Page MenuHomeAleph Objects Inc

Inconsistent skirt measurements
Closed, SpitePublic


We've seen this thru alphas/betas, sincere apologies it doesn't appear a ticket had been submitted for this issue yet.
After a successful X-Axis leveling (both sides visually bottoming out on Z-Lowers) the left side of the X-Axis is about 1mm lower on the left.
This is outside of what the bed leveling matrix is able to compensate for and results in prints being heavily squished on the left, and not so on the right.
I checked the X-End Idler & Motor models and verified that the X-Axis smooth rods are the same height from the bottom of each printed part so I cant quite figure out where this is coming from.

Event Timeline

There are a very large number of changes, so older changes are hidden. Show Older Changes

Hibiscus 0013

front - .36
right - .38
back - .26
left - .26

MikeR added a comment.Apr 25 2018, 8:38 AM

Hibiscus 6

Front: .39
Right: .44
Top: .34
Left: .43

MikeR added a comment.Apr 25 2018, 9:19 AM


Bottom: .28
Right: .43
Top: .34
Left: .37

logan added a subscriber: west.Apr 25 2018, 11:23 AM

@tutley @marcio @west @samantha
Ball is in your court on this one, I'm out of ideas. GL

MikeR added a comment.Apr 25 2018, 3:30 PM

Hibiscus 19

Front: .43
Right: .51
Back: .49
Left: .32

@kent has run a hysteresis test on the z axis of one of these machines experiencing an uneven first layer. He has found a loss of about 350 microns of motion when the z motors change direction (on this particular machine). This is z "flatlining", and will act as backlash. Every time the motor changes direction it has to move about 350 microns before any actual motion is applied to the x gantry. This means that if the auto bed compensation is trying to compensate for anything under 350 microns, no compensation will actually be observed. This could cause an uneven first layer if the build platform isn't very flat. I ran a similar test on my machine which is not exhibiting an uneven first layer and found about 150 microns of lash. I have proposed a method to @marcio for a way to automate generating this value by using a probe point, and apply it to a backlash compensation for the z axis. This way we could have a machine specific value similar to z offset that would correct the issue of inconsistent first layer squish. However I am not sure that this method would be worth pursuing since this issue of the lost z motion only affects the first print layer and all following layers level out. If we incorporate backlash compensation then all layers are affected and we could run into bigger problems. I feel the inconsistent first layer issue may not be as big of a deal as it has been made out to be and most users wont notice/care if the first layer is a tad more squished on one side of the first layer.

Based on the data presented above it appears none of the skirts vary more than about 220 microns, and in fact most are under 150 microns.

This issue is happening because the belt has some elasticity to it and because of bushing friction/drag. I think the issue could be reduced by minimizing the amount of compression on the z compression bushings or removing the compression of the bushing altogether as well as increasing the amount of tension on the z belts to reduce the impact of the springiness of the belts.

We will see this same issue with Quiver.

@logan @karrad @marcio @kent @samantha @Steven any thoughts?

@tutley @kent If we can take one that is having issues and get a value, and implement compensation on that machine then get before and after print samples I feel that should give us some answers. I don't think it should be overly difficult to determine if that introduces more issues or not, though it may take more time that we have before release or at least start of production. May be something to look at for a post-release firmware update at this point, but as you said I don't think most users will have a problem with it for now. Personally, I wouldn't mind, but I would want to solve it eventually for some more niche prints. And, if we do have a potential solution this is something that could bump the quality of the machines to a level more expected in some industrial applications I think it should be investigated but I don't feel it should effect product launch.

@tutley For clarity, is it possible for us to determine what the z-backlash was on Gladiola? I think it would help for understanding the gravity of or lack thereof of the issue.

logan added a subscriber: clynch.Apr 27 2018, 9:59 AM

@tutley @samantha I guess we truly wont know if customers will notice or not yet and it depends on the severity as well. If we can get them under 150 microns in difference it may not be noticeable to the average user. But that is entirely conjecture at this point.
I feel it necessary to point out that this hysteresis WILL affect the print beyond the first layer, it may not always be visible but it will affect the print dimensionally. Check out @clynch's comment on T1814 there is a good example of how our users in the field might experience this issue beyond the skirt/first layer.
If we decide that this is acceptable for our users to experience, to me that would be a drop in product quality that is inconsistent with current expressed goals of positioning ourselves as a premium product for professionals, based on having no knowledge of previous generations of the Mini exhibiting this issue. We can certainly verify this but less hysteresis is inherent in the gladiola design because of the drive rods vs belts.
As far as whether or not this should affect product launch; we have in the past sent machines to reviewers with known issues. I don't feel it necessary to reiterate the results of that occurrence here.
Even if the issue isn't solved entirely (FW compensation) I think we should do anything we can to minimize how apparent the hysteresis is prior to launch and AFAIK that includes making sure the frame is square and the belts as tight as we can comfortably run them.

@logan I definitely agree that everything we can do mechanically should be done, I don't promote making poorer quality machines to compensate for in an update later. But as far as the firmware adjustment is concerned I feel we should look into it as time allows presently, and the nature of firmware makes it much easier to update later than mechanics. Would be extra nice if we can get it sorted quick enough to have first machines shipped with it, but being so close to launch and not having much time test is a concern.

@logan yes you are correct that some layers beyond the first will be affected but not all as the layers will quickly level out. I believe what caused @clynch's issue with the movable joints, was the "elephants foot" issue associated with a squished first layer.
I don't really agree that there is no knowledge of this issue on previous minis. I have definitely seen a lot of prints in the cluster and elsewhere that are more squished on one side. Im sure most anyone who has worked in cluster has also seen issues with inconsistent bead height on the first layer on many foxgloves and gladiola machines. @EricNugent @Shawn feel free to add any input on this.

I do think we need to correct this issue, however i feel it is an issue that has been around in some form, and is now being perceived as a brand new issue.

@benmalouf I will get some info on the gladiola hysteresis for comparison

@benmalouf I tested 4 older minis with @samantha

Hysteresis Results:

  1. 100microns (Gladiola1)
  2. 50 microns (Gladiola2)
  3. 50 microns (Gladiola3)
  4. 50 microns (Foxglove1)

@tutley if you average that it is under 1/6th of the hysteresis we're seeing on Hibiscus.
I meant no knowledge on my part, but as far as uneven first layers in the cluster, how many of those were on machines that had a clean nozzle and a level X-Axis? That cant be confirmed and there are other factors in those occurrences that we have eliminated from this situation through our troubleshooting. We can confirm that it happens consistently on Hibiscus, with a clean nozzle, a level X-Axis, matched washers and spacers, and sheetmetal that is all within spec. While it MAY have been around in some form, never to this level of severity and consistency as far as I am aware. So I disagree that it is a difference of perception vs previous machines making this appear as a new issue. To clarify; this is something I have not seen previous machines do consistently when calibrated and operated properly (cluster techs cant, and often dont watch every single print start). Marlin auto-leveling has its limitations (clearly) and I have always known that it can be at times inconsistent, but I have never seen it this consistently wrong.

@lansky can you find any record of customer complaints of uneven first layer that were not solved by leveling the X-axis and ensuring the probe sequence completes cleanly on either TAZ 6 or Mini 1.04 and previous?

@tutley we can confirm that we have seen issues with first layer issues in the past on foxglove and gladiola machines. Inspecting parts on the floor should help identify just how different the squish looks on the back of parts, not to mention the failure rates we see in cluster which doesn't make it downstairs. Over time bearings will wear down which magnifies this issue as well, which warrants the need for constant preventative maintenance in cluster as well. Taz6 Z upper has a very common occurrence of the variance in first layer, so this isn't just an issue on mini.

logan added a comment.Apr 27 2018, 2:06 PM

@EricNugent Can mechanical issues not inherent in the design be eliminated as causes of those issues? Essentially, is it consistent? Does the same machine exhibit the same traits on every print despite anything you've tried mechanically (down to inspection of sheet metal)?
Those are important questions to answer if we're comparing the two. The best solution we found was ensuring the frame is as square as possible, but that does nothing to address hysteresis.
First layer differences of parts on the floor are often clear when they do occur, granted, and I have no exposure to the failures the cluster sees other than what is received on the floor and the condition of the machines I have performed repair on for the cluster in the past. From that, I agree that the wear on cluster machines and the ways in which we need to improve preventative maintenance contribute to the issues the cluster sees.

I feel as though delving into further comparison may be distracting from the need to solve or alleviate this issue as best and efficiently as possible. Let me know how I can assist with solving this through manufacturing processes, I will be happy to work to implement the solution.

@logan the issue is typically resolved with inspection of parts and repair. So it is not consistent.

I cannot speak for mini2 as we only have a very limited amount of experience printing with it in cluster. Once we receive units that are approved for part production we can actually begin providing the feedback on the issue you are experiencing.

I understand that you have experience with the machine and are ultimately interested in providing a quality product; however, I don't think this conversation is approaching the issue correctly as I fail to see how the including comments of other technicians in separate departments should factor into the data needed to resolve these issues.

I hope we can resolve the issue soon and will do my best to provide any information that I can with the tools provided.

@logan i definitely agree that this issue is worse on hibiscus.
Anyways i think mechanically we wont be able to do much more to correct his issue with our deadline. I think since we have already put in place that the z bushing compression is set at calibration to be the point that it just holds the axis up, there is not much we can do here. Unless we go to where the bushing just adds enough drag to make the axis fall slower than freefall, but it still is able to fall under its own weight. I personally think that this should be strongly considered. The firmware prevents most cases of the z motors being off, and if there is power loss the z brake board will prevent the axis from falling.
Other than that we can tighten z belts a bit more, and ensure the frames are relatively square.
Im open to more ideas!

Marlin adds experimental automatic backlash compensation in software. So far our tests in R&D have shown that this improves the first layer quality by quite a bit.

The new code works by measuring backlash at each probe point and then taking the minimum reading as the common backlash which can be corrected for in software (typically one side of the printer will exhibit more backlash than the other... this cannot be corrected for in software). You can see the measured backlash value by running "G29 V1". The FW will only correct up to .5 mm in backlash, and attempts to measure it at a 0.005 mm resolution.

Occasionally, the electrical contact is broken before the nozzle has cleared the washer and there will be an abnormally low backlash measurement. In such instances, the first layer will be uneven. This does not happen very often, however, but it does happen.

kent added a comment.May 1 2018, 2:10 PM

@marcio what is the logic behind using the minimum reading as the common backlash? You say that there is a tendency for backlash measurements to come out lower than actual, so it seems like we should not use the lowest of the readings because they might be inaccurate.

marcio added a comment.May 1 2018, 2:45 PM

@kent: The reason is that the same backlash correction is applied to all points (we don't have independent Z motors). So, if I add the maximum backlash, then the points with the least backlash will now have backlash in the opposite direction introduced by the backlash compensation code. I was trying to avoid that.

But in truth, it probably doesn't matter that much. We could probably pick an average, which would mean that at none of the four points backlash would be totally corrected for, but the difference would be split among the points.

marcio added a comment.May 1 2018, 2:59 PM

@kent: I'll try it with a maxium to see what effect it has.

tutley added a comment.May 9 2018, 2:28 PM

@logan @kent So far from what i can tell the new firmware has made a huge improvement in the first layer quality.

It seems we are going with firmware for release. That is unless something comes up.

Can we close this as resolved?

MikeR added a comment.May 9 2018, 2:36 PM

@tutley we may want to leave it open, we have seen an improvement in the first layer but calibration is still seeing some weird layer lines in the calibration octopus that could maybe be firmware related.

kent closed this task as Resolved.May 9 2018, 3:02 PM
kent claimed this task.

@MikeR @tutley since this ticket was originally created to address the issue of inconsistent first layer bead height, and that issue's root cause (lash) has been identified, addressed, and confirmed as fixed (the first layer squish on a printer with lash compensation is much more consistent than on one without), I think we can close this ticket. @MikeR if we are still seeing weird layer lines in the calibration octopus, we can open another ticket to address that issue separately. I'm with @tutley; I think we can close this as resolved. If inconsistent skirt height becomes an issue again, we can open it back up.

oliver reopened this task as Open.Jul 18 2018, 3:35 PM
oliver added a subscriber: oliver.

I have been unable to have the lash compensation completely fix the inconsistent bead height of the first layer. I went ahead and took the values of a several probe runs using the G29 V4 command and took the first 3 and found an equation for the plane. The numbers were fairly consistently showing that the front left corner being the highest and the back right being the lowest.

The graph is a a plane plotted from the average over several probe runs that I ran. I believe Eric has pointed out before but I think the cables or the chain may be propping up that corner just enough that the auto bed level may not be able to compensate for.

kent renamed this task from Inconsistent skirt measurements LOG HERE to Inconsistent skirt measurements.Jul 23 2018, 10:23 AM
logan added a comment.Jul 24 2018, 7:39 AM

I can confirm @oliver's finding that backlash compensation doesn't completely fix the issue. This is confirmed by a customer in the forum thread posted by @kent above also.

I have test results that indicate our bed plates are being distorted by the Y-Bearing holders. The test is time consuming to perform, so this has only been done on one machine so far.

Bearing typeWith CompensationWithout
Stock Mini 2.406 L -.498 R = .092.398 - .510 = .112
Press fit.344 - .405 = .061.370 - .480 = .110

The skirts were made without any Z-offset adjustments in between. My regret is not obtaining bed leveling data from this machine, the main goal was to see if press fit Y-bearings on the Y-axis would result in a more even skirt with compensation. It did in this case. I think if this test were repeated on more machines we would see flatter bed plates.
This is happening because of the reaming process, which hasn't proven to be viable for the X/Y Bearing holders. It works well for Z-axis because we've allowed for some misalignment in the bearings. The X-ends it is actually necessary I think because of how often the part warps off the bed in the front so this allows us to use more of those parts than previously. However it is causing a lot of problems throughout production with the X/Y axes and that fallout needs reduced. The two most common red tags are bearings too tight (when fully fastened into assembly), and inconsistent bead height.
I think a simple return to press fit X/Y with a controlled broaching process for those that are too tight individually would result in far less assembly fallout and part fallout being quite reasonable. One has to remember that at the time press fit bearings were used, the only reason the part would fall out was due to the bearings being too tight which can now be corrected by cluster QC techs using a controlled broaching process.

logan raised the priority of this task from Normal to High.Jul 24 2018, 7:40 AM
kent added a comment.Aug 6 2018, 12:15 PM

We just had a machine come in on RMA reference number 16921 and helpdesk ticket reference 174154 reporting this issue.

I have a red tag unit with this issue, despite being unable to locate any mechanical issues and running the calibration gcode with backlash compensation enabled.
I checked the bottom of the frame with a straight edge to ensure that it isn't bent.
I have checked Z-pulleys and belt tensions
I have broached the left Z-axis (heavy side) even though it didn't seem any tighter than the right side.

Anyone else have any ideas?

a few things to check:

bloom on the y ends could cause the bed to sit crooked
also bloom on the y bearing holders

possible slight bend or twist in y rods causing weird probe data

check the y mount flanges are parallel to the frame

maybe a bent bed plate

i will post more if i think of anything else

@tutley I will try a new set of Y-bearings and see if I get an improvement and check rods/mounts while I'm at it. Thanks, I will keep y'all posted

logan added a comment.EditedAug 21 2018, 1:12 PM

@tutley I checked:
rods are straight
all four corners of Y-axis mount flanges rise to same height
no bloom on Y-axis Idler or Rod mount inserts

I found the left bearing holder was reamed considerably crooked, replaced it. Right side was pretty much perfect.

No improvement, the skirt is still visibly heavy on the left and higher on the right.

@logan if you put a straight edge off the bed to the z lowers is it higher up on the z lower on left side than on the right?
@kent you had a machine that was doing something similar. did you ever figure out a fix?

oliver added a comment.EditedAug 21 2018, 1:28 PM

@tutley The inconsistent skirts have been a constant issue for awhile now on a lot of machines, usually we can get them to be with in tolerance but one side will still be squished and the other side will be high. It is also always been the same corner that is too squished and the same corner that is too high.

To clarify the back lash has helped to get a lot more to pass but it does not fix the entire problem.

logan added a comment.Aug 21 2018, 1:56 PM

@tutley Yes, by about 1mm, but I am not sure why, everything meets spec as far as I have been able to tell by inspection of parts
The customer's machine noted above we did not manage to fix, we just waited for a good one to come off the line. And by good one I mean actually good, not just within spec. The machine sent back to the customer was very close all around, under 50 microns difference between measurements if I remember correctly.

Whatever is causing this is consistent at least, I personally have not seen an inconsistent skirt that was heavy on the right and light on the left, its always the same side

@logan yea i am not sure what is going on with this but at least it is consistent. should be easier to pin down

logan added a comment.Aug 21 2018, 2:27 PM

@tutley I tried leveling the X-axis a mm higher on the left, I don't see much if any improvement

oliver added a comment.EditedSep 28 2018, 3:18 PM

Want to reference this post T3779 by Kent in quiver for possible causes of this issue.

Also I spent a bit of time while I fixed red tags to investigate some of the issues. The printer referenced in the aforementioned ticket will be first as it came in for a uneven skirt.

Z Lower Left4mm
Z Lower Right5.5 mm

The next measurement was taking from the top of the flange for the y idler and mount to the top of the y rods.

Y Motor Mount Left16.11mm
Y Motor Mount Right16.07 mm
Y idler Left16.06 mm
Y Idler Right15.93 mm

I tried swapping out the Y idler with one that showed to be closer but it had no discernible effect on the skirt or the z lowers.
The numbers from the probe as well;

S1 = -0.366mmS4 = 0.223mm
S2 = 0.00S3 = 0.348mm

This printer had a distance from the 4th point of 0.24mm and a Z back lash of 0.18mm.

As far as I could tell the angled bed had no discernable effect on the skirt height. Which the next printer help to further confirm this.

oliver added a comment.EditedOct 1 2018, 7:49 AM

For this next printer was red tagged for failing the quick circle test but I also noticed that the bed was non-straight.

Z-Lower Left4mm
Z-Lower Right6mm

The Y idler and motor mount measurements were as followed;

Y Idler Left = 15.96mmY Idler Right = 15.82mm
Y Motor Mount Left = 15.74mmY Motor Mount Right = 15.81mm

Z Probe Data

S1 = 1.145S4 = 0.211
S2 = 2.106S3 = 0.471
Planarity = 0.70Z backlash = 0.23mm
oliver added a comment.EditedOct 1 2018, 9:15 AM

Hypothesis: Is the uneven skirt a symptom of issues with the x-axis instead of the slanted bed?
I did 2 tests one with a shim for leveling and 2 without, the final had a new x-bearing.

2mm ShimMin= 9.83mmMax= 10.22mm
No ShimMin=9.79mmMax= 10.21mm
New X-BearingMin= 9.93mmMax =10.13mm

There was no noticeable difference until I swapped out the x-carriage so this was another case proving that the x-carriage was not helping.

oliver added a comment.Oct 1 2018, 9:18 AM

There was no noticeable difference until I swapped out the x-carriage so this was another case proving that the x-carriage was not helping.
Following after I did several skirts at different z-offsets trying to get the skirt to be within tolerance. Surprisingly I did not see it failing too high on one side and too squish on the other. And it wasn't until I reflashed the firmware that I started to see z-offset affecting the skirt. This is same problem as T3304.

oliver added a comment.Oct 1 2018, 9:18 AM

As a final solution to try and get it within tolerance, I disassembled the Y and looked at bearings for bloom, did not see any bloom but I did find out the double bearing was not reamed straight and this was still causing issues with the Y-axis at times.

oliver added a comment.Oct 1 2018, 9:19 AM

Y-Axis Single

Top = 15.23mmBottom = 15.13mm

Y -Axis Double

Top = 15.07mmBottom = 15.43mm

New Y-Axis Double

Top =14.98mmBottom = 15.13mm

This had a affect on where the y rods were sitting as well.

Y Idler Left = 16.06mmY Idler Right = 15.96mm
Y Motor Mount Left = 15.92mmY Motor Mount Right = 15.13mm
oliver added a comment.Oct 1 2018, 9:19 AM

Z Probe Data

S1 = -0.160S4 = -1.335
S2 = 0.281S3= -1.555
Planarity = 0.66Z backlash = 0.25mm

E-steps were at 420 and it extruded 98.5 out of a 100mm. The skirt ranged from 0.377 to 0.481 just barely in tolerance now.

It's my theory that the uneven bed is increasing the amount of inconsistency in the probing.

logan added a comment.Oct 1 2018, 12:28 PM

The skirt ranged from 0.377 to 0.481 just barely in tolerance now.

That is a difference of 104 microns and is well within tolerance. It would have been well within tolerance even before the specification was expanded for Mini 2.

Is the uneven bed causing any issues with a tight x-carriage?

What do you mean by that? The drag of the X-carriage is irrelevant to the height of the skirt and I fail to see how the two could be related...

I am not seeing any helpful conclusions in the comments posted above.

Back on track; we know that this issue is caused by hysteresis in the Z-axis. Some of that hysteresis is inherent in a belt driven Z-axis (belts stretch), the rest is the result of parts or assembly out of spec causing an increased drag in the Z-axis.
When resolving red tags for this issue, we must stay focused on reducing the hysteresis/drag of the Z-Axis.

logan added a comment.Oct 1 2018, 12:28 PM

I personally resolved two machines with this issue today by disassembling the X-Axis and re-assembling according to the OHAI.
The process went as follows:

  1. Remove X-carriage cap/tool head
  2. Remove X-Axis drive belt
  3. Unplug both Z-motors/push to bottom of travel
  4. Loosen X-Axis smooth rod set screws
  5. Reset X-axis smooth rods ensuring no outward or inward tension is applied in the process
  6. Re-torque X-Axis smooth rod set screws
  7. Retension X-Axis drive belt and complete reassembly
  8. Run skirt and measure

Both instances I was able to correct the results as follows:

1 before.30.51.210
1 after.38.475.095
2 before.30.50.200
2 after.347.467.120

I meant to say is there any correlation between a tight x-carriage and inconsistent skirt. A tight x-carriage can arise from a multitude of issues such as a bad ream of the x-carriage.

Also none that explains why there is quite a significant slant of the bed or where that slant is coming from. I had presumed that the slant was causing an inconsistent bed skirt which could still very much be true as I see nothing in your numbers that would discredit it.

logan added a comment.Oct 1 2018, 1:10 PM

I meant to say is there any correlation between a tight x-carriage and inconsistent skirt. A tight x-carriage can arise from a multitude of issues such as a bad ream of the x-carriage.

Again, what does X-carriage drag have to do with skirt height? How could excessive drag in the X-carriage affect the height of the first layer?

You still failed to determine the cause of the bed slant, correct?
Both machines above have an obvious visible slant to the bed YET after reducing the drag of the Z-axis, the skirts are well within spec, as shown. It is true that the more error is present in the bed itself, the more Marlin will have to compensate for, this involves up/down directional changes of the Z-axis as the print head moves around the perimeter of the print surface. The inconsistency being shown is the result of hysteresis, the Z-axis is not traveling as much as it thinks it is.

oliver added a comment.Oct 1 2018, 2:09 PM

Perhaps I'm still not understanding what your trying to say. Why is there a slant in the y-axis? What in that stack of tolerance of the y-axis is adding up enough to overcome what marlin can handle?

I did not notice quite a significant change in the probe values when I replaced the Y-Double bearing with one that wasn't so misaligned. Just trying to gather data and help investigate a problem that is cropping up in red tag printers.

kent removed kent as the assignee of this task.Apr 1 2019, 3:21 PM
anolen removed a subscriber: anolen.Jun 24 2019, 11:12 AM

Update: This is still an issue that is occurring in calibration and is hampering the number of printers that they can get calibrated in a day. The calibrators are printing multiple calibration prints because they will dial in the z offset using the skirt and go to try and print the final calibration print, only for it to come out being too squished.

@marcio i know this was brought up before from @kent.
i think having the ability to store probe data and use it instead of running through the auto leveling every print may help this issue

logan added a comment.Jul 15 2019, 1:13 PM

Just for the record, this has also been reported with Redgum T8647 but not on Quiver to my knowledge.

@logan @tutley

If this has been brought up with Redgum, could it be due to some inconsistency with the toolhead? Since they both use the same toolhead.

well quiver technically uses the same toolhead as well. its still a titan aero with the same wire connection for probing and stuff so i doubt it is related to the toolhead.
i think it is most likely due to the belt driven z and backlash. I just think quiver may be applying the lash correction better
on mini 2 and redgum the lash is measured off all 4 washers and averaged, on quiver the lash is measured off the cube.
This may be why we are seeing a difference between mini2 and redgum vs quiver

logan added a comment.Jul 15 2019, 1:37 PM

@tutley that sounds reasonable to me. Redgum takes backlash values off of the cube also but it isn't activated for the calibration print. That may be a solution for Redgum.

The root of the issue for Hibiscus I think is that our bed plates aren't sitting level physically. We've had some machines recently that of you bottom out the Z Axis against the lowers, the nozzle is able to touch the left side of the bed but not the right. So while it may know that the bed isn't level, it is outside of the range of motion and therefore can't correct for it.
I believe we've also established previously that there are ranges in which the bed can be out of level that Marlin cant effectively compensate for.

I think we should take a look at why our beds aren't physically level and how we can correct that.

tomc removed a subscriber: tomc.Jul 15 2019, 2:10 PM

I am currently testing out a theory that I have had for awhile and that is that the cable chain is causing enough of a upward force on the bed plate. As far as I have seen its been consistently squished on the left side where the cable chain is pushing up on the bed plate.

@oliver the heater pad wires are on the left as well, there should be clearance, but maybe some units are slightly crushing the wires/silicon pad

@tutley There is still inconsistency between sides but removing the cable chain did seem to show a noticeable shift in the z offset.

Either this is showing that the cable chain does have an effect on the level of the bed or each print is unrepeatable.

@oliver that is inconclusive because it shifted by the same amount on both sides. If it were that the cable chain was causing enough of an upward force to impact the squish, you would see more squished on the left and less squished on the right. Then removing the cable chain you would see left and right ~equal

Talking to the calibrators, they also brought up the fact the x end motor side has a lot more weight that will be pushing downwards on the axis more than the x-end idler side. This could also be a potential cause into why the bed appears to be not level.

logan added a comment.Jul 15 2019, 3:11 PM

@oliver That does nothing to explain why the nozzle can touch the bed on the left and not on the right with the X-Ends bottomed out on the Z-Lowers. Try to answer the question "How level is the bed?" and then you might be heading on the right track to "Why is the bed not level?"
Take a Mini 2 to the granite in QC and use the height gauge to measure the height of the bed washers

I'm seeing this problem of uneven skirts on red tag printers we are trying to correct. The calibrators are also have major problems with getting a skirt to print in spec, due to the skirt being uneven. I have tried remounting the toolhead on the machine I am currently working with, as well as lots of z-offset adjustments, with strange behavior. At a z-offset of -1.20 the high side of my measurement is a .51 mm, which is out of spec, with the low end being around .36. Then when I adjust the z-offset from -1.20 to -1.21 suddenly my values on the high side go from .51mm to .38mm with the low end being around .26mm/ That is with just a .01mm change on the z-offset, which should not cause that drastic of a layer change.

@bigmansas Can you post the gcode file you were using to test?

@karrad Sure, it's straight off the SD card that goes out with the machines.

This comment was removed by karrad.

Opened wrong gcode, file, my mistake. Ignore previous comment

I have a machine here that I can not get within spec on fw version but was able to get it in spec with fw version The other interesting thing was when I flashed back and fourth between the 2 firmware versions I noticed that I needed a different z offset or else my print would come out squished.

@marcio Was there anything changed in marlin that could affect how the printer performs its autocal or utilizes its z offset between the 2 editions.

marcio added a comment.Aug 1 2019, 8:28 AM

@marcio Was there anything changed in marlin that could affect how the printer performs its autocal or utilizes its z offset between the 2 editions.

I cannot know this for sure. Marlin is always in a state of flux. When comparing two versions so far apart, it is really hard to get a handle of what may or may not have changed.

oliver closed this task as Spite.Oct 10 2019, 1:09 PM
oliver claimed this task.

Could not figure this out.