Hi, I just completed my calibration to about 0.05mm, and after over 1 year of basically not printing anything because I couldn't print on the entire bed, I finally have it working in large part due to the fantastic work of the duet firmware (also hardware, the duet wifi steppers are amazing), however I have a few questions that I am hoping someone that has a duet and has already done the calibration could answer.
1) Is it normal for the calibration to go off between print jobs? Currently I'm running hollow test cubes (just the walls) to tune the slicer settings, and I have found that in between prints I might get a 0.053 deviation and then get something like "before 0.076 after 0.062" in the next print, and then get this "before 0.071 after 0.062" in the print after that.
I'm assuming those deviations are somewhat normal and are a mixture of vibration from the print job, imperfections on the fsr triggering and homing triggering, and effector tilting (really need to order 713maker effector one of these days), but would like to know for certain if its normal or if its an hardware issue I have.
2) What do you do regarding plastic coming out of the extruder when doing a G32 before each print, so far it doesn't seem to have been a significant problem, but is there any known way to automatically eliminate it, I currently do the G1 E-5 F2000, G32, G1 E5 F2000 which ofc doesn't completely solve the issue.
3) While I'm currently using S6 calibration, I noticed S7 calibration changes diagonal rod length, wouldn't that cause issues with printing accuracy?
Thank you.
Questions after calibration
Re: Questions after calibration
Micael, looong time Duet user here...
1) No, it isn't normal for calibration to go off but the LSQ fit calculations most likely will report slightly different results after probing for lots of reasons. You are looking for convergence and precision. You seem to be converged around 0.062 deviation. That's actually pretty high, I like to see less than .03 deviation. I would wager that your FSRs are not completely unconstrained in Z and that results in a little "false height" to trigger them (i.e. more force is required which is observed as a larger Z value). There are three things you can do:
You can work on your mounting system to remove inconsistencies.
You can use dc42 G30 "H" parameter to compensate for height errors at each probe point.
You can use grid based compensation and not worry about it. I think David recommends if you use compensation, that you do not want to implement the H probing height corrections in the previous option. I'll try to verify that with David today.
2) I heat up, then retract 2mm and then wipe nozzle with leather. Unless you have a wipe station, there is no automatic way to eliminate it.
3) Yes, don't use it! Stick with S6. It turns out to be pretty easy to actually measure arm length and use that value. Then print a calibration cube and make minor adjustments based on measuring that.
1) No, it isn't normal for calibration to go off but the LSQ fit calculations most likely will report slightly different results after probing for lots of reasons. You are looking for convergence and precision. You seem to be converged around 0.062 deviation. That's actually pretty high, I like to see less than .03 deviation. I would wager that your FSRs are not completely unconstrained in Z and that results in a little "false height" to trigger them (i.e. more force is required which is observed as a larger Z value). There are three things you can do:
You can work on your mounting system to remove inconsistencies.
You can use dc42 G30 "H" parameter to compensate for height errors at each probe point.
You can use grid based compensation and not worry about it. I think David recommends if you use compensation, that you do not want to implement the H probing height corrections in the previous option. I'll try to verify that with David today.
2) I heat up, then retract 2mm and then wipe nozzle with leather. Unless you have a wipe station, there is no automatic way to eliminate it.
3) Yes, don't use it! Stick with S6. It turns out to be pretty easy to actually measure arm length and use that value. Then print a calibration cube and make minor adjustments based on measuring that.
Sublime Layers - my blog on Musings and Experiments in 3D Printing Technology and Art
Start Here:
A Strategy for Successful (and Great) Prints
Strategies for Resolving Print Artifacts
The Eclectic Angler
Re: Questions after calibration
http://forum.seemecnc.com/viewtopic.php ... 00#p100319
See my post yesterday. I am going to put all of this in a blog post so it is easier to find.
See my post yesterday. I am going to put all of this in a blog post so it is easier to find.
Sublime Layers - my blog on Musings and Experiments in 3D Printing Technology and Art
Start Here:
A Strategy for Successful (and Great) Prints
Strategies for Resolving Print Artifacts
The Eclectic Angler
Re: Questions after calibration
Thanks for the reply.
Ok so the program reporting slightly different results after probing between prints is not out of the ordinary?
Right now I think my deviation issue is in large part due to how I setup the H points, at the start I had the H points inverted (using negative values when it should be positive and vice versa), so thinking I had mechanical issues I completely stripped my hotend of any cables, which was when I did my current H values setup, and during that time after having inverted the H values to the correct form I was getting somewhere between 0.032 to 0.026, I also did the H values at an higher bed temperature than what I'm using for PLA, since I will end up printing more ABS than PLA, this might contribute a bit since my onyx is not particularly flat and neither is my snowflake for that matter.
The only reason I haven't redone the H values is because the cut of my aluminium heatspreader should arrive this week and hopefully also my printbite, so I would need to redo my H values in a couple of days.
That being said its entirely possible that my FSR setup is in need of some tuning, while the holders and plungers were machined and rotated very well outside the printer, it is likely that they aren't going up and down nearly as well once they got on the printer, since my onyx and snowflake aren't perfectly flat its possible the screws are entering at a very slight angle and as such angling the plunger, making it harder for it to slide, I know for a fact that one of my fsr is easier to trigger than the other 2, so I might do some sanding on the other 2 after I mount the aluminium heatspreader if its still needed.
Tried the grid based compensation, good feature, especially if we ever get the HE280 accelerometer integration (for those that have it, since it seems superior to fsrs to me), or for people that have the IR proble, but right now to get the best results with FSR it requires manual H parameter compensation in the file, in the end got significantly better first layers with the tuned G32 with H parameters, than I did with the untuned grid based compensation, which was still impressive because even with an untuned version I was able to print in the entire bed (just not with consistent thickness), so clearly tons of potential there, especially for people with significant bed problems.
BTW here is my heightmap file (which I'm not using), if anyone is interesting in comparing, this is a default rostock max v2 onyx setup with FSR
Hmm I see so the only way to remove those slight pieces of plastic really is to just do it by hand, that's a shame, in my prusa I actually had a nice trick where my glass was smaller than where my print head could go, so I was actually able to wipe it off there, but ofc I also didn't use the nozzle to probe the bed.
I forgot to ask a very important question, which is regarding the alignment of the towers, I have a digital angle gauge which I used to tune my towers, however my bed is at a slight angle so if I say 0 the gauge in the middle of the bed (facing towards the front of the printer), but then I rotate it (facing towards the side of the printer) I will get a slight angle different due to the angling of the bed, in this case how should I align the towers?
Should I 0 the angle gauger close to the tower at the correct angle, or should I 0 it at the center, then rotate it and remove that value from the angle the tower should be at (for example if after rotating it gives me -0.3º, that tower would only need to be at 89.7º to make the 90º), which way would give the best results?
Also how precise do the towers need to be? Because tuning the towers in the rostock is a complete nightmare, every time you move one tower you screw up another, I got them to about 0.2º error, and while I'm sure I could do better it takes a lot of unscrewing, budging, screwing and repeat to improve on this
Finally I'm assuming this doesn't make much of a difference, but is there a correct position for the fsr, currently I have my FSRs between the towers, should I put them in front of the towers instead?
Ok so the program reporting slightly different results after probing between prints is not out of the ordinary?
Right now I think my deviation issue is in large part due to how I setup the H points, at the start I had the H points inverted (using negative values when it should be positive and vice versa), so thinking I had mechanical issues I completely stripped my hotend of any cables, which was when I did my current H values setup, and during that time after having inverted the H values to the correct form I was getting somewhere between 0.032 to 0.026, I also did the H values at an higher bed temperature than what I'm using for PLA, since I will end up printing more ABS than PLA, this might contribute a bit since my onyx is not particularly flat and neither is my snowflake for that matter.
The only reason I haven't redone the H values is because the cut of my aluminium heatspreader should arrive this week and hopefully also my printbite, so I would need to redo my H values in a couple of days.
That being said its entirely possible that my FSR setup is in need of some tuning, while the holders and plungers were machined and rotated very well outside the printer, it is likely that they aren't going up and down nearly as well once they got on the printer, since my onyx and snowflake aren't perfectly flat its possible the screws are entering at a very slight angle and as such angling the plunger, making it harder for it to slide, I know for a fact that one of my fsr is easier to trigger than the other 2, so I might do some sanding on the other 2 after I mount the aluminium heatspreader if its still needed.
Tried the grid based compensation, good feature, especially if we ever get the HE280 accelerometer integration (for those that have it, since it seems superior to fsrs to me), or for people that have the IR proble, but right now to get the best results with FSR it requires manual H parameter compensation in the file, in the end got significantly better first layers with the tuned G32 with H parameters, than I did with the untuned grid based compensation, which was still impressive because even with an untuned version I was able to print in the entire bed (just not with consistent thickness), so clearly tons of potential there, especially for people with significant bed problems.
BTW here is my heightmap file (which I'm not using), if anyone is interesting in comparing, this is a default rostock max v2 onyx setup with FSR
Heightmap.csv- (1.48 KiB) Downloaded 228 times
I forgot to ask a very important question, which is regarding the alignment of the towers, I have a digital angle gauge which I used to tune my towers, however my bed is at a slight angle so if I say 0 the gauge in the middle of the bed (facing towards the front of the printer), but then I rotate it (facing towards the side of the printer) I will get a slight angle different due to the angling of the bed, in this case how should I align the towers?
Should I 0 the angle gauger close to the tower at the correct angle, or should I 0 it at the center, then rotate it and remove that value from the angle the tower should be at (for example if after rotating it gives me -0.3º, that tower would only need to be at 89.7º to make the 90º), which way would give the best results?
Also how precise do the towers need to be? Because tuning the towers in the rostock is a complete nightmare, every time you move one tower you screw up another, I got them to about 0.2º error, and while I'm sure I could do better it takes a lot of unscrewing, budging, screwing and repeat to improve on this
Finally I'm assuming this doesn't make much of a difference, but is there a correct position for the fsr, currently I have my FSRs between the towers, should I put them in front of the towers instead?