check the play of your carriage and arms, there shall be none

I agree, that X angle looks very suspicious.
The other thing to check.is whether the metal contacts and the balls at the corners of the bed are free of dirt and grease - as the probe method takes analog readings, a poor contact at any one of the contacts can skew the calibration. I'd also check that all of the bed springs are seated correctly.

Thanks for all the suggestions and help.

I have gone through and checked the tightening and alignment/straightness on everything and now get slightly more sensible values.

Diagonal 160.00, delta radius 82.05, homed height 181.82, bed radius 75.0, X 2.39°, Y 2.21°

The bed is definitely slightly higher in the middle vs the edges by about 0.2 Z steps, what is the best way to correct for this? shouldn't the auto bed levelling be calibrating that out?

The calibration can compensate for a non flat bed, if the default is homogeneous. However, there is a price to pay:

- The calibration establish a fictive geometry for the machine, to have in its 'local space' a flat bed. That means that the dimensions of your parts cannot be accurate. You could cheat the program by modifying measurement offsets, but in this case it will be impossible to have adhesion on the whole bed.

- All your layers will have the same shape as your bed

The calibration actually works as if there is a there is a dip in the centre of the bed. Its setup to assume that the build surface is flat and that *for some reason(s)* the centre point requires the probe to go down further in order to trigger. Therefore, there is a command in the calibration scripts that artificially offsets (lowers) the centre measurement point... so if your bed triggers the centre point 0.2 mm above the other points then the offset is going completely the wrong way for you. How flat is the other side of your bed? If it's flatter then you can flip the balls and BuildTak to the other side... RRP seem to think the BuildTak should come off okay, but I've not tried it myself.

I've also just noticed that there's significant variance in you endstop positions,I'd probably be more worried by that. Have you bent the endstop levers to 45 degrees at the notch mark on the the lever? It makes a huge difference to the calibration. If you have, are the endstops secured well or are the screws loose and causing the endstop to move? Is one or more of the acrylic sides bowing in/out causing the endstop position to be slightly off?

When its calibrated well the endstop offsets should be +/- 0.2 mm at the most, and similarly +/- 0.2 degrees for the tower angles. I'd worry less about the radius for the time being - indeed, my rod length and radius aren't set to what the CAD or manual measurements say they should be, but on my last 100 mm calibration piece the maximum dimension error was 0.25%, and all angles and flat surfaces were as they should be.

I was thinking that the calibration can compensate accurately for the endstop position.
And if you enter the Z-offset in the config file, you will have an accurate starting point, so why trying to adjust physically the endstop position, which is a bit challenging ?

The builtak can come off if you go slowly, but I lost some glue on an edge and edge is damaged where I used it to pull the ensemble.

Yeh, in theory it can, but I'm saying to check the endstops because values that big indicate a significant mechanical error if they are correct.
If this was a system where the endstops were mounted on brackets that were arbitrarily screwed to some aluminium extrusion towers (eg a Kossel), I could believe those values. But if the Fisher is assembled correctly the only error that those values should represent are the small variations in distance between the microswitch and the bed, and any differences in the microswitch's trigger point - these combined values should easily be less than 0.5. If they're not then something is wrong. Why go through the pain of making the calibration struggle to compensate for significant mechanical errors. Really, calibration should be used to compensate for minor deviations or build errors. The values seen here are not minor.

FWIW, I was getting values like that until I bent the microswitch arms, at which point the values dropped to <0.2 mm and it was able to calibrate the build plane flat very quickly.

Thanks for everyones suggestions, I bought a new buildtak on amazon, glad I did, as peeling off the old one broke it. I also bent the arms a bit more, and affixed them with cable ties rather than the bolts.

The X axis is looking much better, and the cube is coming out with square faces but slightly oversized at ~15.27mm, which I guess could be down to slicing or extruder steps/mm or filament diameter.

Diagonal 160.00, delta radius 82.13, homed height 181.93, bed radius 75.0, X 0.61°, Y 0.51°
Endstop adjustments X-1.52 Y0.85 Z0.66

That's better, but it's still not right. The homed height of ~182 is a bit odd too. The default height as 180 and it usually calibrates to less than that. I'd keep looking to see why your x endstop is so different to the others.
Are the ball contacts on the edge of the bed clean and tight? Same with the contacts they connect with on the printer?

I uploaded a target pdf in the Simplify3D topic (I think), you can put that in the bed while running the calibration to verify that head is moving to the correct position. If the head always closer or further away from the centre relative to the targets then the rod length is wrong. If it's hitting either side of the targets then the radius/endstops are wrong. Your error is around 2%, the same as mine was - my rod length needed increasing slightly. Just try checking this before assuming your slicer settings need tweaking.