What should I avoid when creating *.stl files?

[Jsk]

I just received the first batch of 3d printed parts that I designed for my current ship build. Most of the parts printed successfully, but others failed. I do NOT want my own 3d printer. That's a hobby unto itself and I have plenty of hobbies. However, I would like to know how to design *.stl files with a minimal chance of failed prints. Let's face it, since I'm not doing the printing I'm paying for someone else's time and material. I want to avoid wasting their time and my money.

 

My process has been this: I model the parts in Wings3d (my modeling application of choice for many years), which is a subdivision modeler. Wings will export *.stl which I open in the Prusa slicer to see if the slicer catches any errors in the file. Mostly, I think *.stl is a pretty basic file format so the program used to model the part shouldn't matter too much. I don't think there's much in the *.stl that can cause problems when printing but I'm not entirely certain. Occasionally, the slicer will identify an error which I can't find in the origin model so I'll open the *.stl in Blender, and then export it again as *.stl. After that, Prusa typically will not report any errors.

 

I have not added any supports. To me, that seems like something that would be more specific to the slicer program used in the actual printing process. Is that a correct assumption?

 

Also, the guy who printed these for me said that something in the files had 'blown up the printer and he had to clean the tank'. I'm not sure what that means. I've used him to print purchased *.stl files and he's been pretty reliable. A number of the parts I had designed came out very foreshortened as if they had collapsed in length. Is that something that could be caused by the *.stl file?

 

Another question I have is about warping. My understanding is that most warping is due to the curing process where the resin inside the casting cures at a different rate than the outside layers. As a modeler, is there anything I can do to alleviate this?

 

For example, in the picture below, would the 8 inch gun--which is mostly solid--have less of a tendency to warp if it was hollow? if there was a channel running down through the center almost the entire length?

 

The 6-pounder guns under the 8-incher suffer from being too thin in the barrels and shoulder rests. That's just me not knowing the parameters of the printer used and that lesson has been learned.

 

But the 6-inch barrel at the bottom of the picture suffers from something else. I also had a number of ventilators that somehow seem to have shifted during printing. Is there anything in the *.stl that could cause that or is it simply an error in the slicer used to print?

Any other advice for a 3d print noobie?

 

TIA!

[thibaultron]

Most of what I see seems to stem from a insufficient number of supports.

 

Here is an example of one of my typical cannons with supports, in 1/24th scale.

 

What scale are your prints in?

 

 

[Jsk]

These are for a 1/200th scale model. So the 8" gun (the big one) is 68mm in length. I'm not sure what slicer the guy printing the items is using and I'm assuming his program allows him to add the supports. This particular part came without supports attached but it looks like there were 10 in-line running along the bottom. The 6" gun did have a number of supports attached similarly. Other pieces do seem to be adequately supported.

[thibaultron]

That's a very small scale! The only thing I can think of is to print them vertically with the barrels farthest from the build plate, and the widest part toward the bottom. There is not enough of the smaller guns to tell me what they are supposed to look like.

 

Also have him cut back on the cure time, long cure times can warp thin parts. Playing with the exposure times might help.

[Jsk]

Thanks. Yeah, they're small. The small guns are suppose to be 3 and 4 pounder Hotchkiss (I think). They're no more than 0.5 inch from muzzle to butt. I've tweaked the model with beefier barrels and shoulder stocks and will try again.

[thibaultron]

As the parts are so small. you might want to have a solid base with the parts then supported by sacrificial standoffs. This makes it easier to remove them without risking separating them from a larger support. I did this for two small step plates for a HO scale tender bunker I designed.

 

The graphic below shows this with thin walls going from the base to the curved base (horizontal in this view) that the step is mounted to. One of the support walls is circled in red.

 

[CCClarke]

Lots of good information here.  I'll try to add a little more.  Since you aren't doing the printing, I won't get too technical.

 

These are requirements before starting any resin print:

 

1. The printer build plate is level/calibrated.

2. Slicer settings are calibrated for the resin being used.

3. Printing is performed in a temperature-controlled environment. 

 

Your first picture speaks volumes.  Most parts, (especially long, thin ones) are tough to reliably print parallel to the build plate, (no matter how well supported) without problems, though there are exceptions.   Parts need to be angled -or vertical, depending on the design when hanging from the build plate. 

 

All of the "Pro" slicers, (usually an upgrade from the vendor's free version) include auto-orientation.  In most cases, the slicer-determined part angle is adequate, but it depends on the design of the part.

 

You can print parts vertically, (suspended from the raft by supports) if they follow certain design guidelines.  A part with gentle contours is ideal.  Any overhangs parallel to the build plate require supports, but orientation is an important consideration for best results.  The first two pictures I've attached illustrate this point, with two types of orientations, (Parallel to the build plate and angled) based on the part's geometry depicted. 

 

The torpedo afterbodies are vertically printed, with supports under the mid-point where the forebody will be bonded.  Since its a flat surface, and subject to micro-sagging, a little sanding will quickly remove any support artifacts.

 

FYI: The control surfaces are oversized-since these are meant to be handled by young students as part of a STEM outreach program I'm involved with.  Magnets are bonded inside the body in the two places the torpedo contacts the stand, (also containing magnets) allowing the model to be removed and passed around.

 

The slope of the top section flares away from the main body gently, requiring minimal support structure.  A (unseen in the shots) hole inside the torpedo propeller shroud where the propellers live also serves to prevent suction forces from deforming the part during the build plate lift cycle.  The supports that run parallel to the body are latticed, meaning the supports brace one another, minimizing the chance of a support failure that would lead to deformation of the area they're supposed to be supporting on the model. 

 

Layer lines are minimal, and the supports are easy to remove.  Note the rafts on the build plate.  These use less resin than a full raft and make it easier to remove the parts, minimizing breakage.

 

Note the (3) torpedo mounts are angled.  (Each torpedo requires two.)  This ensures the mount's flat faces remain flat.  I could have angled the torpedoes, but there would be many more support artifacts on the model that have to be filled and sanded afterwards.  I'd rather wait longer for a vertically oriented print, (more layers than if it were at an angle = longer print time) than fill support pock marks with resin manually, then cure and sand.  (I'm lazy that way.)

 

The 4th and 5th pictures below illustrate thin, round part orientation for submarine sail handrails.  Angled "skate" style rafts are used to minimize breaking during removal from the build plate. The part orientation works with the curve of the handrails, so they are built-up gradually while printing.

 

-Notice the supports where they attach to the rails.  Most are interlocked and all feature pointed tips where they contact the part. 

 

The four upper and lower rudders have a 3mm hole for a brass shaft.  See how the holes gently slope upwards at the bottoms and then curve upwards at the top?  No supports are needed, and the holes have perfect geometry and more importantly, dimensional tolerance, which is press-fit.  Even the bridge compass, (the two smallest parts) have tiny handles about .1 mm thick.  Properly angled, these print perfectly.  

 

Modeling:

 

You said you're using a Sub-D modeling app.  Are you "freezing" the sub-D polys with a high enough sub-division level before saving the file?  Higher levels minimize faceting insuring the best possible surface finish.  A lot of designers mistakenly believe a slicer will make the model's geometry smooth.  Slicers can add vent holes but cannot change the model outward appearance otherwise. 

 

As mentioned, sacrificial stand-offs can be used for certain parts but require removal/re-finishing post-print.  The 6th picture shows an example, with the sacrificial supports highlighted in orange.  The last picture shows how I set it up for printing in the slicer - I added supports to the lower sacrificial supports, but the uppers were left alone.  The lowers would probably print fine without supports, but why take a chance on a long print?

 

The stern of the VA class submarine is printed vertically.  The hull slopes inward, supporting the previous printed layer without supports.  The stern planes, dihedrals, and shroud stand-offs all use sacrificial supports modeled into the stern.  These angle outwards, supporting the outward angled parts of the stern and get cut out after curing, without the need for standard supports.  You want to model them very thin and/or add a slight beveled inset where they meet the hull for easier removal.  Properly removed, a little sanding erases any hint they ever existed.

 

Modeling error-detection/correction: You already know what a PITA this can be to properly correct.  Occasionally, errors are detected by the slicer when the part is imported.  Most of the Pro-version slicers are able to eliminate most problems, but I try to correct them in the modeling phase before exporting as an OBJ/STL. 

 

If you have a feature in your modeling program that allows you to merge nearly coincident points in the mesh that's a plus.  Merging points can create 2-point polys, which will be identified by the slicer as non-manifold polygons.  Those have to be found and removed as well.  

 

I'll stop here.  If you have any specific questions related to printing that I haven't answered clearly enough, feel free to ask questions and I'll try to do better.

 

CC

 

Edited December 4 by CCClarke

[CCClarke]

I'll show you my two go-to tools I reach for often when post-processing parts after cleaning and support removal, which begs the question:  Is your friend the printer giving you completed/cured parts with or without supports removed?

 

But first, The Why

 

Most 3D prints will have tiny pock marks where the supports contact the part.  It takes a little practice, but with experience, you can manually place the supports in areas that any leftover support blemishes aren't visible.  This is a good reason why the (vertical) orientation of the stern in the last shot above prints so cleanly.  Most of the support structure is located inside.  Even though I usually use the Automatic Orientation and Support functions in Chitubox Pro, I still add a few where I think they'll improve the part quality.

 

Just for educational purposes to illustrate the differences between vertical and auto part orientation, I've included one sliced vertically-oriented Stern, and two shots of the stern from different angles after slicing in Auto Orientation mode.

 

The print time differences aren't too drastic, (displayed on the left side of the slicer screen) but the number of supports contacting the exterior is.  Which one would you rather touch up after the supports are removed?

 

To fill the tiny pock marks, I apply resin to the area, smooth it with a cotton swab, feathering the edges and use a UV curing wand to cure the area.  Another swab dipped in IPA, followed by a couple of swabs dipped in water get the surface ready for sanding.

 

I've found this reciprocating sander from DSPIAE to be invaluable to reach places my other sanding tools can't and it quickly smooths the surface.  There's a tip for every occasion supplied along with a few sheets of pre-cut adhesive-backed sanding paper. 

 

The Phrozen Cure Beam UV wand (a pair are supplied) is the perfect way to cure resin after applying it to the hull for support contact repairs, interior spaces that a standard cure station can't illuminate properly, or bonding parts and filling the leftover seams after filling them with resin.  These have saved me hours of re-work and I highly recommend them.  Both are available on Amazon.

 

CC

 

 

 

 

[Jsk]

Wow, a lot of information to absorb here. Thanks so much. It's all good. Most of the parts I get back from the printer still have the supports attached so I can determine the orientation in which they were printed.

 

Does it make sense to orient the model and model the supports in the modeling program? The few slicer programs I've looked at do that for you. I assumed it would be rather slicer/printer specific but perhaps it's just more for convenience sake.

 

I've been rather old-school in my modeling thinking that the fewer triangles in the model the easier and more effecient to print. But it sounds like that's not really the case.

[thibaultron]

Lychee, at least, allows you to save any model back to an STL file. I put in the supports (both automatic and manual), then save the supported model to a STL file, for future use.

[CCClarke]

22 hours ago, Jsk said:

Wow, a lot of information to absorb here. Thanks so much. It's all good. Most of the parts I get back from the printer still have the supports attached so I can determine the orientation in which they were printed.

 

Does it make sense to orient the model and model the supports in the modeling program? The few slicer programs I've looked at do that for you. I assumed it would be rather slicer/printer specific but perhaps it's just more for convenience sake.

 

I've been rather old-school in my modeling thinking that the fewer triangles in the model the easier and more effecient to print. But it sounds like that's not really the case.

A higher poly count can bog some computers down during the modeling phase, but I have never experienced detrimental effects on the slicer's performance.  Look at models on ThingVerse and note how faceted many are because the modeler didn't add enough geometry.  There's a noticeable difference between a 48-sided cylinder and one with 360 sides as far as smoothness/finish.  It's a common mistake that can ruin an otherwise nice effort.

 

Orientation angle depends a lot on the geometry of the object being printed.  If designed well, I can print hull sections vertically without a problem and all support attachment points remain hidden, cutting laborious post-support removal cosmetic restoration for a smooth finish.  For smaller, organically-shaped parts, I'll start with auto-orientation before adding additional supports manually.  Optimal printing results are a combination of experience and a well-designed part. Mistakes are part of the learning process if they aren't repeated. 

 

Keeping a Print Log adjacent to the printer for every print has been invaluable, allowing me to track how many hours I have on the printer's screen, maintenance and record problems and solutions.

 

I've added a handy guide I hope will illustrate my post more clearly.

 

CC

Proper Part Orientation.pdf

Edited Saturday at 05:54 PM by CCClarke

[Lieste]

With FDM printers a high model resolution is useful (smoother shapes are possible), but the print result is often better if you allow the slicer to cull to a minimum spacing between points - the changes in direction can result in 'bumps' in the surface as the head overshoots path points which are too close together. By default this can be a touch aggressive or off... but you can tune these settings to suit the model type and printing speeds desired. This can also modify shape accuracy for CW/CCW path loops which can ruin surface finish where the error is excessive for the size and curvature.

Resin can generally give much better results, faster and with smaller/finer detail, but there is more mess potential.

[thibaultron]

When I CADed cannons, I started with 48 surfaces on a round part. For anything 1/48th or smaller, this was fine. at 1/24th faint coarseness could be seen, if you looked closely. This was with a resin printer, at 35um layer heigth printed at 45 deg.. You did have to look closely though. I drew the rest of the 3D files at 72 facets, and no problems were seen. With a filament printer, I have no idea, as I don't have access to one.,

[Lieste]

FDM printers use the slicer to construct a series of straight line segments for each polygon in the layer - with a maximum deviation from a point or from the straightline before inserting a path-point and a minimum distance between sequential points set in the slicer. Generally you get better results from a higher poly/smoother surface, but the slicer is faster if the surface detail isn't excessive. A long straight  is stored as two points. A curve is stored as more points, and the slicer geometry is 'similar' once you get above a strongly faceted shape. There can be directional artefacts if you have lower deviation accuracy, and surface noise if the minimum distance is kept too low (the defaults can be set an order of magnitude too small for clean paths, and introduce too many points, but with limited accuracy, and some tuning and test prints are useful to improve both speed and surface finish/detail - 'measles' is how I'd describe this particular fault.).

[CCClarke]

Since the OP is using resin printing, I'm trying to avoid discussing FDM printing in detail, but comparisons are useful for anyone on the fence regarding which technology is best for their needs.

 

I've attached a few shots for comparison for the same shapes but with differing numbers of sides to illustrate polygon faceting.  Each set shows the mesh on the surface and the surface without the mesh.  (Multiple display modes are essential to object modeling. I hope the all-quad polys in the examples are large enough to show the differences.

 

The sphere on the left has 24 sides with 12 segments.  The one on the right is doubled with 48 sides and 24 segments.  If you look at the edges closely, you can see the faceted surface more pronounced on the sphere on the left.  The cylinders have 24, 48, and 180 sides from left to right.  Increased geometry creates a smoother surface.  At some point, adding additional geometry becomes overkill.  If you print tiny parts, you can get away with less geometry, but large parts will show the difference clearly.  It all depends on what look you need based on the part being printed - unless you're a masochist that enjoys sanding.

 

All STL (and OBJ) files are triangle-based, meaning a four-point polygon is halved diagonally.  Th eg-code printers use is triangle-based. Triangles also eliminate non-planar polygons, since a triangle can't be "bent" no matter how the three points are connected.  Non-planar polygons don't render properly, but that's irrelevant to this discussion.

 

When modeling, I always use quads, (four-point polygons) and they're automatically converted to tris, (three-point polygons) if the file is exported as an STL or OBJ.  If I need to reverse-engineer an OBJ file, the triangles must be combined into quads for ease of mesh manipulation. 

Edited Monday at 03:19 AM by CCClarke

[CCClarke]

Just found this site hidden in a folder.  When manually orienting a print, the first question that comes to mind is, what angle is best?  I often opt for 30/30 (pitch and roll) but sometimes it isn't suitable because the part is too large and those angles would have it extend beyond the build plate.

 

Here's one way to determine the best print angle for a resin printer that might be of help:

 

Angle calculator for smooth surfaces in resin printing – RC87