Height Map and Leveling Issues + Head Movement

jens55

@tdm418 said in Height Map and Leveling Issues + Head Movement:
why shouldn't the value be the same regardless of the probe's location, as that is pretty much a fixed distance?

Because the glass isn't flat, the mechanics aren't perfect and there really isn't anything in this world that is a 'fixed distance'. Heck, even continents move!

jens55

An interesting thought since I also have this kind of a sawtooth pattern ....
If I manually move to a given point on the bed while approaching it from the left, take 10 measurements while staying at that point, then move further right and return, this time approaching the same point from the right and take 10 measurements .... should I see a discrepancy of the second set of numbers compared to the first set?
I think I will run that test and see what happens, should be interesting to see the outcome!
Is there a way to probe the bed that doesn't involve the linear travel and uses a more random pattern ?

Veti

@fcwilt said in Height Map and Leveling Issues + Head Movement:

Try writing a macro that does a series of G30 S-1 commands

here is a link to the complete m48 macro
https://forum.duet3d.com/topic/6962/m48-measure-z-probe-repeatability-and-print-to-serial-output

please run at different positions

tdm418

I am traveling for the next couple of days. I'll try it and let you know when I return.
Thank you

tdm418

So I have once again gone through the entire machine, tightening and testing. I did notice that there was some wobble motion in the Z axis during Mesh Compensation. So I reconfigured the leadscrews to get them a bit off the center of gravity for the bed and tore apart the guides. There was a backplate that had cracked, and whether that was the issue or not, the bed now moves smoothly and with no unwanted motion.
Unfortunately, the height map is no better than it was initially.
A great suggestion and link from @fcwilt and @Veti - The results are below. I have only had time to run about five tests,one in the center and at each corner:

11/22/2019, 11:25:01 PM M98 P"0:/macros/BLTouch/BLTouch Uniformity Test"
G32 bed probe heights: 0.138 0.140 0.140 0.140 0.138 0.135 0.138 0.135 0.135 0.138, mean 0.138, deviation from mean 0.002

11/22/2019, 11:19:56 PM M98 P"0:/macros/BLTouch/BLTouch Uniformity Test"
G32 bed probe heights: -0.130 -0.130 -0.130 -0.130 -0.132 -0.130 -0.132 -0.132 -0.132 -0.130, mean -0.131, deviation from mean 0.001

11/22/2019, 11:26:56 PM M98 P"0:/macros/BLTouch/BLTouch Uniformity Test"
G32 bed probe heights: -0.350 -0.345 -0.345 -0.345 -0.350 -0.352 -0.354 -0.350 -0.354 -0.354, mean -0.350, deviation from mean 0.004

11/22/2019, 11:28:40 PM M98 P"0:/macros/BLTouch/BLTouch Uniformity Test"
G32 bed probe heights: 0.465 0.463 0.463 0.465 0.463 0.463 0.465 0.461 0.461 0.463, mean 0.463, deviation from mean 0.002

11/22/2019, 11:30:19 PM M98 P"0:/macros/BLTouch/BLTouch Uniformity Test"
G32 bed probe heights: 0.758 0.760 0.758 0.758 0.758 0.760 0.758 0.760 0.758 0.758, mean 0.759, deviation from mean 0.001

It seems like this is indicative of pretty good uniformity, so unless I'm missing something, I still have not found my gremlin.

Veti

can you level the bed that its mostly planar by using the leveling knobs?

i.e so that all the points where you probed with m48 macro report values close to 0

tdm418

I spent hours yesterday leveling the bed with the probe and it is just not possible.
Wherever I run either the macro or even just a single G30 command, I can adjust the bed at that point based upon the returned value, and then go on to another corner. I have done this for all four corners and gotten the returned values almost perfect. Using the macro, I am getting deviation values from .001 to .003, typically .001. However, the bed is not level. If I move the printhead away from the coordinates where I have just probed, I get a very different height value. And probing four corners only exacerbates the problem. Even though I am not using mesh compensation, the returned values for bed height are as erratic as the mesh compensation map, so they are useless.
A new interesting, or frustrating, problem that I have encountered is that when using the G30 command tied to specific x y coordinates, the x coordinate is always 35mm less that what is entered. If I use the command G30 P0 X180 Y180 Z-99999 S-1, the bed is probed at the proper y coordinate, but the x coordinate is 145. If the x coordinate is less than 35, it will push the head to a negative position. The documentation indicates that it is allowable with the G30 command to probe outside the printable area, but something is causing this 35mm subtraction.
I'm still at a loss here...

infiniteloop

@tdm418 Please don’t tell me that you can’t level your bed somehow, your 9-point mesh posted above says that your print bed is almost flat. Just like you, I have a bed with four adjustment screws, so you might have a look at my macro and adopt it for your purposes. However, my setup differs from yours:

The X/Y origin is in the center of my bed
Instead of a BLtouch, I use a mechanical probe which triggers an optical end stop
The offset of my probe differs from yours (mine is X0 Y9 Z2.20)

; levelbed.g

; Assisted manual bed levelling.
; NOTE that we must re-home the Z axis after any adjustments!

; Screws start at the rear left, they are counted clockwise


M98 P"homez.g"				; home Z (will deselect tools, too)
G1 Z4					; go down to a proper probing height
G29 S2					; disable mesh grid compensation
M400					; wait until all moves are finished

M671 X-124:118:118:-124 Y116:116:-116:-116 P0.5	; position of the adjusting screws, these make 0.5 mm per turn
; NOTE: Y coordinates are NOT adjusted for probe offset

M401					; deploy the mechanical Z probe

; Now, probe the bed and return adjustment values:

G30 P0 X-124 Y107 Z-99999		; rear left
G30 P1 X118 Y107 Z-99999		; rear right
G30 P2 X118 Y-125 Z-99999		; front right
G30 P3 X-124 Y-125 Z-99999 S4		; front left, calc deviation of screw heights
; NOTE: here, Y coordinates are adjusted for probe offset

G1 X0 Y-9 Z4 F10000			; put head over the center of the bed (Note: compensate for probe offset)
G30					; probe to set Z height
M402					; retract the probe

My “NOTE:“ comments tell me where to apply the probe offset to the coordinates and were not - there you might find the source of your frustration with wrong X coordinates. However, my case is a bit different as I have to disable my tools for probing, so it’s better to consult the documentation.

fcwilt

Hi,

What are your probe offset values?

While it is quite possible to use the G30 form of G30 P# X## Y## Z-99999 to help level the bed I found using G29 worked better for me because it yielded a picture (the height map) of what was wrong.

Now the range on my X axis is -150 to 150, the Y range -100 to 100.

So I used this to specify a grid of just 4 points near the bed adjustment screws: M557 X-140:95 Y-145:145 P2:2

Notice the :95 above - this was needed to allow for the probe's Y offset so it didn't try to probe too far to the right and hit the endstop.

This 4 point probing didn't take long and then I could take a look at the height map to see what was going on. I could hover over the 4 points to get the numbers but the visual representation helped me more than the numbers.

It wasn't long before I could look at the height map and have a good idea of the adjustments I needed to make. Usually running through the procedure 5 or so times was enough to get the bed level (+/- 0.002).

Of course this is only at those four points. But once the bed is mostly level then running G29 with a tighter grid yields the information needed for mesh compensation to work while printing.

Frederick

tdm418

I should have specified - It is impossible to level with the BLTouch as it has started reporting weird values. Of course I can level the bed with a feeler gauge and the leveling screws. My very first post on this subject is a picture of the mesh compensation height map after physically leveling the bed. Again, this happened overnight and I can't find an explanation for it.
Probe offset values: G31 X35 Y0 Z1.202 P25, that is the "missing" 35mm. Thank you both for pointing that out. And since the probe is reporting erratic z values, the number of points that I probe does not matter.

Phaedrux

Perhaps time to clean the pin? I often get little wisps of filament gather up like cobwebs around the print head and sometimes they get stuck on the BLTouch pin.

infiniteloop

@tdm418 said in Height Map and Leveling Issues + Head Movement:

It is impossible to level with the BLTouch as it has started reporting weird values.

Sigh! If you need help with the BLtouch: I’ve counted more than 200 threads having this thing in their titles, all in all, I found more than 600 mentions…

In case you look for a reliable replacement, you might consider this design (recommended by @Veti) or this one.

tdm418

My most recent height map still has the geometric pattern. It has improved from the initial one, as the bed is now recognized as being level (sort of), and the deltas are decreasing. I would post a new picture, but am getting a server parsing error whenever I try to upload the screenshot.
I can't rule out the probe being the culprit, but the irregularities seem too consistent and rhythmic to be a probe failure.
@Phaedrux , I have tried cleaning the probe. Like with rebuilding the Z axis, the motion (to my eye) seems smoother, and the height map gradually improves with every new attempt, but it is still indicative of a different problem.
@infiniteloop , you agreed initially that this was a mechanical problem. The BLTouch is only six weeks old and has given me no trouble. In fact, it was one of the add ons that really seemed to improve the print quality. It looks like you have done some research, do you think that a failing unit would cause this sawtooth pattern?
Again, thanks to everyone for their input and suggestions.

fcwilt

Hi,

Well as dc42 mentioned the probe height is changing when going in one direction versus the opposite direction. Thus the peaks and valleys.

You will have to determine what about your hardware is changing as the probing changes direction.

Are your belts really tight?

Frederick

tdm418

Yes, the belts are tight enough that they vibrate when plucked, like a guitar string. I can't replicate any unusual motion even though I have physically twisted and pulled on everything in an attempt to replicate it. This is way beyond frustrating, but hopefully it will turn into a learning experience that can be corrected and shared.
I'm going to keep after it, but would still greatly appreciate any suggestions since I'm out of ideas, for the time.

jens55

Have you run the tests that I suggested? Pick a position to measure. Approach that position with the print head from one direction, do a bunch of repeated height measurements. Move the head to the far side and then turn around and approach the same measurement position from the other side and repeat height measurements. This should show clearly if the sawtooth surface map is generated by mechanical slop somewhere.
If the above was done on the x axis then you might repeat the same thing on the y axis just for the heck of it.
I am planning on doing this test on my printer but have not had n opportunity to try it. In theory it should be definitive.
... I did say 'theory' .... YMMV

tdm418

I'm hoping to get to that one tomorrow. Will definitely let you know the results.

infiniteloop

@tdm418 said in Height Map and Leveling Issues + Head Movement:

you agreed initially that this was a mechanical problem.

Yep, indeed. A functional BLtouch will not produce such a pattern, and I can’t imagine any reason why a dysfunctional probe should change its mind every second row in such a predictable manner.

However, I am no BLtouch expert, I don’t own one (and am glad about this, because I read tons of help requests in literally hundreds of threads). Instead, I tried different probing techniques until I finally understood the simplicity and beauty of the beforementioned design idea: a pin, an optical endstop and a cheap servo for retraction - that’s it.

After this confession, back to work: In case you still can generate said “sawtooth“ pattern, try to probe a single point repeatedly (yeah, I see @jens55 had the same idea) - are the results consistent? If not, have a closer look at your BLtouch, else: welcome to the mechanical age (somehow reminds me of “Myth“).

So, if the probe is ok, could you try to reduce the travel speed between probing points to something below 200 (“T“ parameter in M558), use at least 9 repetitions („A“) and set the recovery time (“R“) to half a second? All this makes the mesh probing painfully slow, but if you get better results than before, you should reconsider the mount of the BLtouch: it might wobble.

fcwilt

@tdm418 said in Height Map and Leveling Issues + Head Movement:

Yes, the belts are tight enough that they vibrate when plucked, like a guitar string. I can't replicate any unusual motion even though I have physically twisted and pulled on everything in an attempt to replicate it. This is way beyond frustrating, but hopefully it will turn into a learning experience that can be corrected and shared.

Well keep in mind we are talking very small mounts of unwanted movement, nothing that you could see with your eyes.

Have you tried making the belts less tight? You may have too much tension causing the some parts of the gantry to deflect/deform.

Frederick

tdm418

@jens55 - So I tried your idea, and found nothing. I came at the same point from + and - x and y, as well as a few diagonals. In general, the measurements were with .003, with a max of .005. Great idea, unfortunately, this thing is extremely uncooperative.
@fcwilt - The belt tension makes almost no difference in the height map at all.
What I did notice, though, is that no matter what the tension, the map seems to constantly improve. That got me thinking that there is a bearing or wheel with a "flat spot" that is gradually working itself out. That would seem to be impossible, of course, to have anything grabbing with such regularity, but I can't deny that everything that I try does improve it. So I have pulled apart every moving piece of this thing (again) and am testing and lubricating them all, even though the bearings are sealed. I even lubricated the power supply, just in case.
I had the machine in Corpus Christi with me, so maybe somehow there is a trace of oxidation that I can't see that is slowly wearing off now that I'm back in a very dry climate. Not a good theory compared to all the fantastic ideas already presented, and the regularity of the pattern would seem to preclude this as well, but I'm really grasping at straws here.
Still no "smoking gun" but I'll see how it does after a few test prints.
Oh, just kidding about the power supply...