New v3 Bed Leveling/Calibration Problems

[ToddZilla]

I started asking about this in the Facebook group a few days ago, but now that my forum account is activated I'm moving the discussion here.

I finished my V3 on Monday and have been trying to work through problems with getting a level bed. I've run through all the calibration/setup advice in mhackney's build thread as well as advice he and others provided on Facebook, but still haven't tracked it down. I modified calibration points in the EEPROM, tried removing 3 bed mounts (since reinstalled), checked that my towers are square, adjusted end stops, etc. I've done my best to check for warpage in the plastic bed tray, bed heater, and glass plate but haven't found anything obvious. Using a micrometer, I checked the thickness of the glass around its circumference and found a maximum variation of around 0.08 mm. This evening I've been trying to adjust the height of the bed supports to eliminate high & low spots I'm seeing in my calibration ring print, with mixed success (just keep moving the high/low spots around). I tried rotating the glass 120 deg, but again that just moved the high/low spots elsewhere.

My most recent attempt is pictured here:

I marked the print near each tower and used my micrometer to measure the thickness. There's about a 0.1 mm difference between the thickest part (0.4 mm near the X tower) and thinnest area I measured (0.3 mm near the Z tower). I couldn't even attempt to measure the thinnest part of the print since it was too thin to even get off the glass as anything other than scrapings. Next I think I'm going to try printing a spiral or a series of concentric rings and see how things change towards the center of the plate.

I'd greatly appreciate any additional ideas/advice to get this sorted out. Thanks!

[ToddZilla]

Today I attempted a manual calibration using a feeler gage (as described in one of the appendices of the V2 User Manual) to adjust the end stops and manually edited the horizontal radius value in the EEPROM. I also added a few check points mid-way between each tower at the same radius (135 mm). The Z0 position at each tower and in the center were pretty good according to the feeler gage, but the nozzle was a bit high at each of the midpoints. I went ahead and printed another 1 layer calibration circle and the results aren't too surprising. Things look pretty good near each tower (especially X and Z), but the nozzle is way too high between the towers (see photos).

At this point I'm not sure what to do next. I just started looking at RollieRowland's Delta Auto-Cal Tool, and may give that a try next.

Again--appreciate any ideas thrown my way. Thanks!

[mhackney]

You've checked your towers to make sure they are not leaning? N

The effect you are seeing reminded me of something I read last year: http://forums.reprap.org/read.php?178,492382

David Crocker implemented the dc42 firmware on Duet that has the best delta auto calibration around (bar none). He knows his stuff. His observations (and I've validated this with my own machines) is that there is a relationship between motor steps/mm and diagonal rod length (your arms). The observation is that belt stretch or out of tolerance, etc can produce non planar movement like this. David asserts that it is actually the product of diagonal rod length and motor steps/mm that matter. So it you tweak your rod length a but, recalibrate and test and repeat you might be able to resolve this. I know one of my mini kossels has arms configured that are 4 mm shorter than actual and it prints dead flat now. Read David's posts and give it a try.

[ToddZilla]

I did check for tower lean--as near as I can tell using a precision steel ruler, they're all within 1 mm bottom to top (though there's probably a better way to measure it). I'll go check out the post in the reprap forum. Thanks!

[ToddZilla]

Okay, after another 4-5 hours, I'm still stuck in almost the same place. Based on the Reprap forums thread referenced above, I adjusted my manual calibration points to 2/3 of the build radius (assuming 140 mm max build radius, that gives 93 1/3 mm). I repeated my manual calibration and horizontal radius adjustments in EEPROM until all 4 points were equal using a feeler gage. After that I started changing the diagonal radius setting until the midpoints between each tower matched the other 4 points per the feeler gage. That's 7 points that seemed to check good as far as z height. I then attempted to print another calibration donut (see photos).

I measured the thickness of the print at 5 points nearest each tower and between the towers (except XY, which was garbage) and came up with the following: X - 0.010 in, Y - 0.010 in, Z - 0.010 in, XZ - 0.015 in, YZ - 0.020 in. Again, XY was trashed and unmeasurable.

At this point I'm going to pull the bed components again and check for warping. All the bed mounts are tight except the one closest to the Z axis on the X-Z side, which I loosened/lowered to correct the thin prints I was getting in that area earlier. After adjusting diagonal radius, that area is thicker than at the towers, but still thinner than the areas between X-Y and Y-Z. If I understand what I read on the Reprap forum, the diagonal radius should affect all 3 mid-tower regions equally, but that's not exactly what I'm seeing.

[alvingm]

I have the same exact problem as yours and already did exactly most of the troubleshooting you've done and still getting the same result when printing that donut

Anyways, will watch this thread and will also continue troubleshooting on my side as well

[ToddZilla]

I haven't had much time to work on my printer this week, but I just noticed there were changes posted earlier today to the dev branch of the firmware on GitHub. The changes specifically deal with/affect calibration, including configuration.h parameters DELTA_DIAGONAL_ROD, CARRIAGE_HORIZONTAL_OFFSET, and Z_PROBE_SENSITIVITY. There's also added code to average the measured double-tap distances if they're within tolerance of one another (assuming I understand the code correctly).

I'm thinking of compiling and installing the dev firmware to see if it improves the problems I'm having.

[mhackney]

I posted in the other thread about this an hour and a half ago. I'm getting excellent results with the fix. It was related to accelerometer sensitivity. So do grab it and test it. You MUST clear the EEPROM with the special clear_eeprom sketch first.

[ToddZilla]

I also just noticed that all six of my bed supports are upside down (I installed them the direction shown in the original picture in the Dozuki--didn't notice the text instruction stating the picture was incorrect). Since that's now been fixed in the instructions, I went ahead and flipped mine all over and am removing the washers (otherwise the clamps don't hold the bed firmly at all). Next up is the new firmware...

[ToddZilla]

Actually, now that I've removed all the washers, the bed clamps still don't hold the glass firmly in place--might be because they've now developed a curve from sitting up over the higher edge of the glass the past 2 weeks. Hopefully a little manual bending will fix that, if not I guess I'll hit SeeMeCNC up for a replacement set.

[mhackney]

That's what happened to mine too. I am heating them up in hot water and reshaping them.

[ToddZilla]

After flipping all my bed supports, bending my bed clamps, and installing the dev firmware things have improved, but I still can't get a truly level bed out to max print radius (see photos). The areas between the YZ and XY towers are still high (more so on the YZ side) and I have low spots on the XZ side near the Z tower and on both sides of the X tower. I tried adjusting the bed supports between the two different prints shown in the photos, but after flipping the bed supports I can only loosen the high areas so much before the clamps completely stop gripping the glass plate. I'm not worried about the stringing on partdaddy's calibration print, as I'm still using 210/60 temps for PLA, which I know are on the high side. I'm just trying to get a calibration/level that works across the whole bed surface.

When I get some time, I'll try manual calibration again and post the results. Once my PhD is done, I'm seriously considering trying to implement a hybrid Repetier/Octoprint version of the 7 factor auto-cal that's present in the dc42 firmware using the accelerometer (but that will take at least a few more months). In the meantime, maybe the guys at SeeMeCNC wouldn't mind me dropping by with my printer--it's only a 4 hr drive...

[mhackney]

ToddZilla, firstly congrats on getting this far with your Ph.D. What is your thesis? That last push is tough but well worth it.

Your first photo actually looks quite good as far as layer height. I think for any real world print - except maybe a 12.5" disk - this calibration would do an excellent job. Interestingly, my attempts with the new firmware also gave me a low area exactly where yours is. I started manually tweaking the Z Max by adding .03mm to it and testing. After a couple of additions I am now able to get a good calibration print across the entire bed. Perhaps a little thin between X and Z but we are talking about .05mm variations which on a disk this size and machine like this is excellent. Now that I've done the manual offset I can use that to calculate the Z Probe Height in EEPROM. By default it is set to -0.200. I forgot to write mine down (I had calibrated it before) before zeroing out the EEPROM for the new firmware. But it is easy enough to calculate. After you do your initial calibration, write down Z Max length. Then as you tweak and get to the correct Z height, simply subtract the calibrated Z Max Length from the new one:

Z max tweaked - Z max original calibration = delta Z

Then ADD this value to the Z-probe height parameter, which is -0.20 in the default firmware:

Z-probe height + delta Z = new Z-probe height

Store the new calculated Z-probe height in EEPROM and run a test calibration. Double check your Z Max Height to make sure it makes sense. Test print and maybe tweak a bit if needed.

Here are my results:

Calibrated Z Max: 402.219 mm
Z Max after tweaking: 402.410 mm

delta Z = 402.410 - 402.219 = 0.191 mm

Original Z-probe height: -0.20

New Z-probe height = -0.20 + 0.191 = 0.009mm - which is "0", .001 mm is meaningless precision on these printers. I round everything to the hundredths (.01) place for final setting.

So my Z-probe height is now set to 0.00 mm. This makes sense in my case because of the way I currently have my bed mounted. There is essentially no flex in it at all. The Z-probe height is basically a fudge factor in this case to compensate for the deflection due to probe impact. Since the nozzle actually is the point of impact, the Z-probe height should be 0 in theory. In practice, any mechanical movement (flex, belt stretch, etc) will result in an artificially long Z max. The Z-probe height can be used to remove that error, which is why it is set to -0.20 mm in the default firmware. But I think if you do a good job tightening your bed mounts, you should be able to reduce the amount of flex.

[ToddZilla]

mhackney, my dissertation is on the interaction of ultrasonic Lamb waves with fatigue cracks in aluminum plates--related to structural health monitoring of aerospace vehicles. I'm supposed to be done by the end of the year, but many days I swear I'll never make it; feels like 90% done and 90% left to go...

I'm pretty happy with the improvement after trying the new firmware (at least now the center of the print bed is very usable, where before I had a big height difference between the edge & the center), but there are obviously still a few problem areas. I'm still relatively new to 3D printing, but I'm more concerned about the low spots in my bed level/calibration (nozzle too high so the filament doesn't even stick) that, while improved, I'm still seeing between the Y & Z towers (3rd photo in my last post). There's another one between X & Y, but it's definitely improved with the new firmware.

I'll give your technique of adjusting z-probe height a go next time I have a few hours to spend on the printer. Thanks!

[mhackney]

Cool. Although my Ph.D. is in inorganic chemistry I spent the early part of my career fabricating high tech ceramic materials. We had a huge grant for the non-destructive evaluation of these materials. You can imagine what the application was, I am not allowed to talk about it for another 12 years.

So, tweak your Z a little like I described, you should be able to find a happy sweet spot where you can stick across the entire bed.

[cloneit3d]

ToddZilla, firstly congrats on getting this far with your Ph.D. What is your thesis? That last push is tough but well worth it.


Store the new calculated Z-probe height in EEPROM and run a test calibration. Double check your Z Max Height to make sure it makes sense. Test print and maybe tweak a bit if needed.

Please let me know how to store the Z-probe height or offset in the EEPROM. I know I can edit the values in configuration.h and then flash the EEPROM again however there must be an easier way to do it. I am mostly looking for the Z-offset so I can more efficiently set my first layer height and I know it can be done in Matter Control if you user the Matter Slicer however I prefer to use CURA.

Thank you-

Tim

[ToddZilla]

cloneit3d: I'm using the EEPROM Repetier Editor Plugin in OctoPrint to make changes to the EEPROM settings without having to reflash.

[morgandc]

Another option is to use MatterControl and edit the EEProm with a USB cable.

[mhackney]

You can use mattercontrol to edit the eeprom and not use it to slice. You can also easily change the default slicer in MC to Cura if you want to.

[cloneit3d]

cloneit3d: I'm using the EEPROM Repetier Editor Plugin in OctoPrint to make changes to the EEPROM settings without having to reflash.

Ok/ I am using Octoprint- where can I get the plugin?

Thank you-

[cloneit3d]

You can use mattercontrol to edit the eeprom and not use it to slice. You can also easily change the default slicer in MC to Cura if you want to.

Ah ok- another great option! Many thanks all for the ideas, much appreciated!

[Turpinator]

You can also use repetier host which allows for curaengine slicer, and does have the ability to edit eeprom with a gui like in matter control.

Now, I haven't used octoprint at all (well, I tried it once for like 2 mins), but I use repetier host and server, it still allows you to print remotely, whether you use the web interface, or the full on repetier host (to connect to repetier server on your pi/whatever). As with octo print, the setup is pretty simple, so if you've got some time, you can compare features and possibly make the switch.

[briankb]

Where did you get those calibration files? I've searched thingiverse and yeggi and can't seem to find them. I created one in spaceclaim engineer then in openscad but both render with segments within Simplify3D and nothing as smooth as the one you originally posted.

Code: Select all

difference(){
    cylinder(h=0.25,d=275,center=true,$fn=100);
    cylinder(h=1,d=245,center=true,$fn=100);
};

Can you share the link of the files for the STL files you are using to calibrate and fine tune your Max?

UPDATE:
Got a reply from Simplify3D support within an hour, maybe less. Telling me to check the output resolution in SpaceClaim, and that was the issue. And as it happens it is the same in OpenSCAD so the code above was updated to add "$fn=100"

I also found one of the calibration files on repables via another forum post by partdaddy which is http://repables.com/r/800/ still looking for the donut one but it is easly generated in CAD or that above openscad code and exported to STL.

[mhackney]

If you are asking about the big donut one that was printed in red in the first post in this thread, I posted the STL for that in my V3 build thread here.