Auto Calibration / Auto Bed Leveling million questions - Paging DC42
Posted by clearlynotstefan
|
Auto Calibration / Auto Bed Leveling million questions - Paging DC42 December 09, 2016 07:02PM |
Registered: 7 years ago Posts: 154 |
Hey helpful people I'm posting one thread with all of my questions on the various software calibration options we have, because frankly I don't really get them and their relationships to each other yet. This is a rambling post of many questions, any help on any of them would be appreciated.
Background: 2040 based delta with aluminum vertexes. I am very confident that my horizontals and verticals are the exact length, exactly parallel and assembled very study. I have the DC IR probe, and use magnetic rods that are sturdy and within .01mm of eachother in length. My carriages ride smoothly on the v-slot and don't wobble on any dimension. I'm using a duet wifi with 1.17beta (with bed leveling). I can't rule out some degree of effector tilt across the x-y plane and have yet to set up H parameters for the bed.g file.
Now a series of random questions so I can better understand the calibration options we have available due to DC42's tireless efforts to make these printers awesome.
How does bed.g work exactly and to what degree do various design shortcomings affect the results. For instance, I have reliable endstops mounted exactly evenly at the top of my verticals. I am getting some degree of "endstop adjustment" compensation when running bed.g. That suggests that the probe is detecting different heights at different points, which could be caused by an unlevel bed, or alternatively by effector tilt skewing the results of probing. What am I expected to do with the endstop adjustment number? Should I be actually moving my endstops to get 0? Should I assume that those numbers actually represent effector til or a not level bed? When I run bed.g then lower the nozzle to an effective X0 Y0 Z0 (sometimes needing adjusting on the z and a g92 Z to set that as the height) I get bed crashing at certain extremes on the xy axes. That suggests I should use H parameters in my bed.g to compensate for effector tilt right?
Now we have bed leveling on 1.17. With that I have more questions still. I'm familiar with the theory from having used it on my marlin cartesian, but I don't understand the relationship of 29 to bed.g now.
Should I use one or the other, or both?
Does G28 rely on information from an accurate bed.g? It seems like H would be ignored because its essentially offsetting effector tilt by probing many places. In that sense, how is bed.g with accurate H parameters any different from just using grid leveling?
What actions change or overwrite the matrix calculated by G28? If I run bed.g, then G28 does it overwrite the auto-calibration? If I run G28 then bed.g does it overwrite the grid data?
If after doing either or both of the above, I choose to lower my head to an acceptable height and use G92 Z - does it overwrite the grid? or does it set that as the new Z0 and continue compensating on the XY?
Yesterday I ran a g28 grid and did a larger print. My first layer was PERFECT, but as the height grew it turned out the entire print was being skewed heavily in one direction, a giant slant! Why did this happen? Did effector tilt skew the results of the G28? Effector tilt obviously matters more to the probe than it does to the hotend (assuming a centered hotend), thus unless G28 is using H parameters it could overcompensate by adjusting for tilt at the probe not at the nozzle, right?
Basically, you're me: Printer is built like a tank, stable endstops, accurate dimensions, IR probe. Assume some effector tilt is real. Currently, running 6 factor bed.g returns numbers inconsistently but always low. Your prints don't stick well with bed crashing close to the x tower. What do you do? I can't really figure out where effector tilt is coming from, but evidently it exists at least in a small way (I can't manually jiggle my effector on any plane). Additionally, you want your prints to be accurate and STRAIGHT, thus bed leveling beyond a tiny amount scares you because the skewing that could take place as the print grows taller with compensation. Whats your next move? Sort H parameters and scrap G28 altogether? Use both? Adjust endstops? Measure something or other?
I appreciate your help sincerely folks. Yesterday I printed a small test cube in the center of the bed (where far dimension level issues couldn't hurt it) without G28. It was my first print on this new monster and it came out perfectly, I was so happy! I just want to get the leveling and calibration stuff sorted so I can move on to bigger, accurate, and straight prints and I'm struggling conceptually with what bed.g and grid leveling actually do, and how I can adjust things physically to be less reliant on software calibration.
Edit: And Axis Compensation. Whats that about?
Edited 1 time(s). Last edit at 12/10/2016 01:46AM by clearlynotstefan.
Background: 2040 based delta with aluminum vertexes. I am very confident that my horizontals and verticals are the exact length, exactly parallel and assembled very study. I have the DC IR probe, and use magnetic rods that are sturdy and within .01mm of eachother in length. My carriages ride smoothly on the v-slot and don't wobble on any dimension. I'm using a duet wifi with 1.17beta (with bed leveling). I can't rule out some degree of effector tilt across the x-y plane and have yet to set up H parameters for the bed.g file.
Now a series of random questions so I can better understand the calibration options we have available due to DC42's tireless efforts to make these printers awesome.
How does bed.g work exactly and to what degree do various design shortcomings affect the results. For instance, I have reliable endstops mounted exactly evenly at the top of my verticals. I am getting some degree of "endstop adjustment" compensation when running bed.g. That suggests that the probe is detecting different heights at different points, which could be caused by an unlevel bed, or alternatively by effector tilt skewing the results of probing. What am I expected to do with the endstop adjustment number? Should I be actually moving my endstops to get 0? Should I assume that those numbers actually represent effector til or a not level bed? When I run bed.g then lower the nozzle to an effective X0 Y0 Z0 (sometimes needing adjusting on the z and a g92 Z to set that as the height) I get bed crashing at certain extremes on the xy axes. That suggests I should use H parameters in my bed.g to compensate for effector tilt right?
Now we have bed leveling on 1.17. With that I have more questions still. I'm familiar with the theory from having used it on my marlin cartesian, but I don't understand the relationship of 29 to bed.g now.
Should I use one or the other, or both?
Does G28 rely on information from an accurate bed.g? It seems like H would be ignored because its essentially offsetting effector tilt by probing many places. In that sense, how is bed.g with accurate H parameters any different from just using grid leveling?
What actions change or overwrite the matrix calculated by G28? If I run bed.g, then G28 does it overwrite the auto-calibration? If I run G28 then bed.g does it overwrite the grid data?
If after doing either or both of the above, I choose to lower my head to an acceptable height and use G92 Z - does it overwrite the grid? or does it set that as the new Z0 and continue compensating on the XY?
Yesterday I ran a g28 grid and did a larger print. My first layer was PERFECT, but as the height grew it turned out the entire print was being skewed heavily in one direction, a giant slant! Why did this happen? Did effector tilt skew the results of the G28? Effector tilt obviously matters more to the probe than it does to the hotend (assuming a centered hotend), thus unless G28 is using H parameters it could overcompensate by adjusting for tilt at the probe not at the nozzle, right?
Basically, you're me: Printer is built like a tank, stable endstops, accurate dimensions, IR probe. Assume some effector tilt is real. Currently, running 6 factor bed.g returns numbers inconsistently but always low. Your prints don't stick well with bed crashing close to the x tower. What do you do? I can't really figure out where effector tilt is coming from, but evidently it exists at least in a small way (I can't manually jiggle my effector on any plane). Additionally, you want your prints to be accurate and STRAIGHT, thus bed leveling beyond a tiny amount scares you because the skewing that could take place as the print grows taller with compensation. Whats your next move? Sort H parameters and scrap G28 altogether? Use both? Adjust endstops? Measure something or other?
I appreciate your help sincerely folks. Yesterday I printed a small test cube in the center of the bed (where far dimension level issues couldn't hurt it) without G28. It was my first print on this new monster and it came out perfectly, I was so happy! I just want to get the leveling and calibration stuff sorted so I can move on to bigger, accurate, and straight prints and I'm struggling conceptually with what bed.g and grid leveling actually do, and how I can adjust things physically to be less reliant on software calibration.
Edit: And Axis Compensation. Whats that about?
Edited 1 time(s). Last edit at 12/10/2016 01:46AM by clearlynotstefan.
|
Re: Auto Calibration / Auto Bed Leveling million questions - Paging DC42 December 10, 2016 05:38AM |
Registered: 10 years ago Posts: 14,672 |
A lot of questions! I have been meaning to write a detailed description of how to detect geometrical errors in a delta printer and how to use auto calibration, and now that grid bed compensation in RRF and some other matters are sorted I may finally get time to do it this weekend. In the mean time:
1. You do not need to physically adjust the endstop positions after calibration because the adjustment is done in firmware. Just try to get the endstop heights equal to within 1mm or better.
2. Auto calibration assumes no effector tilt, towers parallel to each other, and a flat bed. If you do have effector tilt, then you can use H parameters to compensate for the effect of effector tilt on Z probe trigger height. But it is better to eliminate effector tilt as far as possible, and the piece I write will cover the common causes.
3. The sole purpose of H parameters is to correct for the Z probe trigger height varying with XY position. In deltas using a Z probe that is offset from the nozzle such as the IR probe, varying effector tilt is a common reason for trigger height variation. Other systems such as FSRs under the bed support may also show trigger height variation, e.g. if you probe close to one of the FSRs then triggering may happen with lower pressure than if you probe at the bed centre.
4. Recent versions of RRF can calibrate for bed tilt as well as endstop positions, delta radius and tower position. This is done by specifying 8 factor calibration instead of 6, or 9 factor calibration instead of 7. However, unless you can probe at large radius (i.e. well outside the triangle formed by the towers), the effects of bed tilt and misplaced towers on the probed heights are indistinguishable. So I recommend that you use a carpenter's square or similar to get the bed perpendicular to the towers, then use 6 factor calibration. Or, if you know the extent of bed tilt, you can declare it in the M665 command.
5. If after calibration the head is too low everywhere, the fix is simple: reduce the Z probe trigger height in the G31 command in config.g. Likewise if the head is too high everywhere, increase the trigger height setting.
6. Your slanted print is probably caused by missed steps, if by slanted you mean that it doesn't go straight up with increasing layers, and that you have no axis compensation configured.
7. On delta printers, the main reason for using G29 is to compensate for residual errors after running auto calibration, to get a better first layer. it is most likely to be needed on large printers.
8. Auto calibration results are preserved until you power the printer down, or you override them by sending M665 and/or M666 commands manually. G28 and G29 do not affect them. However, if your endstop switches do not always trigger at precisely the same height, auto calibration will correct for the actual trigger points when it is run, so you are best not running G28 again.
9. G29 grid probing results are unaffected by G28 but are cleared when you run auto calibration. However, you can reload the previous grid results if you wish by running G29 S1 after auto calibration to load them from the heightmap.csv file, in firmware 1.17dev8 and later.
10. Axis compensation is primarily for Cartesian printers whose axes are not quite at right angles. It could also be used on deltas to compensate for a known bed tilt.
HTH David
Edited 6 time(s). Last edit at 12/10/2016 05:43AM by dc42.
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].
1. You do not need to physically adjust the endstop positions after calibration because the adjustment is done in firmware. Just try to get the endstop heights equal to within 1mm or better.
2. Auto calibration assumes no effector tilt, towers parallel to each other, and a flat bed. If you do have effector tilt, then you can use H parameters to compensate for the effect of effector tilt on Z probe trigger height. But it is better to eliminate effector tilt as far as possible, and the piece I write will cover the common causes.
3. The sole purpose of H parameters is to correct for the Z probe trigger height varying with XY position. In deltas using a Z probe that is offset from the nozzle such as the IR probe, varying effector tilt is a common reason for trigger height variation. Other systems such as FSRs under the bed support may also show trigger height variation, e.g. if you probe close to one of the FSRs then triggering may happen with lower pressure than if you probe at the bed centre.
4. Recent versions of RRF can calibrate for bed tilt as well as endstop positions, delta radius and tower position. This is done by specifying 8 factor calibration instead of 6, or 9 factor calibration instead of 7. However, unless you can probe at large radius (i.e. well outside the triangle formed by the towers), the effects of bed tilt and misplaced towers on the probed heights are indistinguishable. So I recommend that you use a carpenter's square or similar to get the bed perpendicular to the towers, then use 6 factor calibration. Or, if you know the extent of bed tilt, you can declare it in the M665 command.
5. If after calibration the head is too low everywhere, the fix is simple: reduce the Z probe trigger height in the G31 command in config.g. Likewise if the head is too high everywhere, increase the trigger height setting.
6. Your slanted print is probably caused by missed steps, if by slanted you mean that it doesn't go straight up with increasing layers, and that you have no axis compensation configured.
7. On delta printers, the main reason for using G29 is to compensate for residual errors after running auto calibration, to get a better first layer. it is most likely to be needed on large printers.
8. Auto calibration results are preserved until you power the printer down, or you override them by sending M665 and/or M666 commands manually. G28 and G29 do not affect them. However, if your endstop switches do not always trigger at precisely the same height, auto calibration will correct for the actual trigger points when it is run, so you are best not running G28 again.
9. G29 grid probing results are unaffected by G28 but are cleared when you run auto calibration. However, you can reload the previous grid results if you wish by running G29 S1 after auto calibration to load them from the heightmap.csv file, in firmware 1.17dev8 and later.
10. Axis compensation is primarily for Cartesian printers whose axes are not quite at right angles. It could also be used on deltas to compensate for a known bed tilt.
HTH David
Edited 6 time(s). Last edit at 12/10/2016 05:43AM by dc42.
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].
|
Re: Auto Calibration / Auto Bed Leveling million questions - Paging DC42 December 10, 2016 06:43AM |
Registered: 7 years ago Posts: 154 |
Thanks for going through that monster post! One point of clarification, the slanted print did actually have G29 compensation on it. It stuck like glue and was uniformly squished getting my hopes up! Then it moved over .2mm or so per layer. It makes sense, if g29 disregards H parameters it can't possibly be a usable option on machines with effector tilt, all it does is compensate for an actually level bed that it interprets as being out of level from effector tilt, and since effector tilt is a much bigger problem to a probe off-center than a hotend centered, it'll be a nightmare! Given that, I don't think I'm skipping steps. I'm greatly looking forward to your post on these topics, I know I have a well constructed machine and great firmware options to help but without having a real gameplan or understanding of the calibration script I always find myself lost as how to proceed. For instance, having appropriate H offsets in my bed.g I understand will mitigate the effect of effector tilt for purposes of bed.g calibration, so say my delta radius returns more accurate. Does whatever hokus pokus happens following a bed.g then affect my print, I have no idea!? And so I'm stuck guessing and trial and erroring my way through things. It stands to reason that if my bed.g is factoring in a slant via H calibration for effector tilt, that all movements after that would also compensate for that tilt, but even with good H parameters I get a perfect .01 at 0,0, but scrape my IR probe further out on a plane. G29 did me no good as it interpreted the tilt in the effector as tilt of the bed and sent my layers off sideways to try to make it right haha. I'm convinced my problem is now either effector tilt or an actually crooked bed (its think borosillicate and braced on the horizontals so it should be perfectly flat relative to the rest of the printer. If its effector tilt I'm at a loss for where to look, I can't jiggle the effector, my rods are all within .01 of each other rod and move smoothly, my rods are spaced the same on both ends and across the printer etc.
Once again thanks for all the help, I look forward to your article, I think a little clarity on the process behind G28 and bed.g will help a lot of duet newbies like me!
Once again thanks for all the help, I look forward to your article, I think a little clarity on the process behind G28 and bed.g will help a lot of duet newbies like me!
|
Re: Auto Calibration / Auto Bed Leveling million questions - Paging DC42 December 12, 2016 12:32AM |
Registered: 8 years ago Posts: 155 |
|
Re: Auto Calibration / Auto Bed Leveling million questions - Paging DC42 December 12, 2016 05:17AM |
Registered: 7 years ago Posts: 154 |
|
Re: Auto Calibration / Auto Bed Leveling million questions - Paging DC42 December 12, 2016 05:26AM |
Registered: 7 years ago Posts: 154 |
DC, I'll add my next question here because its relevant. Clearly I'm having difficulty calibrating a printer with a great frame but something off effector wise. I have a bullseye level coming tomorrow to see if I can detect effector tilt, I'm running on the assumption that effector tilt is whats causing my calibration issues due to the nozzle offset. I'm confident my rods are solid (magrogs by Haydyn) but my effector itself might be skewed. Is it possible to manually calibrate this thing without a probe? ie home printer, jog till the nozzle is right in front of a tower, feeler gauge it till its perfect, set that endstop offset etc. I'm in a bit of a catch 22, until I can print a better effector, I can't calibrate to print a better effector haha. i'll try the H parameter again and auto-calibrate with the tilt compensation, but I didn't get far with that on my first try.
|
Re: Auto Calibration / Auto Bed Leveling million questions - Paging DC42 December 12, 2016 06:57AM |
Registered: 7 years ago Posts: 154 |
JESUS IT WAS THE EFFECTOR. I printed a godawful version of it (skipped steps, nozzle dragging etc, eventually it stabilized around layer 8, pushed the ball studs straight through the bottom few layers, figuring this is as good as it'll get). Calibrated 6 factor, returned 0.08 or so first shot, babystepped to a solid z0 and fired up another effector, first layer stuck like glue with no nozzle dragging and no detectable tilt! I believe 2 of my balls were too close together somehow. The relief is palpable, I was starting to get very frustrated! That said, I still would appreciate a walkthrough of how this calibration works. It's a little confusing to try to sort out how effector tilt effects the various calculations. Not to mention I have no idea how to find those points without relying on the script, which makes sorting through problems and narrowing options really frustrating.
Thanks for all the help thusfar Mr. DC. The joy of watching this print run flawlessly after such struggles is amazing and I owe much of it to you!
Thanks for all the help thusfar Mr. DC. The joy of watching this print run flawlessly after such struggles is amazing and I owe much of it to you!
|
Re: Auto Calibration / Auto Bed Leveling million questions - Paging DC42 December 13, 2016 02:31AM |
Registered: 8 years ago Posts: 301 |
|
Re: Auto Calibration / Auto Bed Leveling million questions - Paging DC42 December 13, 2016 04:11AM |
Registered: 7 years ago Posts: 154 |
[www.thingiverse.com]
Basically a heavily modified version of this. It turns out my first print was a little skewed causing some magnets to be slightly closer together than they were at the carriages. Now I'm getting much less (but still some) effector tilt. I'm still not getting a good first layer on bigger prints.
Basically a heavily modified version of this. It turns out my first print was a little skewed causing some magnets to be slightly closer together than they were at the carriages. Now I'm getting much less (but still some) effector tilt. I'm still not getting a good first layer on bigger prints.
|
Re: Auto Calibration / Auto Bed Leveling million questions - Paging DC42 January 05, 2017 06:45AM |
Registered: 7 years ago Posts: 154 |
|
Re: Auto Calibration / Auto Bed Leveling million questions - Paging DC42 January 05, 2017 11:07AM |
Registered: 10 years ago Posts: 14,672 |
Quote
clearlynotstefan
Dc, did you ever get that post out? I still can't get a good first layer down across the whole bed, and I've made further mechanical improvements to boot! I'd love to hear all the things mentioned in your post above! Thanks for everything!
Yes, see the thread "New wiki pages..." that I posted a few days ago.
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].
Sorry, only registered users may post in this forum.