Precision of U2 is 0,06 or 0,020 mm ?

[celso-santos 11]

When we specify the settings in Cura 14.03 it says that high quality is 0,06 mm. Can we really get the 0,020 mm quality? Are there design problems to achieve this level of ultra high detail?

Celso

 

[gr5 2,097]

No design problems.

But if you want higher quality than .05mm you should get a smaller nozzle first. The nozzle hole diameter is .4mm. The radius is therefore .2mm. You can't get good details in X/Y any better than .2mm without getting a smaller nozzle.

So there is not much point in going higher than around .06mm for Z axis. I never go thinner than .1mm.

Also I have a print that took about 40 hours at .1mm. If I went to .02mm layers it would take 200 hours or 8 days to print. More than a week! That's too long to tie up my printer.

 

[illuminarti 18]

I've repeatedly asked for references to 'Quality' to be removed from the Cura UI and quickprint settings names, because I think it's totally misleading. It implies that certain levels of settings are 'good' and others are 'bad', and that's simply not the case. It depends what you are printing. Some are faster, some are slower. Some are better able to reproduce finer details than others.

But it's wrong to think that thick layers = poor quality = bad to use, and thin layers = good quality = good to use. It's just not that simple.

 

[jonnybischof 60]

Just for reference:

I recently printed something with 0.03mm layer height:

There is a huge difference in the top chamfer quality shown in the pictures. It turned out pretty much perfect, no more visible "stairs".

BUT if you do overhangs with 0.03mm, you can get very bad results because the 0.03mm plastic is much more fragile than a "strong" 0.2mm strain. I'm not saying it doesn't work, but it does require a very well tuned machine and a good quality filament.

 

[celso-santos 11]

Jonny and iluminarti

1. How can we add pictures to our comments in this forum?

2. Why when we open a gecode file in Cura it shows the sliced parts but not the parameters that created it. Shouldn't we see all the data related to the file?

Celso

 

[IRobertI 517]

1: http://umforum.ultimaker.com/index.php?/topic/4525-how-to-upload-an-image-to-the-forum/

2: I would assume it is done that way to avoid overwriting your settings. You can however tell cura to import the settings with File -> Load profile from GCode

 

[znib 2]

Just for reference:

 

I recently printed something with 0.03mm layer height:

.

 

If I'm not mistaken, the smallest step an Ultimaker can do in the Z-axis is 0.02mm, right? So shouldn't the layer hight be a multiple of 0.02mm?

I assume if you set it to 0.03mm it will do one layer at 0.02mm the next at 0.04mm and so on. If Cura calculates the amount of fillament needed for a 0.03mm Layer it will over extrude by 33% on one layer and under extrude by 33% on the next.

Of course these are just asumptions, can someone confirm / disprove them?

[illuminarti 18]

No, the steps per mm setting for the z-axis on a UM2 is 200 - i.e., a step is 0.005mm.

A 0.02mm layer height would therefore be 4 steps. Interestingly though, as I understand the code, the planner drops any movement segment that is less than 5 steps, so it looks like a z-change by itself of 0.02mm wouldn't happen; instead the layer change would happen as part of the next x-y movement that required more than 5 steps total.

But when printing at 0.03mm, that's 6 steps, so not an issue - and they should all be even sized layers.

 

[poshman 0]

So - regarding resolution, isn't it another way to set things up to obtain what is requested?!

IF - guess there's an awful lot of job behind that - but if we could divide the perimeter into e.g. 0.02 mm layers, finish 5 of those - and then make the infill in 0.1 mm... That should save a lot of time but still get us the surface finish we want..

Is this possible?...

From my point of view most likely a nice to have thing as 0.1 mm resolution is very fine for me.. Of course, chamfers and radius would look better in higher res.

But with this method it should be possible to speed up normal prints as well.

Perimeter at 0.1 resolution and infill with 0.2 or 0.4 ?!

 

[jonnybischof 60]

Actually I never thought about whether it was doable or not - I just set it to 0.03mm and it came out fine

 

I was wondering about something similar. There may be models where you'd want to get different layer heights for different "stages" of the model. Like: make the base with 0.2mm layers, but the details that come on top of the base with higher resolution.

Or of course, have a model which consists generally of high resolution, fine layers, but only the bridging layers are solid 0.2mm. I haven't tried it but I guess bridging at 0.02mm works pretty badly...

This shouldn't be much of a problem, as there are other features that also create large flow fluctuations (like setting different speeds for infill and shell. Never tried that though). But it is of course a lot of work to actually implement such a feature...

 

[Daid 303]

No, the steps per mm setting for the z-axis on a UM2 is 200 - i.e., a step is 0.005mm.

A 0.02mm layer height would therefore be 4 steps. Interestingly though, as I understand the code, the planner drops any movement segment that is less than 5 steps, so it looks like a z-change by itself of 0.02mm wouldn't happen; instead the layer change would happen as part of the next x-y movement that required more than 5 steps total.

But when printing at 0.03mm, that's 6 steps, so not an issue - and they should all be even sized layers.

Actually, the Z change is never done as a single move up, it's done as an X/Y/Z move towards the first point on the perimeter.

G1 F2100 X111.76 Y93.37 E80.17075

;LAYER:3

G0 F9000 X111.90 Y93.10 Z0.75

 

This also causes less Z blob.

 

 

Bridging as 0.2mm layers works better then at 0.1mm layers (or thinner). So in some future you might get the option to do bridges at a different thickness.

 

Variable layer-height slicing is something that is in my head. But I also know there is some patent about it. Generic idea is that you look at the angles in the model, at high angles you use thinner layers. Smoothing the changes in layer thickness is also important so get a better consistent look. However, doing this in my current code requires some major changes.

But, it is in my head. I know how to do this. It will take significant brain time.

 

[Dim3nsioneer 557]

Bridging as 0.2mm layers works better then at 0.1mm layers (or thinner). So in some future you might get the option to do bridges at a different thickness.

 

Today's teaser no. 2... :cool:

 

[gr5 2,097]

Daid maybe you already do this but it would be nice to add comments in the gcode where bridging is occurring in case someone wants to experiment with a plug in like: over extrude or decrease speed, or increase speed, or over extrude for the first mm, then underextrude the rest, or change acceleration, or whatever.

 

[illuminarti 18]

Daid -

I just submitted a pull request for some UM1 Marlin code to do with acceleration blending. It's something I fixed last summer, but never fully tested, or submitted before. It looks like the UM2 Marlin has the same problem. Basically when moving in multiple axes - as on the layer change - it tries to ensure that no one axis exceeds its acceleration limit. This is done by taking into account what proportion of the total move distance (and hence time) is spent on each axis. However, when it finds a violation, it sets the acceleration to the raw value, not the scaled one : in other words, the acceleration value used isn't the one that would pass the test which caused it to be used in the first place: it's a lower value.

This causes the acceleration used for 3-axis moves to be lower than it needs to be - and so, for instance, layer changes end up slower than they could be.

Please take a look at it, and let me know what you think.

Actually, the Z change is never done as a single move up, it's done as an X/Y/Z move towards the first point on the perimeter.

G1 F2100 X111.76 Y93.37 E80.17075

;LAYER:3

G0 F9000 X111.90 Y93.10 Z0.75

 

This also causes less Z blob.

 

 

 

[Daid 303]

Today's teaser no. 2... :cool:

Well, I have lots of ideas. So do not expect everything to make it in a release.

I did an experiment with "doubled up" layer thickness for infill. Did not win as much time as had expect. The code was quickly hacked, so didn't commit it. I also did it for both sparse and top/bottom infill. And for top/bottom, "double upping" is a bad idea. Except when it's a bridge.

Now, for bridges. The current bridge detection sucks. It's done per layer, which is wrong, and the detection fails a lot of times, and has false-positives. I have a test model testing a lot of cases right now. But I haven't make a proper detection yet.

 

[Dim3nsioneer 557]

Tea time small talk by Daid... :lol: No, really, thank you for sharing this. It's really interesting to read about!

I once thought of writing a plugin which could double up whole layers in the case of a purely vertical structure. Did you double up only the infill or everything except the outermost shell for which one can now set a separate speed in Cura?

 

[Daid 303]

I once thought of writing a plugin which could double up whole layers in the case of a purely vertical structure. Did you double up only the infill or everything except the outermost shell for which one can now set a separate speed in Cura?

I doubled-up everything that is yellow in the GCode view ;-)

What I did, is, I check for every even layer where there was infill in this, and the layer below it. Then those areas are "doubled up" areas. Then I substracted the "doubled up" areas from the normal infill areas. This means there are no gaps or missing infills where you have slopped areas or the model ending on an odd layer instead of an even layer.

Then the doubled-up areas where printed with twice the extrusion.

It worked, but the top was really ugly. So I should only do it for sparse infill I think.