Blender Design Iterations

I just put up a sub on Thingiverse, based on board game art from Underwater Cities. There was another sub, part of a bigger "replacement parts", but it was relatively plain.

Final Piece

I went through about 14 different versions - with each one getting closer to the final piece. The first useful piece was adding a camera in the side panel, so I could look at the picture of the sub WITH the model.

Version 7 - Adding Camera View

You can see in the upper right panel, you can see a bunch of pieces in an Outliner panel. In the middle right, you can see a camera shot in a 3D Viewport. (There's a lower right, where a properties panel is.) The jpeg of the box front was added using shift-A>Add Image, and I changed the rotation and scale to match the idealized object.

The iterations were roughly 3 categories of work:

1. Getting it to look like the sub in the picture (70% of the total work)
2. Making it printable (25%)
3. Giving it characteristics of a game piece (5%)

Getting it to look like the sub in the picture

A lot of this process is just taking standard shapes and modifying them. The buoyancy tanks are the same — I took a cylinder, did some loop cuts for the ridges that stick out. I copied a couple of UV spheres, turned them on the sides, aligned them with the ends of the cylinder. Select the vertices of the spheres that weren't visible and delete those vertices. Remove the ends of the cylinder, then boolean>union the spheres to those ends. In Edit>Vertex>Remove Double Vertices, to clean up problems. You can either copy it, or use Modifiers>Mirror>Apply.

Honestly, the shoulders and sacs were the messy parts. Shoulders were a lot of proportional editing from a cylinder. Look under the shoulders, there's a strange "sac". It was a isomorphic sphere, that was stretched on the z-axis, added a wireframe modification, and then I removed some vertices to make planes. This next part falls under making it printable: I tried to make it a pure mesh sphere, but it never quite turned out....so I took the inner areas of the holes and turned them into faces. (And the other internal faces I removed, so it would be a manifold object.)

I probably spend way too much time cleaning that up.

Making it printable

The side headlights broke off occasionally — I tried to make them interesting by mounting them to a sphere, then mounting the sphere to the sub. The angled headlights would break off when pulling supports.

The buoyancy tanks were round on the bottom. I'd get weird layers too often down there, so I used Pivot Point>Active Element...selecting a bunch of vertices on the bottom of the main platform, and the lowest 3-4 pairs of vertices on the buoyancy tanks...then Scale>Z to have them scale to the last selected vertex. (As a reminder, when doing this, it can be helpful to look straight on, use "Z" to go into Wireframe mode.) This flattening took care of the weird layers.

The braces that reached the buoyancy tanks were just bars to start — I made them thicker as they entered the sub's body, but eventually also extruded them down on the z-axis through the tanks...this eliminated the tiny interior supports, as well as giving it a more solid-looking feel (necessary for the next part).

Giving it characteristics of a game piece

It was too tall (z-axis, solved by shrinking the dome and the dome cage on the z-axis, then scaling out on the x- and y-axis.). The shoulders and sacs needed to expand in the x- and y-axis (primarily by using Scale).  Keep in mind, these were all separate pieces...making it easy to scale each piece, but also taking longer because I wasn't doing a single operation.

I put the angled headlights forward more on the model, to give a cleaner look. I did not include the "probes" that were on the right side of the sub's art...they were going to be too thin and easily breakable.

These were all relatively easy fixes. If I was more of a designer, I would probably have stronger (and earlier) design principles to fall back on.

Overall Learning Experience

If you're sweating a piece, maybe you're doing it wrong. I probably worked on the mesh sacs for too long. There were simple brute force methods that would have been fine. The dome brace was done as a separate piece, I should have just made it part of the dome. The buoyancy tanks on the bottom, I wanted perfectly round — for no real necessity. (The tanks had those brackets going all the way around, so that first layer was trying to keep those, but then have the tanks.) I printed multiple versions, trying to make the smallest change for the best result.

Cura Lulzbot Edition - The Good, the Bad, and the Ugly

I've been using the "Lulzbot" version of Cura for the whole time I've been printing. It uses a different number versioning system than the original Ultimaker Cura. (I'll refer to it as Cura LE.)

I have an ancient desktop OSX machine that is still functioning. Apple maxed out its upgrades at 10.9.5 which has caused some small headaches that will only get worse. For now, it works. It handles Blender with ease. It's stable. There's old software that I don't have replacements for. And it is maxed out in Cura "Developer edition" through Lulzbot - a test build from August 2017...every upgrade I've tried to install has failed.

The Must-Solve Issue

Lately though, the 2017 Cura LE has been problematic. The amazing Micro .25 SL toolhead prints fine objects with an uneven top relatively well. Sometimes on the bottom, the walls don't meet up with the infill, but it's been mostly minor. But give Cura LE a flat top, and...it gets bad.

You can see the bottom dino meeples have what appears to be support showing through - it's actually the top printing layer...and does not get any better than that.

Dinosaurs on the top, showing the "best results" for the side on the printer bed.
Dinosaurs on the bottom showing bad cross-hatching/spiraling for their top layer.

On this T(eal) Rex, you can also see the wall not meeting up with infill. Photo is a little blurry, not the dinosaur's fault...it wasn't moving THAT fast.

T-Rex from the printer bed side - walls and infill separating

The older Cura LE is fine for printing with the .5mm nozzle, and the Micro *could* print okay on non-flat surfaces, but why make bad prints if you can solve it? Incidentally, the upcoming experience has been summarized. I probably burned about 6-8 hours this weekend troubleshooting the logs, then trying various solutions.

The Good (eventually) - Ubuntu 18.04

Let's start with the Good — this morning I managed to get my laptop running Ubuntu 18.04 to run "one-step behind current", i.e. Cura LE 3.2.32. I finally have access to ironing, in case that was necessary for the fix (it wasn't necessary to resolve the surface issues, but it's nice). The best stories are one with an element of hope.

The Bad - OSX 10.9.5

Don't bother, it's not going to work. (I tried the various OSX installs on the Lulzbot site.) Some versions would just crash and terminate, not generating a log. The best case scenario, was when it generated a log as it failed to launch. You can poke around Console>Applications>cura (stderr.log) and learn some troubleshooting.

The "useful" error log:

  File "cura/CrashHandler.pyc", line 17, in <module>
ImportError: dlopen(/Applications/cura-lulzbot.app/Contents/Resources/lib/python3.5/lib-dynload/PyQt5/QtCore.so, 2): Symbol not found: _LSCopyDefaultApplicationURLForURL
  Referenced from: /Applications/cura-lulzbot.app/Contents/MacOS/../Frameworks/libQt5Core.5.9.1.dylib
 in /Applications/cura-lulzbot.app/Contents/MacOS/../Frameworks/libQt5Core.5.9.1.dylib
  File "/Applications/cura-lulzbot.app/Contents/Resources/__boot__.py", line 81, in <module>
  File "/Applications/cura-lulzbot.app/Contents/Resources/__boot__.py", line 66, in _run
  File "/Applications/cura-lulzbot.app/Contents/Resources/cura_app.py", line 86, in <module>
    import cura.CuraApplication
  File "cura/CuraApplication.pyc", line 4, in <module>
ImportError: dlopen(/Applications/cura-lulzbot.app/Contents/Resources/lib/python3.5/lib-dynload/PyQt5/QtNetwork.so, 2): Symbol not found: _LSCopyDefaultApplicationURLForURL
  Referenced from: /Applications/cura-lulzbot.app/Contents/MacOS/../Frameworks/libQt5Core.5.9.1.dylib
 in /Applications/cura-lulzbot.app/Contents/MacOS/../Frameworks/libQt5Core.5.9.1.dylib

It turns out that the modern libraries of PyQt5 (Qt 5.10+?) aren't compatible with 10.9.5. This won't stop someone from publishing software that "works" with 10.9.5...their code is fine. It's just PyQt5 that will reject you. There's a suggestion of this problem here: https://github.com/Ultimaker/Cura/issues/4466

The Ugly - Back to Ubuntu 18.04

How can it be both Good and Ugly? Because it is. I'm pretty sure that the Lulzbot developers are Debian folks. The Ubuntu instructions...*should* work. "Jessie" is Debian 8, "Stretch" is Debian 9, and "Buster" is Debian 10...what is Ubuntu 18.04? (You can check /etc/debian_version and find out.) Buster! So...it should be easy enough. But if it's ugly, clearly there is something up.

It turns out that their most recent (3.6.3) Buster build doesn't work with 18.04. (If you want to skip ahead to the solution, go to the Lulzbot site here and grab the older 3.2.32. Afterward, go to the directory you downloaded it to, and enter this command - "sudo dpkg -i cura-lulzbot_3.2.32_amd64.deb") The current Cura LE 3.6.3 is looking for a file (GLIBC_2.28), and Ubuntu 18.04 uses GLIBC_2.27. (The event viewer for Ubuntu 18.04 is called Gnome Logging, and you can find it as an application installed on the general build.) The problem is called out in the Cura backlog, but I don't expect it to be fixed again (although it had been previously fixed the 3.2.32 version) Someone else mentioned the same solution here: Ultimaker Cura github page. (Basically, use 3.2.32, install command is done differently.)

You might be able to install 3.6.3 by using "sudo apt-get install glib2.0", to get GLIBC_2.28. I'm not an expert on Linux, and don't know what it could break. (For the most part, I use my laptop for internet, and for portable Blender/Cura...easily refreshed if I needed to do a clean reinstall.) If I hadn't spent the time setting up preferences and printers in 3.2.32, I would go back and try it.

So...what does this mean?

If you're not running straight Debian, you might be stuck with 3.2.32. I wouldn't get too hung up...3.2.32 and ironing seems to work just fine. At least the ironed T-Rex on the left thinks so...

Ironed T-Rex on the left, regular 3.2.32 T-Rex on the right

Credits:

The T-Rex meeple was remixed (thickened) from here: https://www.thingiverse.com/thing:2749410
The Teal PETG is Polymaker's PolyLite.
The Stegosaurus meeple was taken from here: https://www.thingiverse.com/thing:2749406
The Silver PETG is Filastruder.
The Triceratops meeple was remixed (thickened) from here: https://www.thingiverse.com/thing:2749413
The Orange PETG is leftover from MakerGeeks. Sadly, I can't recommend shopping with them anymore...6+ months of not fulfilling an order, I'm still unhappy about that. Used to be so great.

Using Blender to take a 2D image into a 3D cutout

One of the easiest and most magical things, is the speed at which you can turn an image into a 2D mesh object with depth. Calling it 3D is...tough — it doesn't turn a cutout image into a full-formed velociraptor!

This is a phone stand, remixed from a phone holder on Thingiverse — although I also added an extra piece below the holder, to make it more stable. The cutout of the velociraptor is from an image that looked like this:

Once you have the image (jpeg, png, whatever), take it to a (free) website that converts images to SVG - scalable vector graphics format. There's also techniques where you use Inkscape or Adobe Illustrator, but the free websites are pretty much drag-drop-convert.

In Blender, use File>Import to import the SVG.  It will be a microscopic "curve". I typically scale it in x/y about by *40. Important next step: Object>Apply>Scale. If you skip this, it will do crazy things to your object. Object>Convert to "Mesh from Curve/etc". This will turn the curve into a regular mesh. Next, Add Modifier>Solidify>Thickness = 1, and apply it. This gives the mesh depth, instead of just being a plane. It's solid black, which is hard to see details...under Materials, I used the slider to make it gray.

At this point, your image is now a 2D object with depth. If you weren't doing this project in your main blender file, save it, open your main blender file, and then Append>[filepath/file]>Objects>[objectname]. It should be listed in your Scene now, as a mesh object.

I thickened it to about 5mm, Object>Transform>Snap>Origin to Geometry, and then in Orthographic mode, clicked through keypad 1/3/7, adjusting the orientation and rotation of the cutout piece. (For example, in the phone stand, rotated it 90 degrees while looking from above. From the side of the stand, rotated it 55 degrees to match the plane of the upper part of the stand. Once you have your object aligned, scaled, and rotated...I usually hide it, then select the primary object. Add Modifier>Boolean>Difference, then apply.

You're stamping through the main object - make sure nothing is left hanging. For example, if you look at the capital letter "A", If you stamp an "A" through an object, that triangle in the upper half is going to be hanging in mid-air. (Maybe you have your stamp go halfway through the wall of the primary object.) An example of this is another stand, using a Zelda-based image:

Part of the FDM printing process, behind the Zelda icon...the back wall was not a clean print. Probably because the stamp almost went through the wall. A thicker wall, or a shallower boolean could have eliminated it. Not really a problem, as people aren't going to see that back wall, but an imperfection.

Flexible Filament (and buying a printer from Craigslist)

October 2018, I listed a bunch of filaments and my experiences with them. It was an older post than that, but had never gotten around to posting before then...and it was correct for the time. I had picked up some flexible filament earlier in 2018 as well as a toolhead to print flexible filament...but didn't get around until this past month.

The Craigslist Printer from February 2018

When I bought my (used)backup Lulzbot Mini from Craigslist, it was $1000 for the printer, and a spare flexystruder head (which at the time was about $300, designed for TPU use). I paid the guy an extra $40 or so for the remains of his two Ninjaflex (TPU) spools. He had been posting the printer locally for several months without buyers, and I had negotiated the flexystruder head in with his original price of $1000.

Was it a good deal?

Kind of. New retail for 1.04 at the time was $1250, plus tax. The Flexystruder head at the time was $275 — it was the only way a Mini could print flexible filament. The Ninjaflex, was probably break-even, maybe trending towards not. Within 4 months, the glass bed cracked. I also noticed that the "scrub" phase of warmup, it would knock into the plastic frame that held the scrub pad. (And he must have printed a replacement frame.)

Fixing up the Craigslist printer

My 1.03 Mini was 18 months in, and its glass was still good. (I actually stole that bed to the Craigslist one, because the 1.03 was out of commission. It's now over 30 months in use, so maybe the seller didn't treat his 1.04 right. I printed a new scrub pad frame, and manually edited the startup gcode to rub on the pad correctly. (This is awkward, because I have to check a text file every time I tamper with the printer's settings.)

When I bought the printer from Craigslist, I did have the guy run through a sample print, so I could see warm-up and first layers. I just missed the scrub issue. (And forgot the printer came with a putty knife, single-ended pick.) It wasn't the worst deal, but 4-5 months later, Lulzbot announced the Mini 2, and I picked up a refurbished Mini 1.04 from them for $1000. No issues with that one.

I guess I still have the Flexystruder head. And the Ninjaflex. I never got around to swapping out a head to print it. I wasn't really sure what I wanted to print, that would be worth the time to swap the head in and out and change settings around.

Remember the Micro .25 SL toolhead from three months ago (December 2018)? In addition to printing normal filament, it can also print 3mm flexible filament. (I have two 1.04 machines, the refurbished "standard" machine, and the other with the Micro toolhead. The original 1.03 is still in the closet, waiting for me to fix the z-axis.)

Trying out Flexible Filament aka TPU

The orange "Cheetah" line of Ninjaflex was nothing short of magic. 230 degrees, just like PETG. The bed was 40 degrees. Printed very easily, with a little bit of extra stringing. With two shells, and 20% infill it was spongy but firm (for a solid object). The filament feels similar - it has a slight give, but firm overall. If you flick it, there's a bit of bounce as it returns to its semi-rigid state.

I printed a voronoi-style ball, about 3.5cm across - it bounces a little. The TPU also made for an excellent 20-sided die, with just the right amount of bounce.

The white Ninjaflex was probably a more generic TPU. It was much more flexible. Printing a solid object like the orange TPU, you could definitely tell it was a soft version...you could gently pinch the same model and it would collapse all the way through. It was still solid — I couldn't tear it apart with my fingers. The solid object would still stand upright and not deform. More stringing occurred than with the orange.

Swapping out TPU seemed problematic. It felt like it was harder to get the initial flow correct, and also difficult to swap back out to PETG. Normally switching from one PETG color to another PETG, is usually 60 mm of extrude to flush out the previous filament. After 120, there were still bits of PETG in the TPU. I probably won't be going back-and-forth between the two often, but it's much easier than swapping in toolheads (maybe a 30-minute process, and then another 30 to verify all of the settings are correct.)

Overall, the Cheetah was really good. I could see using that in designs. The more basic white was flimsy feeling. I'm sure it has a purpose, but something to design for.

Printing Miniatures

Printing miniatures is a popular topic for 3D printing. If you play something like D&D, you never have enough of the right miniature. "I need 20 goblins." "I need some plant-tree creature, small-size, on a base." Thingiverse has a ton of models. Hero Forge lets people design figures, then buy the figure's stl. (And then send the .stl to a friend of theirs to print, if they can.) My friend's Starfinder figure will be an example further down.

My Lulzbot Mini is a workhorse, but uses a stock .5mm nozzle. This means it prints faster than smaller nozzles, as it can push through more melted plastic.  A lot of other printers use .4mm as their standard nozzle...and people generally stop designing a "successful" print when they get it right...which means sometimes I'll get a model that doesn't print well on my bigger nozzle diameter.

A Finer Nozzle

Maybe you have a collection of nozzles, or you buy an extra nozzle off of Amazon for $15. (I've bought some Microswiss ones in the past.)

[The next paragraph will be something specific to a Lulzbot printer...]

Enter...the "Aerostruder .25mm SL Micro" tool head. It's released through the Lulzbot site for $250, needs a $50 converter to fit my Lulzbot Mini 1.3/1.4. You can guess it's a .25mm diameter nozzle, which allows for higher resolution. It also has a custom fan shroud to blow in air from more angles...allowing for better cooling and better resolution.

[...back to the regular broadcast...]

I swapped the new toolhead on one of my 1.4 Mini's, as a long-term switch out. You also need to flash the firmware so it recognizes to print with a smaller nozzle, which is a little bit of an effort. Once you load the default .25mm Micro profile, do NOT change the e-steps. The documentation and Cura is a little wonky....it tells you to copy your e-steps when you change your nozzle...don't. The new profile will have different e-steps, and it should. Imagine that you are shoving filament through at the same rate as before...but you have a smaller nozzle. It won't go through, and you'll have problems.

In Cura, I set up a new printer profile for this "new" printer. It makes it easier in the future if I swap back to a different toolhead. I went to my Octoprint instance, and changed the nozzle size there — if you extrude 30mm, you want that based on the smaller nozzle...

New Layer Heights

It also took some new habits, when it came to printing miniatures. You now have much finer detail available to you...I started printing at .05mm height, which was always so difficult with a bigger nozzle to do it right. (Fine detail for a .5mm nozzle profile was usually .15mm.) Honestly, a .1mm layer height is completely fine - you won't notice much of a difference. I'd recommend starting with .1mm, and seeing how you like it. (My experience is usually about 2 hours at .1mm layer height, including the trees listed below.)

Meshmixer

Miniatures typically have dynamic poses — pointing their sword, or a spell shooting out from their outstretched hand, which means supports. Normal supports can be okay, but sometimes you're missing details in key spots. I started playing around with Meshmixer - a free-to-hobby software released by Autodesk, which can allow for a lot of leeway when it comes to adding tree supports. I'll be talking about my experience in 3.3.15, in case you're reading this in a future that my assistance does not make sense.

A couple of tips...first, when you import an .stl in, first step is Edit>Align [Accept]. This will bump your miniature up into the correct plane. If you don't do this, it sometimes cuts off the lowest part of your model. Next, try letting Meshmixer do the orientation (Analysis/Orientation [Accept].). Hopefully it'll preserve your details better than the stock "as the figure stands".

The real work is Analysis>Overhangs>Generate Support. You may want to try Meshmixer's baseline attempt, before you do a deep dive. Some key settings: Density (how much tree support you want), and Post Diameter (how thick the tree posts are). More density = more support...but it can also introduce more scarring as well as make the support more difficult to pop off. Less support means that parts can droop. A thicker post can be useful - if you print a bunch of really tiny posts, it increases the chances of some getting loose and gumming up the whole print job. The adverse effect — you can't print the posts as closely together, regardless of density...which means you risk more droop between supports.

With a difficult miniature, I was running a density of 80 (a bit high), post diameter of 2mm (a bit thin)...and the Starfinder miniature looked a bit like this:

Export your figure. Now, keep in mind, this is no longer a normal .stl. Don't use supports - you've already added supports. If you generate supports in your slicer....it will not know that your tree supports are anything separate from the figure. It will add more, which defeats the purpose of tree supports.

Finally, I'd urge you to use a "raft" when printing. Usually when I print with Cura, I do a quick skirt. The drawback of trees is you're going to be printing a lot of small support bases onto the bed, and if any of them come loose....not good. The typical 'bad' aspect of a raft is that it needs to be pulled off of whatever model you're making...but that's great for these tree supports. It makes removing them that much easier, beyond making it reliable for the support's bases to stick to the raft.


Good luck!

Blender Gears

I've been using gears for a couple of different projects.

My first foray into gears using Blender, was using the Add Mesh: Extra Objects plug-in. It opens up another "base" object in the Add Mesh > Gears > Gear options. It creates a gear without the faces connecting the inside of the center hole (a non-manifold model). It's easy enough to create these faces - alt-shift each inside ring, the bridge the edge loops (through Control-E for the edge menu, then Bridge Edge Loops, or spacebar-find Bridge Edge Loops).

Breaking it down

• Teeth is...teeth - the distance from the inside ring to the outside tooth tip is Base+Dedendum+Addendum
• Radius is the distance from the middle of a tooth, to the opposing side's middle tooth
• Width is the z-axis height
• Base is the inside of the ring to the outside of the ring
• Dedendum - the distance from the "middle" of a tooth, to the outside ring (increasing it, also shrinks the hole in the center)
• Addendum is the tip of the tooth, distance is figured again from the middle of the tooth to the tip
• Pressure angle is going to the slant of the tips of each tooth - anywhere from 0 degrees (square), to a maximum of 45 degrees (if your addendum is a larger value than your overall size, you'll need to reduce the pressure angle to avoid bad topology)
• Skewness is used for helical and herringbone gears - also known as a helix angle
• Conical angle has one end shrink to a cone - 45 degrees leads to a tip, where 0 is no cone, and 90 is flat (practical uses - 0 to 45)
• Crown makes a crown by elevating the tips and mid-points of each tooth - a measurement of the z-axis difference between the base of a tooth to the connected top of tooth (not a hypotenuse, which would be longer)

Woodgears

That's just how to lay out the gear, once you get your gear designed! I used this perfectly fine option from a site called Woodgears...he does a lot of great woodworking, and the gear tool is useful.

Otvinta's Instant Gear

My current gear-making assistant is Otvinta's Instant Gear. Their branding is "A collection of tutorials & resources for 3D modeling and 3D printing enthusiasts". The useful part, is how it generates a python script to create the gears.

You can then go into a Text Editor window in Blender (usually on the 3D View, lower left hand corner), click on the "+ New" button, to create a new text data block...paste that code, and the press the "Run Script" button. You have the outside outline of two gears, both preset as meshes (back in the 3D View).

Figuring out bad meshes in Cura and Blender

One of my recent quick projects was making a simple "D" with a cut for a magnet. I was using magnets from my first project on Thingiverse, so I started by appending the "white" portion of that Blender file.

The benefits of appending, means that I get the Blender object, without having been converted to an STL. This means if I do edge loops, Blender already knows how to play with it...and I can use my current Blender startup file, instead of the dated one I used two years ago.

I switched into edit mode after the append, selected all of the planes of the magnet-holding portion, and hid all of that. I then selected the rest and deleted it. Looking back, I should have just done Control-I, to invert the selection. Oh well...un-hid the ring of planes and made it a cylinder (select an edge with Alt-A, F to create a face, repeat it for the other end of the cylinder. Renamed it Magnet.

Back in Object mode, I added Text, Edit Mode to make it a D. Object mode, Object>Convert to>Mesh from Curve/Text. I went back into Edit mode to scale the interior of the D in, so it would be a thicker letter. Boolean>Solidify>Thicken 1mm...Apply, easy.

It should be a simple matter just to Boolean>Difference the magnet out of the D...and it was. Opened the STL in Cura, and it was ready to print. But there's a red bit. Huh. (I'm posting the fix above it, the broken original beneath.)

It probably would have printed, but it also means something is wrong in the design.

Running Blender 2.78 on OSX, I re-examine the cutout...it's a slightly different shade. It's an easy enough fix, reseting normals in Edit mode. Mesh>Normals>Recalculate Outside. But it doesn't explain why this was a problem!

Remember my first steps? I used the magnet sizing from the STOP sign...and the cylinder part inside-outside normals were flipped. When I made the magnet cutout, by using those cylinder sides, it used those same normals. You can see the Magnet object that is off-color. (When I was doing work, I hid objects and never saw the color difference.)

Be wary of Boolean cuts with bad normals!

[The solution is: fix the normals on the "bad" magnet, so the colors look correct. When you use boolean on the D to remove the magnet shape, it will have the correct normals.]