@kinginnovator listen to Gr5. it is the voice of experience. Consider - if Cura can't get the Gcode right on a simple model like that then no one would use Cura.
I wrote a macro to read Gcode into AutoCad and I've done that with a couple of hundred Cura generated gcode files.
You can post your gcode, and I can read it into AutoCad and we can see. I'm going to bet that it's exactly correct.
I can do this soon.. I noticed that cura will put seems in the direction of home. Is there a way to put the seems opposite of this? If I can show you guys that I can control the direction of the skew then this is atleast a step in the right direction..
As I'm looking into your printer, the extruder has a double feeder a'la Bondtech with only one stepper driving both feeders right above the extruder, right? Further the cold side of the upper part of the extruder is of water/liquid type.
Then we'll see the two "heaters" connected to each side of the Y connector supplying the nozzle.
This is a "huge" printer based on Ultimaker gantries, -so I wonder what kind of steppers are you using here? Also, -what kind of main board etc.?
Edit; This is a really nice way to change / improve things the way you want!
Thanks
Torgeir
Instead of playing with the shape of the part. go to PREVIEW mode in cura and use the horizontal (not vertical) scroll bar to play through a few layers. Maybe also set PREVIEW mode to show the speed setting as a color. THIS is where you will see something non-symmetrical. Some "weird" move that doesn't happen very often. If you can find the movement that causes you to lose a step then this gives you a huge hint.
I've personally done exactly this. It's usually in the infill or travel moves among the infill or within "combing" moves within the infill. It's usually not the outer shell (outer walls) lines. Or sometimes it's on a z move possibly. Particularly buggy is if you move Z and another axis at the same time as the original programmers didn't expect this. But cura probably won't do this anyway. But it could be moves just before or after the layer change.
5 hours ago, Torgeir said:
As I'm looking into your printer, the extruder has a double feeder a'la Bondtech with only one stepper driving both feeders right above the extruder, right? Further the cold side of the upper part of the extruder is of water/liquid type.
Then we'll see the two "heaters" connected to each side of the Y connector supplying the nozzle.
This is a "huge" printer based on Ultimaker gantries, -so I wonder what kind of steppers are you using here? Also, -what kind of main board etc.?
Edit; This is a really nice way to change / improve things the way you want!
Thanks
Torgeir
Hey thanks! Its actually fed by a flexible shaft drive cable from an external stepper (nema 24). The shaft has a worm gear at the end that drive both Bondtechs simultaneously. This allows for some seriously high flows (as much as or more than a pellet system). Yes 2 heaters each 400watt potential.. This is all designed to achieve more that what's ever needed lol..
43 minutes ago, gr5 said:Instead of playing with the shape of the part. go to PREVIEW mode in cura and use the horizontal (not vertical) scroll bar to play through a few layers. Maybe also set PREVIEW mode to show the speed setting as a color. THIS is where you will see something non-symmetrical. Some "weird" move that doesn't happen very often. If you can find the movement that causes you to lose a step then this gives you a huge hint.
I've personally done exactly this. It's usually in the infill or travel moves among the infill or within "combing" moves within the infill. It's usually not the outer shell (outer walls) lines. Or sometimes it's on a z move possibly. Particularly buggy is if you move Z and another axis at the same time as the original programmers didn't expect this. But cura probably won't do this anyway. But it could be moves just before or after the layer change.
This is a great idea! I will do this tomorrow for sure to 'see into' what could possibly be causing this.... It makes zero sense if I can print perfect circles and walls in both directions but then have skewing happen on more complex shapes.. I know this will take time to figure out... My hunch is something that normally would work fine for a smaller printer (much less inertia) is hidden in the code and was never meant for something this large. I know that when I select randomize for the z seam this appears to solve everything via randomizing each layers start and stop locations and gives an overall average on the layers with everything working out in the end.. But I really need to pin point this because it's driving me crazy to have come so far yet have this happen at this time.. It's very expensive to build such a machine.. I really appreciate your ideas, knowledge and caring of my current situation and It will not go unnoticed!
1
Okay finally getting somewhere! I printed a test ring (double path wall). By manually choosing the seam location (front,back and right ) in 3 different prints I can control the direction of the skew (see picture). I also made a short video. I've also included the gcode and cura file. I think that it's actually printing perfect layers but something is shorting the distance most likely when it starts a new layer at the seam location when it tries to make a tiny quick movement to start the new layers outer path. This could be do to the inertia within a larger system like this not being able to correctly respond to such a tiny movement. What are your thoughts on this? Is there a way to compensate for all tiny movements as such or atleast alter the gcode to compensate? All ideas are welcome..
https://photos.app.goo.gl/AE3CrfcWjPtKfzah8
Below is a screenshot from AutoCad. This is all layers in the plan view and everything is concentric.
Start slow and with low Accel and Jerk. Going slow will allow any shaking/vibration to settle down before the head moves on. You have something that is creeping during the print. If you have any "Wipe distance" set it to "0" to avoid a short jerky move at layer change.
Design a solid cylinder so you can use spiralize and set the Bottom Layer Count to "0" so only the walls build. With no stops and starts you should get a nice straight model.
You have a new deal going here and you need to build a baseline between what it likes and what it hates. In order to do that you need to define both ends of that scale.
Edited by GregValiant
I've tried everything... No luck.. Please see list of true's below:
1. I can control the direction of the skew based on seam location.
2. If I choose random seam the skew goes away but I suspect every layer will be a bit off from the previous.
3. I can print a single wall/path line to any geometry without any skewing.
4. I can print a single layer perfectly over and over and over and over..
All the truths above are the bounding box to solving this pandora of fun!!! I have acceleration set to 80mm/sec super slow. I've tried higher accelerations with the exact same results. I've turned off jerk movement and also tried with it on. Wipe distance has been (0).. Most everything has been set to (0).. GR5 said this " Or sometimes it's on a z move possibly. Particularly buggy is if you move Z and another axis at the same time as the original programmers didn't expect this. But cura probably won't do this anyway. But it could be moves just before or after the layer change." This could be it???? My gut tells me it's happening at this moment of movement (see picture again)... This is the only thing I can think of that's different than printing a single layer line/path only.. Btw yes If I print spiral the problem should go away but that is not a solution to the actual problem that will effect all prints hear forward....
You could just edit out those tiny movements circled above by hand and see if that helps.
Are you using Marlin? What firmware are you using? I think at this point i would recommend you get klipper firmware. This will require hardware changes but it's worth it. Klipper is about 10 years more modern than Marlin. I guess the more important part is that it is a complete rewrite so it will have different bugs than Marlin and the guy who wrote Klipper is pretty damn smart and made very different choices (software architecture choices).
The people who wrote marlin are also very smart but it runs on a very weak processor. They made some tough choices to get it to work on an arduino. Klipper doesn't have cpu restrictions and so the code is less likely to be buggy and it can handle weird gcodes better.
Your issues are slightly more likely to be in travel moves (blue lines) than in extruding moves because travel moves are usually faster. Moves this short with acceleration so slow *shouldn't* move any faster but... there are a few bugs in marlin where it violates acceleration limits.
Did you know there are two (two!) different max accelerations in Marlin? One is meant to be *the* acceleration in every move - it tries to always accel and deccel at that rate. The other is a "max accel" which I think is double which Marlin stays under in special circumstances. It's a big kluge. A hack. It's ugly. I forget what circumstance might make it go over the acceleration.
I still think you should also mark your servo shafts and your couplers with a tiny mark using a sharpie to be *absolutely sure* nothing is slipping! This is a very very common problem that people deny exists until they are shown proof.
Hi @kinginnovator,
I'll see you're using Cura 5.3.1, the "new" Arachne engine. Have you ever tried the old slicer like Cura 4.13.1? This is the last stable version of the old Cura as far as I know (well it's a version 4.13.2 enterprise version).
The old version is more predictable for various reasons. There are fewer small corrections/movements that can benefit your printer.
In your printer definition setting there is only one 3mm filament, I assume you fix all this in the firmware.
You are using 900mm for X/Y, so how many steps/mm? Of course, since the Z height is 1000mm, how many steps/mm here? It's a bit hard to understand your problem without these numbers.
Thanks
Torgeir
I'm going to focus on the mechanical.
If you print a taller model...does the skew move back (lessen)?, does it stop moving and then continue straight up(?), or does the skew continue to increase at a constant rate/angle?
There is a lot of weight there and it's a pendulum. I'm thinking maybe the bearing footprint of the X beam and print head isn't sufficient. One mm of "offage" at the nozzle might be because of hundredths of a degree of rotation at the slide bearings.
The wider the footprint of those bearings, the less the deflection will be. You won't get perfect rigidity, but you need to strive for it.
55 minutes ago, gr5 said:I still think you should also mark your servo shafts and your couplers with a tiny mark using a sharpie to be *absolutely sure* nothing is slipping! This is a very very common problem that people deny exists until they are shown proof.
The set screws go into a pre drilled hole guaranteeing no slippage..
1
55 minutes ago, Torgeir said:Hi @kinginnovator,
I'll see you're using Cura 5.3.1, the "new" Arachne engine. Have you ever tried the old slicer like Cura 4.13.1? This is the last stable version of the old Cura as far as I know (well it's a version 4.13.2 enterprise version).
The old version is more predictable for various reasons. There are fewer small corrections/movements that can benefit your printer.
In your printer definition setting there is only one 3mm filament, I assume you fix all this in the firmware.
You are using 900mm for X/Y, so how many steps/mm? Of course, since the Z height is 1000mm, how many steps/mm here? It's a bit hard to understand your problem without these numbers.
Thanks
Torgeir
I will definitely try this! Very possible it could be tiny changes from the older version! Thanks 👌
54 minutes ago, GregValiant said:I'm going to focus on the mechanical.
If you print a taller model...does the skew move back (lessen)?, does it stop moving and then continue straight up(?), or does the skew continue to increase at a constant rate/angle?
There is a lot of weight there and it's a pendulum. I'm thinking maybe the bearing footprint of the X beam and print head isn't sufficient. One mm of "offage" at the nozzle might be because of hundredths of a degree of rotation at the slide bearings.
The wider the footprint of those bearings, the less the deflection will be. You won't get perfect rigidity, but you need to strive for it.
This is an older picture. We got rid of the slide bearings in the x,y and have gone to a track bearing setup.. lighter and much simpler.. if I can control the skew this eliminates alot.. In theory it shouldn't be mechanical.. I can print a single line tube 4 inches with no skew repeatable. I just might try klipper too at this point if I'm able to... we'll see... going to dive into everything today and even try an older version of Cura possibly...
1 hour ago, gr5 said:You could just edit out those tiny movements circled above by hand and see if that helps.
Are you using Marlin? What firmware are you using? I think at this point i would recommend you get klipper firmware. This will require hardware changes but it's worth it. Klipper is about 10 years more modern than Marlin. I guess the more important part is that it is a complete rewrite so it will have different bugs than Marlin and the guy who wrote Klipper is pretty damn smart and made very different choices (software architecture choices).
The people who wrote marlin are also very smart but it runs on a very weak processor. They made some tough choices to get it to work on an arduino. Klipper doesn't have cpu restrictions and so the code is less likely to be buggy and it can handle weird gcodes better.
Your issues are slightly more likely to be in travel moves (blue lines) than in extruding moves because travel moves are usually faster. Moves this short with acceleration so slow *shouldn't* move any faster but... there are a few bugs in marlin where it violates acceleration limits.
Did you know there are two (two!) different max accelerations in Marlin? One is meant to be *the* acceleration in every move - it tries to always accel and deccel at that rate. The other is a "max accel" which I think is double which Marlin stays under in special circumstances. It's a big kluge. A hack. It's ugly. I forget what circumstance might make it go over the acceleration.
Edit out those tiny movements? Like in the g code?
1 hour ago, kinginnovator said:I will definitely try this! Very possible it could be tiny changes from the older version! Thanks 👌
Here are the version options? I don't see it?
54 minutes ago, kinginnovator said:Here are the version options? I don't see it?
Nevermind.. I found it.. About to print a file from this version.. Fingers crossed!
Ok everyone... I loaded a brand new copy of Marlin and also went back to cure 4.13 and the skewing went away!! I even printed at some very impressive speeds (see video link)... I honestly have no real clue to what is different or what changed... I still have lots more testing to do.. I really appreciate everyone's help and you guys are the very best!!!
1
4.13.1 is my go-to for anything mission critical.
There are still some issues with the 5.x versions and Klipper. The team continues to refine Cura for Klipper compatibility.
1
Awesome!! Again thanks for your patience.. it was night and day on this last print.. I just started fresh completely over to eliminate all the rest if the variables.. looking forward to see what she can do
..
1
ArunC posted a topic in UltiMaker 3D printers,
.thumb.jpeg.0b7a05eafc09add17b8338efde5852e9.jpeg)
ArunC posted a topic in UltiMaker Cura,
Recommended Posts
kinginnovator 6
Ok.... So I printed the same part at 45 and both sides skewed this time not a single side. This makes sense because they share the same direction. But I noticed in the firmware where the stop/start lines are at each layer change and I've been following a hunch. So I set z seam to random instead of shortest path and the skewing went away... So I decided to see if I could investigate further on this by printing a very long wall in both the X and Y direction (picture). No skewing took place in either to my surprise.. At this point there is definitely something different happening about the program in how it sees a more complicated geometry. Yes you mention the layers of complications going on behind the scenes yes. I need to figure out a geometry that can point the finger at the monkey aka (demon lol) hiding. The last wall part had no travel moves or possibly so much else when compared to the square frame.. Any ideas on what time of geometry I could use to say "ah ha!" I use Fusion 360.. Ive owned 8 other (small) printers before this one so I definitely think I could be dealing with a very tiny demon where acceleration isn't the same in a smaller machine when scaled to this. There has to be some geometric shape I could print to test the next step.. Remember when I when to random Z seam everything went away and this makes total sense if the same layer isn't repeated electrically/mechanically..
Edited by kinginnovatorLink to post
Share on other sites