Bed leveling between towers

[ewm]

We are having some bed leveling issues with Casper between towers. We have fixed the belt tension, updated the firmware and run the test mention here http://seemecnc.com/blogs/news/14684425 ... -perimeter in order to hopefully find a middle ground which we have found on all points except the mid points around the left tower.

[img]http://orig14.deviantart.net/8951/f/201 ... 9cf0vl.jpg[/img]

We used the dial indicator mentioned in the blog post and we are consistently getting these numbers. We can't find any mechanical issues. We also used a wixey angle gauge and the towers are square. It seems that it's just that one tower on the left being the contributing factor.

[ewm]

We looked into the issue further ensuring the table it was on was flat so no problem there. Noticed the glass was not sitting flat on the bed. Used a square and determined glass was bowed in the center. Removed the glass and found that the onyx bed was also bowed in the center. Loosened the onyx bed to see if it was too tight no change. Noticed that top base plate bows upwards as seen in the photo below. Not sure what to do.

[img]http://orig10.deviantart.net/ce01/f/201 ... 9cf0wl.jpg[/img]

[Jimustanguitar]

This thread might be your solution.

http://forum.seemecnc.com/viewtopic.php?f=36&t=8698

[ewm]

This thread might be your solution.

http://forum.seemecnc.com/viewtopic.php?f=36&t=8698

Getting around the issue with auto calibration is not going to work due to the fact the prints will still be warped in the same way. We need to make accurate prints using nearly the entire bed. We'll will likely use the auto calibration after we sort out the physical calibration first.

[geneb]

You're confusing auto-level and auto-calibration. OpenDACT does things like tweak tower rotations and other things. It actually does a genuine calibration. When the process is done, you can go to 1mm above the bed anywhere the nozzle can reach and the measured distance between the nozzle and the bed will be 1mm. It does this by sampling a number of points and then makes changes to a number of different EEPROM parameters until all the sample points agree, +/- some tiny factor. Once OpenDACT does its thing, you're done. It won't matter what slicer, host or SD vs Computer - you'll be printing with a perfectly trammeled machine.

The essence of what is happening is this - OpenDACT is modifying the "perfect" mathematical definition of a delta printer to match the "imperfect" physical model that the firmware is controlling. OpenDACT is printer-agnostic - it'll work on pretty much any delta that's running Repetier Firmware. (the firmware is the only one known that provides the ability to change all the required parameters via EEPROM tweaks, without requiring a recompile)

Auto-leveling adjusts for bed tilt and will modify the g-code on the way out to the printer in order to compensate for the tilt. This can result in weird looking parts, depending on how bad the tilt is.

g.

[Captain Starfish]

When the process is done, you can go to 1mm above the bed anywhere the nozzle can reach and the measured distance between the nozzle and the bed will be 1mm.

I think he knows this. It's fine if the inaccuracies are caused by a model wrapped around variations in tower angle, height etc. The process will leave the bottom of the print perfectly matching the bed.

Remember: all this autolevel and autocalibration stuff is based on the assumption that the bed is planar - dead flat. It's the reference against which all the calibration and levelling is done. If the bed is non-planar which it appears to be in this case, you'll end up with prints whose bases are equally and accurately non-planar.

ewm - I'm going to assume the newer machines still rest the heater PCB on a spider plate which is mounted on standoffs around its perimeter. If this is the case then any non-planarity of the base is mechanically de-coupled from the spider and therefore bed, which will help. Can you perhaps remove the heater PCB and check for a flat spider to start with? If that isn't flat, you need to make it so.

If you take the spider off and it's still not flat just sitting there, might be worth asking for a flat replacement or finding a local machine shop and getting them to surface mill it flat for you.

If it's not flat when mounted but flattens out when you remove the mounting bolts, you might need to look at why tightening up the bolts is distorting it.

If the spider's flat but the PCB isn't, again, maybe ask for a flat replacement. Or look at inserting appropriate shims between PCB and spider to flatten it out.

You're right though, until you have that surface nice and flat there's zero point in using OpenDACT or bed levelling software because they will just carry the errors through. And 0.02", 0.5mm, is significant.

[ewm]

I wanted to thank everyone for all the help! We spent all day/night yesterday getting Casper back on track and I'm happy to say it's fixed! We remounted the spider and bed making sure to apply weight when doing so to get the bed area level again and after that we used the manual calibration in Rollie's thread. Did the normal mode, then went onto the expert to get those values, followed by going back to the normal mode for a few iterations and managed to level it out. The circle test went quite well, there's a slight high spot in the back just right of the tower but it's a huge improvement overall and should work out just fine.

Thanks again!

[img]https://scontent-sea1-1.xx.fbcdn.net/hp ... 8053_o.jpg[/img]

[geneb]

\o/

[ewm]

Still having some trouble with getting Casper leveled out. We have two areas on the right hand side that are pretty severe and never seem to change despite several iterations with OpenDACT 2.0.5 and 2.1.1. I made a post on the OpenDACT thread with information on what we've done so far. If anyone has any suggestions on what else I can try to fix these spots I would greatly appreciate it. Thanks!

http://forum.seemecnc.com/viewtopic.php ... 625#p79258