bed mesh problem with Delta after upgrade to 3.3 from 2.05_1

T3P3Tony

here is your mesdh.g

; mesh.g
M117 Bed Map - Heating Bed                              ; display message
G28                                                     ; home
M140 S80                                                ; set bed temp 
M190 S80                                                ; wait for bed temp
M140 S0                                                 ; bed heat off
G32
M558 P8 I0 A10 S0.01 R0.4 C"zprobe.in" H8 F1500 T6000         ; HOTEND PROBEset Z probe type to switch and the dive height + speeds
M557 R140 S14                                           ; HOTEND PROBE define mesh grid
G29 S2                                                  ; clear height map
G29 S0                                                  ; run mesh calibration
M500                                                    ; save results to EEPROM
;G28                                                     ; home
M375                                                    ; load heightmap
;M84                                                     ; motors off
M117 End Mesh Probe Calibration                         ; display message

You need to add G30 to probe the bed and establish a Z datum before starting the mesh

best to do that in the center of the bed so I would add a line to move the probe to the center of the bed, then probe with a single G30 at that point, the start the mesh bed levelling

Baenwort

@t3p3tony

Per the Duet Wiki: https://duet3d.dozuki.com/Wiki/Using_mesh_bed_compensation#Section_Establish_a_Z_0_datum_using_the_Z_probe the Delta AutoCalibration (G32) should produce a Z-0 and be used for Delta machines instead of G30.

G30 isn't the recommended way to set Z height for the delta configuration as it is basically a multiple z-axis bot.

If the Wiki is wrong I'm happy to try with a G30.

jay_s_uk

@baenwort I find its easiest to home using G28, then do a single probe with G30 and then run G32 and save the results. That gets the homed height correct. And then from there, a G32 should be all that's needed at the beginning of a print

dc42

@baenwort doing G32 immediately followed by G29 is the correct method. However, one thing I notice is that you change the probing speed (M558 F parameter) between the G32 and the G29 S0. What happens if you remove the M558 line from mesh.g so that you use the same parameters for G32 and G29?

Also, if you are going to use G29 S2, put it before G32 instead of after.

You don't need the M375 command in mesh.g because the height map will be loaded anyway after G29 S0.

Phaedrux

@dc42 said in bed mesh problem with Delta after upgrade to 3.3 from 2.05_1:

Also, if you are going to use G29 S2, put it before G32 instead of after.

If you do this, remove the G29 S1 from the end of your bed.g as well.

Baenwort

Thanks!

I'll give these a try tonight after I get the kids in bed.

This has been frustrating as I like my flex plate system overall but without mesh compensation it isn't as good as glass for large bed filling prints.

I didn't run G29 when I had just glass and a doping agent. But I find that the mag sticker and flex plate added enough nonuniformity that I need it once I get away from the center of the bed.

Baenwort

@dc42 So it seems that the F parameter being different or the G32/G29 S2 order change was what did it. I'm going to try testing tomorrow to see which of the two changes was the causative one.

I hope it is just the F parameter change as I'd like to keep the G29 S1 load in the bed.g as I run a G32 before every print and I'd rather not forget a G29 S1 load.

Baenwort

So I tried with only the F change and it fixed it. The G29 S2 positioning did not affect the mesh to cause the problem I am seeing.

dc42

@baenwort said in bed mesh problem with Delta after upgrade to 3.3 from 2.05_1:

So I tried with only the F change and it fixed it. The G29 S2 positioning did not affect the mesh to cause the problem I am seeing.

I'm glad you solved it. I think what's happening is that the trigger height depends on the probing speed, at least at high speeds, because of the response characteristics of the sensor.

Baenwort

@dc42

Seems so. It is a Duet Smart Probe based system from SeeMeCNC. Being a strain probe I expected maybe some difference but this was a rather large difference in results.

It makes me think I might want to spend some time tuning the F parameter.