LARGE holes too small by much more than nozzle width!

[markaudacity 0]

I am familiar with the difficulty of accurately depositing an inner radius with FDM—someone has described it as "trying to draw a circle with snot" 😄
My experience so far is with this occurring with small holes, and usually by a nozzle diameter or two, but I have a part with a 1.129" round hole that prints at 1.09 and a .631" hole that prints at .603 with a .4 mm nozzle—an enormous discrepancy.
Exterior dimensions are within .002", so I know it isn't a scaling issue.

There has to be a way to compensate for this without having to change my model to dimensions that will be wildly out of spec once I cut them in aluminum.
Halp!

[mkoic 2]

Under “Shell” there is a setting called “Hole Horizontal Expansion” that can do exactly what you are looking for.

[gr5 2,266]

But he wants a different expansion for 0.631 holes versus 1.129 holes.  Note that these are in some weird unit called "inches" - it would be nice if they were translated to mm for the rest of us to comprehend 🙂

 

But seriously, I don't really have a feel for what the error is without opening up my calculator which I'm too lazy to do (sorry).

 

But to your point about having 2 models, one for FFF (aka FDM) and one for aluminum.  I recommend you do have 2 models.  Or use some kind of parametric CAD that makes it trivial to switch hole sizes.

 

In my experience, enlarging all the vertical (keep in mind that the issue is different for horizontal holes) holes by 0.4mm is about right.  But it sounds like you are seeing different enlargement amounts for different hole sizes.  This is strange.  Maybe something else is going wrong.  But I don't have a feel for how much error we are talking about.  Is it less than 0.1mm?

 

Note that if you paid the typical $50,000 for an injection mold, some poor engineer would have to take your model and make tons of changes to it so that it could be made in an injection mold and the changes would be a lot more involved than simple hole sizes.  Even 90 degree corners have to be changed by a bit (not sure if smaller like 89 or larger like 91 degrees) so that after it pops out the corners are back to 90 degrees.  I know this is small comfort.

 

[gr5 2,266]

By the way, Ultimaker has some profiles that are intended for extra high accurate results.  I think they are called "engineering" maybe"?  I forget but here is a summary of how to get higher dimensional accuracy:

 

Line width: 0.4
Wall thickness: 1.2
Top/Bottom thickness: 1.2
Speeds: 35-40 (all speeds, except travel)
Jerks: 20
Horizontal expansion: -0.03
walls: 3
Inital Layer Height = 0.1
Slicing Tolerance = Exclusive
Combing Mode = off
Outer before Inner Walls = Checked

[markaudacity 0]

“Note that if you paid the typical $50,000 for an injection mold, some poor engineer would have to take your model and make tons of changes to it so that it could be made in an injection mold“

Sure, absolutely.  Injection molding is not a rapid prototyping technique, though, and to expect the same level of pre-processing to get a good result from FDM as from injection molding is a bit out of scale, I think? 
 

FWIW, PrusaSlicer never had this problem, even on a trashed Prusa Mk2, so I don’t think it’s entirely universal to the medium. I would expect a bit more from Cura?
 

[I get that inches are stupid, believe me, but since this is an English-language forum and since almost all native English speakers, and thus most native-English-speaking makers and manufacturers, use inches, it’s not that crazy to expect folks to know that .020”=.5mm 😉]

Edited August 19, 2020 by markaudacity