Active leveling does not work for me

[tmostad 5]

The print bed is low in the front left corner and consequently high in the back right corner, of course magnified on prints with a large base as you can see here.

I have tried many times to re-level since I am a noob to this machine (and believing I don't know what I am doing). No difference whatsoever. In fact it seemed to worsen after this last firmware update. I began to think my glass was severely warped (in truth is isn't very flat but it turns out that that isn't the majority of the problem). I shut off the auto leveling and manually leveled then was able to successfully print. I need to tweak it some more but it is workable at the moment. Kind of frustrating to have what seems like a nice feature that I can't use.

Edited February 23, 2017 by Guest

[neotko 1,417]

Maybe your bed glass or bed heating plate is curved? I seen other posts of users having that. Mine had that but I fixed it manually (don't do that). Contact your reseller for more support.

[Daid 306]

From a software perspective, be sure to run the latest stable firmware. There are a whole bunch of improvements on the leveling measurement in 3.5

Also make sure your nozzle is clean before it starts, especially if you use something else then PLA.

There is a bit of variation/curves in the glass, so with the 3 point measurement that we do, there are still going to be high and low spots compared to what we measured. We are looking into measuring a full grid to improve on this area.

[neotko 1,417]

There is a bit of variation/curves in the glass, so with the 3 point measurement that we do, there are still going to be high and low spots compared to what we measured. We are looking into measuring a full grid to improve on this area.

 

Sorry Daid this will sound like a rant, but is more a concern.

Does that mean that we must assume 0.2-0.3 glass deformation? With a flat glass/bed that shouldn't happen. Unless ofc the user has a banana shaft, wooble, etc...

[tmostad 5]

If you took at least 9 points of sample in a grid then you could probably extrapolate accurately enough for most needs. Too many points and the calibration process becomes exceedingly slow. Better yet provide a bed that is flatter by design. I have read that others found flatter glass and that it makes a big difference.

[Daid 306]

Normal window glass is flatter. But more fragile. So that is not to difficult, but impossible for us to reliably ship. And a bit of a safety hazard to put into our product by default.

9 point leveling doesn't have to take a whole lot longer then what we have now. We are working on it, what we can do, is first measure the 3 points as we do now, and then with that data we have a pretty clear indication where the bed is, so we can start measuring very close to the bed after that. We can do the first 3 points with one nozzle, and then the other 9 with the other. So we combine the height difference measurement as well.

[fabreax 15]

Normal window glass is flatter. But more fragile. So that is not to difficult, but impossible for us to reliably ship. And a bit of a safety hazard to put into our product by default.

9 point leveling doesn't have to take a whole lot longer then what we have now. We are working on it, what we can do, is first measure the 3 points as we do now, and then with that data we have a pretty clear indication where the bed is, so we can start measuring very close to the bed after that. We can do the first 3 points with one nozzle, and then the other 9 with the other. So we combine the height difference measurement as well.

 

It looks promising !

[Daid 306]

It's also a bit of a catch 22, people want the leveling measurements to be fast, as they want to be at the printer in case the first layer fails. At the same time, if we take enough time during leveling, the first layer has a much lower chance to fail, so you do not need to be at the printer.

However, we cannot simply say "you will have to trust our leveling system" if it hasn't proved itself yet. And I don't think I have to argue that it still needs to prove itself a bit more.

[tmostad 5]

One thing I don't understand about the current level algorithm is that it doesn't sample the third point right at the adjustment knob though the first two are. The manual leveling process does and it seems to work pretty well for me. I suppose the question is moot given you will be doing something more sophisticated (and hopefully more accurate) in the future. It seems to be a curious choice nonetheless.

[fbrc8-erin 272]

If you took at least 9 points of sample in a grid then you could probably extrapolate accurately enough for most needs. Too many points and the calibration process becomes exceedingly slow. Better yet provide a bed that is flatter by design. I have read that others found flatter glass and that it makes a big difference.

 

Have you tried flipping your glass over and checking to see if the problem changes? If you get the opposite high/low points with the glass flipped, then it's a glass issue, and you should be able to get a piece of replacement glass under warranty.

[Daid 306]

One thing I don't understand about the current level algorithm is that it doesn't sample the third point right at the adjustment knob though the first two are. The manual leveling process does and it seems to work pretty well for me. I suppose the question is moot given you will be doing something more sophisticated (and hopefully more accurate) in the future. It seems to be a curious choice nonetheless.

 

We're researched a whole bunch of different options for the 3.5 firmware update. The current positions give the best results in repeatability and accuracy of the measurements.

The difference between manual and active leveling is that the active leveling needs some flex in the system for good measurements.

[fabreax 15]

With 75 microns layers, the auto level has a strange behaviour :

[fabreax 15]

Another printer with another glass plate (same .gcode file) :

Edited March 2, 2017 by Guest

[Daid 306]

Don't do 75micron first layers (other layers are fine, just the first layer needs to be thicker to catch variations). That's simply asking for problems. The variations in the leveling measurement, the bed and the rods will cause you nothing but troubles.

[fabreax 15]

Don't do 75micron first layers (other layers are fine, just the first layer needs to be thicker to catch variations). That's simply asking for problems. The variations in the leveling measurement, the bed and the rods will cause you nothing but troubles.

 

Thanks. The first layer is at 100 microns, do you suggest 150 microns ?

[tomnagel 126]

The default first layer thickness in all Cura profiles is 0.27mm

[tomnagel 126]

By the way: the default profiles in Cura are really good, when combined with UM materials. For other brand materials, profiles might need some tuning, but are at least a good starting point.

My advice: start experimenting and optimizing only after you have had good results with the default profiles.