Cura model height question

[paxpring 4]

Hey all. I have a short question. Why is it that Cura does not "fill" the entire 3D model with layers? It looks like Cura is compressing the model in the x-direction (see the grey area on top and above the edge on the bottom). It could easily fit another layer. At .3mm layer height it could even fit 2 more layers inside the grey area.

 

I could not find a similar topic so my apologies if this is a recursive question

 

 

[Carla_Birch 116]

It's hard to tell from the image being side on, but is this item single walled you are printing?

 

If so it could be that the top of the model is to thin in terms of wall width for cure to print the top part, without editing "line width" "Print thin walls" or a smaller nozzle size so that it can print thinner walls.

 

Adding the project file from cura to the thread can help people better understand the model and the settings you are using.

[paxpring 4]

1 hour ago, Carla_Birch said:

It's hard to tell from the image being side on, but is this item single walled you are printing?

 

If so it could be that the top of the model is to thin in terms of wall width for cure to print the top part, without editing "line width" "Print thin walls" or a smaller nozzle size so that it can print thinner walls.

 

Adding the project file from cura to the thread can help people better understand the model and the settings you are using.

 

 it is not a single-wall model, just solid. At the moment I do not have access to the printer but I figured that the project file wouldn't help me in this issue as it could very well be just a very specific single parameter I should adjust in Cura. Or am I wrong?

[ahoeben 1,948]

You could try setting the "Slicing Tolerance" setting to "Inclusive".

[paxpring 4]

20 hours ago, ahoeben said:

You could try setting the "Slicing Tolerance" setting to "Inclusive".

 

Hmm okay, I've set that setting to "Exclusive" in order to achieve better dimensional accuracy in the x- and y-directions

[ahoeben 1,948]

As far as I know, that setting only affects the Z axis. But I could be wrong. In any case, setting it to "exclusive" would cause exactly the issue you are seeing.

 

Edit: See https://github.com/Ghostkeeper/SettingsGuide/blob/master/resources/articles/experimental/slicing_tolerance.md

Edited August 7, 2020 by ahoeben

[paxpring 4]

On 8/7/2020 at 1:50 PM, ahoeben said:

As far as I know, that setting only affects the Z axis. But I could be wrong. In any case, setting it to "exclusive" would cause exactly the issue you are seeing.

 

Edit: See https://github.com/Ghostkeeper/SettingsGuide/blob/master/resources/articles/experimental/slicing_tolerance.md

 

How I interpret it was like this cura support page explains. The parameter affects all axes. 

https://support.ultimaker.com/hc/en-us/articles/360012614559-Experimental-settings

[ahoeben 1,948]

3 hours ago, paxpring said:

The parameter affects all axes.

You are correct, I was wrong.

Still, setting it to "exclusive" causes exactly what you are seeing.

[paxpring 4]

Thanks for your help! 

 

I've worked with "exclusive" for a while now and in the accuracy of the x-and-y-directions has improved like expected. However, it bugs me that the z-direction is off by two layers on 0.2mm layerheight. I designed something with a z-height of 11mm and it turned out 10.6mm, which is not dimensionally accurate at all ... I've tried the option "middle" but that reduces the x-and-y-derection's accuracy, just like "inclusive" does.

 

I would like to be able to have accurate dimensions in all directions without any compromise. 

 

How come Cura does not treat dimensional accuracy as their default? I know it is important that prints look good and that that might affect other parameters but I am a strong believer that aesthetics do not have to compromise dimensional accuracy. At least this isn't an issue with PrusaSlicer and my Prusa MK3S.

Edited August 12, 2020 by paxpring

[GregValiant 1,251]

On 8/12/2020 at 9:18 AM, paxpring said:

"I would like to be able to have accurate dimensions in all directions without any compromise. 

 

...I am a strong believer that aesthetics do not have to compromise dimensional accuracy. "

 

Me too.  Lately I believe we have fallen into the minority.

 

I've been thinking about this topic as I watch my machine go back and forth and I've deleted a lot thoughts I had on this because it finally occurred to me that the whole thing boils down to one question...

 

Within the Cura generated gcode file what is the Z height of the last layer?

[DivingDuck 104]

This is an never ending story and maybe one cause why I see so rarely Cura as tool in productive environments. A lot of responsible persons there don't accept one or more failed prints first before they have a usable part, especially when a print took hours for mostly simple shaped parts.

 

[paxpring 4]

On 8/14/2020 at 1:05 AM, GregValiant said:

 

Me too.  Lately I believe we have fallen into the minority.

 

I've been thinking about this topic as I watch my machine go back and forth and I've deleted a lot thoughts I had on this because it finally occurred to me that the whole thing boils down to one question...

 

Within the Cura generated gcode file what is the Z height of the last layer?

 

That's a parameter you have control over in Cura. Or am I missing your point?

 

 

On 8/14/2020 at 12:34 PM, DivingDuck said:

This is an never ending story and maybe one cause why I see so rarely Cura as tool in productive environments. A lot of responsible persons there don't accept one or more failed prints first before they have a usable part, especially when a print took hours for mostly simple shaped parts.

 

 

What do you mean by productive environments?

[burtoogle 513]

Late arrival to this topic. Could you please attach the project file for this print so I can take a look, thanks.

[DivingDuck 104]

52 minutes ago, paxpring said:

What do you mean by productive environments?

E.g. companies that want to use 3D prints for the daily work. They usually have not a 3D expert in the production / tool shop and usually use variants of existing parts where only a dimension like a hight need to be adjust to fit real world dimensions (e.g. like 3 dimensional spacers).

 

[paxpring 4]

59 minutes ago, burtoogle said:

Late arrival to this topic. Could you please attach the project file for this print so I can take a look, thanks.

 

Thanks for your participation, the issue I described in the very first post was due to my personal value setting for the parameter "Slicing Tolerance", so no further investigation of a project file is required. For me personally I would still like to find an answer to the question why this parameter does not have a value for dimensional accuracy in all three directions simultaneously.

 

22 minutes ago, DivingDuck said:

E.g. companies that want to use 3D prints for the daily work. They usually have not a 3D expert in the production / tool shop and usually use variants of existing parts where only a dimension like a hight need to be adjust to fit real world dimensions (e.g. like 3 dimensional spacers).

 

 

Yeh true, although UM is already known for their relatively high level of "plug and play" machines, right? I guess the process of 3D printing inherently requires at least some knowledge on the topic, even with so called "plug and play" 3D printers. 

 

 

Edited August 17, 2020 by paxpring

[DivingDuck 104]

Well, in my case the problem isn't the printer, for me it is Cura and how to deal with software quality. I really don't want to complain anyone, but I too like expect, that a 3D printer software is capable to have a accurate dimensional calculation w/o the need to do discussions about. It shouldn't be a user who have to do the math to recalculate matching layer height's. At least Cura need to inform a user if there is an dimension that is not achievable with given settings. They have the dimensions for the print and do the slicing.

 

And also, if errors like this happen once it have to be fixed immediately with an interim release and not some when with other features. This is something what I expect from an ISO certificated company.

 

To be clear on this, I don't deal with shrinkage of filament (as this is something a designer/engineer have to deal with), but things like this example with variable layer hight (what is a bit different to your problem):

https://community.ultimaker.com/topic/32356-cura-461-calculation-of-body-height-with-use-adaptive-layers-wrong/?tab=comments#comment-264247

Edited August 17, 2020 by DivingDuck

[burtoogle 513]

With regard to accuracy in the horizontal versus layer height, the problem is really akin to digital signal processing in that if the incoming signal contains frequencies that are higher than (1/2) the sampling frequency then the digitisation will be subject to aliasing errors. The analogy suggests that if the model's contour varies with z values with a greater frequency than twice the layer height then information is going to be lost when the model is sliced.

[paxpring 4]

2 hours ago, DivingDuck said:

Well, in my case the problem isn't the printer, for me it is Cura and how to deal with software quality. I really don't want to complain anyone, but I too like expect, that a 3D printer software is capable to have a accurate dimensional calculation w/o the need to do discussions about. It shouldn't be a user who have to do the math to recalculate matching layer height's. At least Cura need to inform a user if there is an dimension that is not achievable with given settings. They have the dimensions for the print and do the slicing.

 

And also, if errors like this happen once it have to be fixed immediately with an interim release and not some when with other features. This is something what I expect from an ISO certificated company.

 

To be clear on this, I don't deal with shrinkage of filament (as this is something a designer/engineer have to deal with), but things like this example with variable layer hight (what is a bit different to your problem):

https://community.ultimaker.com/topic/32356-cura-461-calculation-of-body-height-with-use-adaptive-layers-wrong/?tab=comments#comment-264247

 

Yeh, that's what I meant, slicing the STL is part of the 3D printing process: UM machines come with their default slicer Cura. Dimensional accuracy should be at least one of the default presets in ANY slicing software, so I agree that this is indeed a problem with Cura compared to for example PrusaSlicer.

 

However, the lack of in-slicer feedback and support is a completely different issue. In my opinion, the latter is something ANY slicer can (and should) improve at, not Cura only. It is true that 3D printers (and thus corresponding slicers) still have enough room for development in different directions but I guess your expectations are way above the current status quo of 3D printers. We are not there, yet.

 

 

2 hours ago, burtoogle said:

With regard to accuracy in the horizontal versus layer height, the problem is really akin to digital signal processing in that if the incoming signal contains frequencies that are higher than (1/2) the sampling frequency then the digitisation will be subject to aliasing errors. The analogy suggests that if the model's contour varies with z values with a greater frequency than twice the layer height then information is going to be lost when the model is sliced.

 

This analysis is way over my knowledge of 3D printing, printers and slicing software. However, what I am trying to explain is that in my opinion Cura is lacking a default set of parameters and values to deliver prints which are dimensionally accurate in all directions, a preset just like any other existing preset they have upon installation. It should be possible as this is a non-issue for for example PrusaSlicer. Even more so, I feel like it is a MUST for UM machines as they market their printers to be suitable plug-and-play business applications, which increases the need for dimensional accuracy. 

[GregValiant 1,251]

"However, it bugs me that the z-direction is off by two layers on 0.2mm layerheight. I designed something with a z-height of 11mm and it turned out 10.6mm..."

 

My point was that if I design something at 11mm tall and slice at .2 layer height/.2 initial layer height there should be 55 layers and the last layer should be at 11.00mm.  I am trying to understand how .4mm could be missing from the top of the part by changing from "Middle" to "Exclusive".

Given the above (11mm tall, .2 layer, .2 initial layer) the Gcode file should should have layers 0-54 and end up at 11.00mm.  A final Z height of 10.6 instead of 11.00 is a definite aberration and it flies past troublesome directly to WTF.  I want to understand how and why there is a nearly 4% error in the height.

That's why I asked what the final height in that Gcode file was.

[burtoogle 513]

12 minutes ago, GregValiant said:

I am trying to understand how .4mm could be missing from the top of the part by changing from "Middle" to "Exclusive".

 

Inspired by this thread, I have today submitted 2 PRs to UM that restore those missing layers...

 

https://github.com/Ultimaker/CuraEngine/pull/1306

 

https://github.com/Ultimaker/CuraEngine/pull/1307

 

The current behaviour may be unintended (due to bugs) or intended (due to their intentions), we shall see.

 

[GregValiant 1,251]

Burtoogle - I read your comments in the PR regarding the missing 1st and last layers.  I've been cogitating...

 

We have a cylinder of 25mm diameter and 11mm tall.  It has 55 .2mm deep cuts arrayed around the circumference such that each cut exists on 1 layer only.

When sliced with Middle it looks like this.  The Gcode shows 0-54 layers @ .2mm each and the final Z = 11.00.  This is how it was designed and what I expected to see.

 

When sliced with Exclusive it looks like this.  The first layer has a single feature that is .2mm tall.  Succeeding layers each have 2 features that are .4mm tall (.2 up and .2 down).  The top layer (54 in Gcode) disappears.  The numbering in the Gcode file goes from 0 to 53 so yes, there are two layers missing and the top layer is at 10.6mm.

  

 

A model with a single .2mm deep feature on the top of the model, when sliced with Exclusive, shows that the top layer (and consequently the feature) disappears.  Once again, the top layer Z is 10.6.

 

In either case, the diameter was 24.6mm.  Using a square model, the side dimensions were 24.6mm.  Line width was .4mm.

 

Good Luck with this one.

 

 

Edited August 17, 2020 by GregValiant

[burtoogle 513]

I find their choice of words here very confusing. The underlying operation when exclusive mode is selected is that the polygons that define the outline of each layer are calculated as the intersection of a given layer and the layer above. Intersection is like a logical AND, you need both parts to be present (true) for there to be an output, why didn't they call it "intersection" or "and" rather than "exclusive"? Similarly, the underlying operation when inclusive mode is selected is that each layer's polygons are made by doing a union of the current layer and the layer above. Yes, you guessed it, union is like logical OR. So why didn't they call it "union" or "or" rather than "inclusive"? Perhaps the marketing people thought that using engineering words would be too scary for the gentle user.

[paxpring 4]

11 hours ago, burtoogle said:

I find their choice of words here very confusing. The underlying operation when exclusive mode is selected is that the polygons that define the outline of each layer are calculated as the intersection of a given layer and the layer above. Intersection is like a logical AND, you need both parts to be present (true) for there to be an output, why didn't they call it "intersection" or "and" rather than "exclusive"? Similarly, the underlying operation when inclusive mode is selected is that each layer's polygons are made by doing a union of the current layer and the layer above. Yes, you guessed it, union is like logical OR. So why didn't they call it "union" or "or" rather than "inclusive"? Perhaps the marketing people thought that using engineering words would be too scary for the gentle user.

 

Again, I must admit that your conversation is above my knowledge and by all means, please continue. However I must correct you on something here regarding the highlighted sentence. I've graduated my master on human computer interaction and I can tell you that your comment is quite ignorant towards other users (I do not mean to offend). I fully understand your reasoning but not all people know as much of 3D printing as you do, including myself. For you, your terminology would suffice better. However, based on my personal experience in user-testing and user interaction design, I understand their choice of terminology. For a lot of those "gentle" less experieced or knowledgeable users (like myself), the terms Cura uses right now do mean more than any of the terms you suggested. 

[burtoogle 513]

Fair enough. I guess there's always going to be conflicts of this nature when you have an application (or anything) that is targeted at a broad spectrum of users; engineers to artists, etc. I'm afraid I can only see the world through my engineer eyes so hence my confusion here.

[GregValiant 1,251]

Ghostkeeper's Usage description of "Slicing Tolerance" - "This setting is named after its intended use rather than its functional effect. If you have multiple pieces that need to slide past each other, the theoretical shape of the layers may physically prevent exact fits. In such a case, you can set this setting to Exclusive so that the layers are guaranteed to stay within the bounds of the original volume. Barring warping, sagging and similar deformation effects, this would guarantee that the parts fit within each other and are able to slide past each other.

In reality there are always other effects going on that prevent this. In practice, this setting can be used to get slightly more or less tolerance between two sloped surfaces, as can be seen in the above images."

     I am an ME with 50 years experience scribbling special machines, fixtures, dies, automation, whatever.  It was always a basic tenet that any part be fully described within the borders of the drawing.  Didn't matter who made detail 1 or who made detail 2 - they need to fit together.  Tolerances, location of features, etc., needs to be part of the drawing - which is now the 3d model.

So I have a  philosophical problem with the idea of generating "tolerance" via a software setting.  It is akin to allowing the technician to determine tolerances.  That indicates ad-libbing by the technician and should never be allowed.  Now that the designer and the technician are often the same person doesn't change my view of that.  If two parts need to fit together, they should be designed to fit together.

    Being aware of what type of machine was going to be used to make a metal part was necessary to design a part that could actually be made (as opposed to what we referred to as Optical Illusions).  Being aware of the 3d printer/slicer/firmware capabilities is just as necessary.  An eye must be kept on those capabilities when designing parts of an assembly.

Paxpring - in one of your initial posts you mentioned problems between the PrusaSlicer / Prusa combination and then using Cura with (I must assume) the same Prusa printer.  If the parts slice differently in Cura than they did in PrusaSlicer causing dimensional anomalies within the Gcode this would seem to be an elemental problem between the slicers that needs to be addressed (somehow).  Using the Exclusive setting (with it's side-effects) to make adjustments in dimensions just doesn't sit well with me.  Again, it has the effect of allowing a technician to unilaterally alter a design.

 

And after all of that - the old guy forgot to change Exclusive back to Middle, sliced and printed a model that turned out two layers too short.