Compensate Wall Overlaps - causing zits

[Link 34]

Hi,

 

I have been printing some parts with large rounded edges, these are shells and just two lines wide (using a 0.8 nozzle), i have been getting some random artifacts on the inside and outside of the curves, after plenty of messing about and reading other people seeing the same thing (these artifacts only appear on the curve), it turns out its to do with Compensate Wall Overlaps, if i disable this the artifacts go away.

 

I came across this thread by @smartavionics where it seems Compensate Wall Overlaps is kind of broken and wont be fixed

https://github.com/Ultimaker/CuraEngine/issues/577

 

What do people tend to do with this, if i disable it on small parts i get stringing etc due to the extra material deposited at the overlap site, however on larger parts with curves i get artifacts with it turned on.

 

I am looking at other alternates to these artifacts but simply turning this feature off solves it, maybe my max resolution was too high as per my other thread ?, there are plenty of threads online around Cura and external artifacts on curves, sounds like something is amiss 

 

https://github.com/Ultimaker/Cura/issues/4860

 

would really appreciate some advice/thoughts

 

Thanks

 

[burtoogle 513]

Hello @Link, you could try using one of my builds that has had some improvements made to the wall overlap compensation feature. You can find builds for Linux and Windows at https://www.dropbox.com/sh/s43vqzmi4d2bqe2/AAADdYdSu9iwcKa0Knqgurm4a?dl=0

I don't claim perfection but it does work better than the Ultimaker version (IMHO).

If you are using the overlap compensation, make sure to set a non-zero value for min wall flow so that the thinnest wall lines are removed.

 

Another reason for getting zits is over-retraction. Please check that you are not retracting too much.

 

Finally, make sure to use the no-skin combing mode and set the max comb distance with no retract to something like 10 or 20.

 

Hope this helps.

[Link 34]

14 minutes ago, burtoogle said:

Hello @Link, you could try using one of my builds that has had some improvements made to the wall overlap compensation feature. You can find builds for Linux and Windows at https://www.dropbox.com/sh/s43vqzmi4d2bqe2/AAADdYdSu9iwcKa0Knqgurm4a?dl=0

I don't claim perfection but it does work better than the Ultimaker version (IMHO).

If you are using the overlap compensation, make sure to set a non-zero value for min wall flow so that the thinnest wall lines are removed.

 

Another reason for getting zits is over-retraction. Please check that you are not retracting too much.

 

Finally, make sure to use the no-skin combing mode and set the max comb distance with no retract to something like 10 or 20.

 

Hope this helps.

 

Thanks, what you advise to set the min wall flow too ?

 

Also do you tend to disable overlap compensation todally, what are the downsides of this ?, what is it meant to do when working properly ?

 

 

 

 

 

[burtoogle 513]

2 minutes ago, Link said:

Thanks, what you advise to set the min wall flow too ?

 

It's not very critical, it can be quite large, i.e. 50%. If it is too large, it could introduce gaps into wall lines that would otherwise be complete.

 

3 minutes ago, Link said:

Also do you tend to disable overlap compensation todally, what are the downsides of this ?, what is it meant to do when working properly ?

 

No, I nearly always use the overlap compensation and I do not get zits like you are seeing. That's with PLA and PETG using a 0.4mm nozzle and 0.5mm line widths.

[Link 34]

9 minutes ago, burtoogle said:

 

It's not very critical, it can be quite large, i.e. 50%. If it is too large, it could introduce gaps into wall lines that would otherwise be complete.

 

 

No, I nearly always use the overlap compensation and I do not get zits like you are seeing. That's with PLA and PETG using a 0.4mm nozzle and 0.5mm line widths.

 

I only tend to get zits with the 0.8 nozzle and 0.3 > layer, but turning off compensation def fixes it

 

interestingly Cura says to use the min flow you have to print outer walls before inner will can affect overhang quality ?

 

What is your max resolution set to ?

 

thanks

[burtoogle 513]

1 minute ago, Link said:

interestingly Cura says to use the min flow you have to print outer walls before inner will can affect overhang quality ?

 

If I remember right, you don't have to print outer wall first but the way it works is that the first line that gets printed is extruded at the normal width and then subsequent lines that overlap that first line get extruded at a reduced width (or omitted completely if their flow is < the min flow value). So to get the most accurate outline, you need to print the outer wall first.

 

What is your max resolution set to ?

 

Generally, I just use the default values but I note that (a) they have recently introduced a Max Deviation setting and (b) have increased the default value for max resolution.

 

 

I think (OK, I'm guessing) that the max deviation is now really controlling the accuracy of the print rather than the max resolution. i.e. use longer line segments (up to max resolution) as long as points do not deviate by more than max deviation.

 

Using a coarse resolution makes the slicing quicker and the gcode file smaller but does have the downside of making it even harder to achieve a nice straight z-seam on a curved object. Here's what I mean...

 

Here's a model sliced using the default values for max res and max deviation. I have enabled coasting simply so that it  shows the z-seam position more clearly (I never normally use coasting as I think it is detrimental to print quality). Notice the ragged z-seam...

 

 

If you then change the max res to 0.05 (i.e. ten times finer), you now get a much nicer z-seam...

 

 

So the lesson there is that using a coarse resolution makes it hard to achieve a nice z-seam on curved objects.

 

PS - don't use the sharpest corner z-seam style, it very rarely works well. Always use a user-defined z-seam (with a small max res!)

 

[Link 34]

i have looked into this some more

 

the issue is when you have a hollow object, for example a box and the walls are set to twice the line width, for example 1.6 with a line width of 0.8, this means exactly two walls will fit, i have used the offset command in fusion 360 to create an even wall thickness around the corners.

 

 

 

The issue is clear even in the preview in Cura, with Compensation enabled the print stutters as it rounds the corner, when this happens on the actual print it creates blobs, disable compensation and it runs smoothly round the corner. Cura is implementing some level of flow compensation round curves for a wall which is exactly twice the line width. 

 

Even with a very basic cylinder it does the same, set the thickness to twice the line width and compensations makes the movement stutter 

 

 

 

 

 

 

 

 

 

[burtoogle 513]

I think I know what's happening there, could you please provide the gcode that stutters and also the stl file? Thanks.

[Smithy 1,144]

It is indeed strange, but I often compensate it with a smaller line width. So instead of 0.8, I try 0.79 or something like that, just to get rid of the compensation.

[Link 34]

model.gcodeBOX.stl

[Link 34]

i am using a 0.82 line width

[Link 34]

2 minutes ago, Smithy said:

It is indeed strange, but I often compensate it with a smaller line width. So instead of 0.8, I try 0.79 or something like that, just to get rid of the compensation.

 

i tried that too and as you say it stops the compensation, its like Cura thinks the walls are actually thinning around the corner when it isn't, almost like it slices in straight lines and these lines thin as they turn the corner...

 

i have tried increasing resolution but that doesn't help

[Link 34]

6 minutes ago, Smithy said:

It is indeed strange, but I often compensate it with a smaller line width. So instead of 0.8, I try 0.79 or something like that, just to get rid of the compensation.

have you considered disbaling compensation ?, i tried a couple of prints with it off, and not see a huge impact, but thats not to say i should turn it off !, i dont really know when it really helps a print ?

[Smithy 1,144]

No, haven't tried it yet, but will do so next time.

[Link 34]

it def fixes the issue on hollow items like a box with walls in multiple of the line width, however i have tried some solid models which have infill between the walls and tbh haven't seen any adverse affect of not having compensation on, but as i don't really know what it does i can't be sure i should leave it off...

[burtoogle 513]

Thanks for the files. The gcode didn't look bad as I expected and the model is too large for me to slice for my printer. Could you please provide the gcode and model for the cylinder, please?

[Link 34]

Sure, many thanks for looking into this !, this has the same issue

 

model.gcodetube.stl

[yellowshark 153]

I have never used wall overlap and never had zits. I only print my designs and designs all have the same wall width, irrespective of number of walls, i.e. measured nozzle extrusion width. I do not know what this thin wall lark is all about as I have never found a need to use a non standard design as I design to my environment.

[Link 34]

3 minutes ago, yellowshark said:

I have never used wall overlap and never had zits. I only print my designs and designs all have the same wall width, irrespective of number of walls, i.e. measured nozzle extrusion width. I do not know what this thin wall lark is all about as I have never found a need to use a non standard design as I design to my environment.

 

Like you I design my own stuff, interesting you don't use overlap, how many walls do you use if you designed a box for example ?, 2 walls of your set nozzle width ?

[burtoogle 513]

Thanks for the files. I believe the stuttering is caused by the model's line segment lengths getting "adjusted" by the overlap compensation and so the line segments end up with widely differing lengths, some very short.

TBH, I don't understand how a lot of the compensation code functions as it is not well documented. In the past, I have made some changes and I believe the version I have in my builds is better but it still doesn't behave very well. The current Cura master branch which will form the basis of the next release does make a better job of the cylinder when the line width is exactly 0.8 and the fill gaps between walls feature is disabled. My current build is also OK with a 0.8 line width and there is no need to disable the fill gaps feature. However, increasing the line width so that it is more than 50% of the wall width will cause the bad segmentation to occur again when the overlap compensation is enabled. Sorry, not very good, I know but that's how it is right now.

[Link 34]

By setting the line width to .8 it's actually masking the issue as the wall is 1.64 (2 X .82) so reducing the wall doesn't fix the issue just moves the problem away from the problematic overlap function. This is the method @Smithy uses. It's seems the overlap compensation code remains broken and not likely to get fixed 😔. Unless someone from UM can comment here as it's a common issue. 

 

Anyway many thanks for taking the time to look into it, really appreciate it. 👍🏻👍🏻

 

[burtoogle 513]

9 minutes ago, Link said:

It's seems the overlap compensation code remains broken and not likely to get fixed 😔.

 

They have had years to fix it and not done anything and they weren't interested in using the "improvements" that I have made to that code some time back. It seems unlikely that this will get better anytime soon.

[Smithy 1,144]

2 hours ago, Link said:

Unless someone from UM can comment here as it's a common issue. 

 

It would be great if you could open bug at Github about the issue. You find the link in the help menu in Cura. The chance that someone picks this issue and work on it is much bigger than when it is just placed here.

[Link 34]

On 6/24/2019 at 6:43 PM, Smithy said:

 

It would be great if you could open bug at Github about the issue. You find the link in the help menu in Cura. The chance that someone picks this issue and work on it is much bigger than when it is just placed here.

 

will do

[yellowshark 153]

On ‎6‎/‎24‎/‎2019 at 3:58 PM, Link said:

 

Like you I design my own stuff, interesting you don't use overlap, how many walls do you use if you designed a box for example ?, 2 walls of your set nozzle width ?

Yes spot on. Depending on thickness dims I might use more walls if it proves faster to print that using infill. FWIW I also always use 15% infill overlap.