G30 A command?

I am new to 3d pprinting and I just used the G30 A command on my delta printer. It has been going for about an hour now and hasn't stopped probing. Is this normal?

Sounds like you need better firmware. I can do 7-factor auto delta calibration on my printer in 35 seconds.

---

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].

I was assuming the G30 A command was associated with Rich Cattell's Marlin branch and still under development. You also want to make sure you are using the Testing branch rather than the Marlin-Main (?) branch.

I also believe the Marlin-Main branch defaults to 0.03mm accuracy, which you can change by using a G30 A0.05 command.

The time it takes to calibrate the machine is also somewhat determined by your manual calibration.

It finished after about 3 hours, but it completely stuffed up my machine. When it prints, the dimensions are off by a heap and the bed leveling is worse than before.

Have you checked all moving parts of the printer before auto calibration?

Based on my experience, auto calibration wouldn't fix mechanical problem.
In one occasion, one or two diagonal length were constantly changing which made incorrect calibration by manual.
Auto calibration failed after 100, the maximum iteration and stalled the RAMPS 1.4 controller showing "Calibration success" message in the end.
In other occasion one pulley of the stepper motors were loose it was moving side by side. I found it during auto calibration by hearing unfriendly noise.

After I fixed the mechanical problems, auto calibration finishes within 5 to 7 iteration.

Present


Previous auto calibration


Manual calibration


Edited 1 time(s). Last edit at 04/02/2015 08:42PM by janpenguin.

What were the main issues you fixed manually? Im at iteration 57 @0.5 precision and while most of the values stay the same, I cant tell if hes going in the right direction.

Admittedly I haven't spend much time squaring the frame and my cross braces aren't installed as well, would it be enough to calibrate a 1mm precision to at least start printing?

FWIW, I just flashed Rich Cattell's Testing branch of Marlin to my delta, and ran the G30 A calibration.

It took 6 iterations to converge to AUTOCALIBRATION_PRECISION 0.05// mm :

Quote
Marlin
Recv: Iteration: 6
Recv: Checking/Adjusting endstop offsets
Send: M105
Recv: x:0.2563 (adj:-0.2625) y:0.1250 (adj:-0.9937) z:0.1250 (adj:-1.8375)
Recv: X=ERROR Y=ERROR Z=ERROR
Recv: x:0.0375 (adj:-0.2250) y:-0.0125 (adj:-1.0062) z:-0.0312 (adj:-1.8687)
Recv: X=OK Y=OK Z=OK
Send: M105
Send: M105
Recv: | Z-Tower Endstop Offsets
Recv: | -0.0187 X:-0.22 Y:-1.01 Z:-1.87
Recv: | 0.0000 -0.0187 Tower Offsets
Recv: | -0.0375 A:0.00 b:0.00 C:0.00
Recv: | 0.0313 -0.0062 I:-2.13 J:0.00 K:0.00
Recv: | 0.0188 Delta Radius: 174.0000
Recv: | X-Tower Y-Tower Diagonal Rod: 373.4001
Recv:
Recv: Autocalibration Complete
Recv: ok

So I guess it does converge, as janpenguin wrote above, within 5 to 7 iterations. I am porting dc42's calibration algorithm to Marlin and in principle it should be able to run alongside Rich Cattell's calibration code, so we should be able to compare their performance and accuracy ASAP sometime in the future.

I will take some time to square up the frame.
Im trying to find a precision level that makes its under 100 iterations

@andrew: Great, I cant contribute by programming but let me know when your finished Im happy to beta test it.

Hi @AndrewBCN,

Did you make any progress with porting @dc42's algorithm to Marlin? Do you need any help; I might be able to find some time in the near future.

I have a modified 3DR Simple delta that now includes dc42's IR probe sensor. Previously, I was using FSRs under the bed (with the heated bed supported on silicone pads) and used JohnSL's FSR endstop controller board. I used RichC's main branch of Marlin for the auto calibration and for the most part it worked fine (in about 10 iterations) but I did have to fix the diagonal rod level as a constant otherwise the solution never converged. I had also modified the probe function to take the median of 5 readings rather than the default of finding two consecutive identical readings.

Basically, I liked the FSRs except for the fact it isn't then possible to hard fix the heated bed to the base since it has to rest loosely on the sensors. I didn't like this as, not only was it untidy, but also it loses one of the delta advantages over cartesian printers which is to have a rigid bed assembly. Hence, my recent move to @dc42's IR probe. I have this mounted as close as I dare to an E3d v6 hot end. What I then found was that RichC's original autocal Marlin would never converge. I then moved to his Test branch of code. This is significantly different in the way it works but does converge to 0.09 accuracy after about 7 iterations (15 minutes) with my IR probe.

I have to say thanks to RichC for all the work he's put in on the Marlin auto calibration over the recent years and he has certainly allowed me to get my delta performing adequately. I do think, though, that this empirical approach may be susceptible to convergence problems which is why I'd like to try the least squares fit of @dc42's algorithm.

So, please let me know how it's going.

Best regards