3D Scanning and Printing

27 September 2017 | By Casey Evans

“Good judgement comes from experience, and a lot of that comes from bad judgement. - Will Rogers”

Assignment

Group Assignment: Test the design rules for your 3D printer(s).

Individual Assignment: Design and 3D print an object (small, few cm) that could not be made subtractively. 3D scan an object (and optionally print it).

Class Notes

Unless you're me you should always skip Class Notes. They're not organized or special in any way.

Other People

Make sure to have final project progress. Haha. Okay. I’m on a struggle bus. Arduino is a development environment, form factor, set of libraries and set of boards. GCC compilers. Using microcontrollers. MIT App inventor. Drones from scratch are “too much for most mortals.” Xin Wen – site broken, with 404s. He said to check local vs on the web. She had a good comment that the set screw walks up a bit when tightened so hold down. Neil comments: But not a “gym workout” to tighten screw or you’ll strip it. Clean iron with flux (whatever that is). Inside Out headquarters. Kreg made a video game thing. Fish made interactive furniture. Alexandre Kaspar (this is the guy who helped me) comments that tightening too loosely eventually led the bit to drop out by the end of the job and drag it away. Everything was okay though. EECS may have forgotten to put in a sacrificial layer. Otherwise the offset board will cease to be a reference surface. White dust is a good sign. Suggestion: make another one and try to program one with the other. Ragged end mills tear up/push around copper. You can use a straight edge to deburr it. Alex is working on an ornithopter so he’s thinking about a gear shaft. Adafruit for motors. Anna got it to work too! I may ask her about rechecking my programmer. Not sure if it being a programmer and not a programmee will be a problem. Jiaobao nyan cat “most creative and most annoying thing”-Neil. Vinyl cut the circuit. Anchor with a glue gun on things that will have a mechanical load. Use the transfer tape after (Honghao Deng). Tomas milled with 1/64th and then the 1/100th endmill. Requires everything to be flat and such. MBX-20. Honestly I’m not sure if all of these people are inspiring or demoralizing. The things they do would take me weeks—or at least, concerted effort as if that thing was my job. I have a job. It’s not this, but I would like to incorporate this into my job.

Last Week

Always wash board before soldering. And after because within a few months the oils will etch the board. For ugly spots—reheat, add some solder and then it’s good. Heat junction through braid but you have to add a touch of solder to saturate the braid. I didn’t think of that it seems counter intuitive. Power supplies have internal resistances. Check programmer again because it can sometimes program despite a dead board (ask Gavin perhaps). Offset layer is the most likely part to not be square. What if there was a placement device that you could look at in a microscope that will place and hold it…okay it’d be too much. Anode to cathode (alphabetical order) and line is always on cathode side. Hardware v software issues are hard to debug. With the wick, hold it by the plastic round thing. Neil says “they have a reason to exist.” Wipe off tip of mill with your fingers to get crud off. If you vinyl cut circuits, make the traces thicker. And use a white epoxy film underneath while soldering.

This Week

du means disc space usage. At least in Linux (or whatever operating system Neil is using). Maybe it's a git thing. He doesn't like people committing during class. Murphy's law--anticipate what can go wrong so it doesn't go wrong. 3MB images are too big. Use imagemagick.org/script/convert.php for compressing files and such. 100KB is a good number for web resolution. Recommend backing up the site (especially if you’re not using git). 10MB but not 100MB per week if that makes sense. 4MB movie makes sense. Mp4 most supported but it has patent issues. Use ffmpeg if you want. Set quality or bit rate in HTML5. GitLab note: don’t turn off all the messages. We are masters on our class levels, we are developers on the top level page. Make sure you’re watching. Neil won’t debug so make sure your problem is not a debugging one before you ask—or just ask your TA first. Or comment. After class make sure to check for the test issues for the top level and EECS section. It smells really bad in the class for some reason. Interconnect rivets for vias to “stitch” interconnected boards. Very important but least creative week is this week. Check tilt of bed. Take apart and put together again, perhaps. If you’re seeing uneven milling, that’s not good.

Generative design will be considered in optimization. Don’t ask questions that don’t have specific answers, or questions he doesn’t know the answer to. 1950s kitchen of the future—push the button to make anything. Microwave inadequate though. Just like 3D printer is inadequate. Subtractive machining versus additive machining. Subtractive-assemble parts. Additive-nested parts, not as much waste, helps when later versions would be hard to get in with tools to. Chuck Hull invented stereo lithography (laser to fuse stuff, “not office friendly,” sticky and messy). 3D printer tolerance test is a given thing. PLA-polylactic acid is plant based. “Carbon-negative.” Limited resolution 10/1000” for a voxel. Printers take hours to run. Digital rights management. Post-processing can smooth out the 3D print. We’re fusing together beads. So points are bad. There’s a minimum feature size to ensure that there’s enough material for beads to stick together and not make a sticky mess. There’s also a minimum gap. Overhangs sometimes need supports. Design around the constraints. Anisotropic-stronger in plane. Sparser inside to save material (and time). 3D printing is “not magic.” Building scale printing (wow) with concrete. Two photon is high resolution. Makezine. Using parts to create nanoscale circuit parts. RepRap, printers to make their own parts. Ultimakers are optimized to print quickly. MakerBot from RepRap originally open source. Bad mech engineering and gradually improved. Sindoh great for entry level printing. WOXes are 1,000$, automated leveling and settling, temperature control, camera, beautifully integrated, output not dramatically better but they are consistent. FDM modeling. Formlabs stereo lithography. Print the Legend. Form 2 very high resolution. Resins. Sticky, gooey and messy. “If not it can be imperfectly cured.”-Neil. Unbound powder supports bound powder. Can create cool nested structures. Something about laser metal 3D printing (million dollar dedicated room). Desktop metal is something similar. Metal injection molding. STL is the most common file extension. Ron Rael Berkeley makes ceramics from a printer—it’s not commercial. ASCII or binary. It’s a list of triangles. Use the right hand rule. You have to agree with the printer on units though. Can’t describe color in STL. Marching Cubes algorithm.

Make something small – manage printer. Needs overhangs or parts within parts (ie what a milling machine couldn’t do). Fusion can talk to a mesh. “Meshing is nasty.”-Neil. Meshing approximates by facets so there will be some loss of information. Zbrush, pushing pulling and twisting. Meshlab is fine. Normals in the wrong direction, too many triangles in one area but not in another. Sketchfab for sharing design. Thingiverse too can be used to share your design with the class.

Scanning gets points. Will need to make meshes and such. Million dollar xray tomography. Lidar measures phase, used in cars. Time of flight for each pixel. 3D System in between speckle and photogrammetry. One of the easiest things to use. Projectors can measure how patterns deform on an object and make approximations from that. Optical v geometric properties. Make a model that you can then manipulate. Google’s Tango phone is pretty cool. Hololens. Meta 2. Red camera Leila.

Group project-at what angle do you need a support? And other degrees of freedom.

Recitation

Many random thoughts: Link. Using nature as an example for design. Bring in and image and use spline tools to trace, hard to manipulate an stl. Split face – simulation tab, add constraints, apply a load, set material; preferences, preview, simulation, meshes may be asymmetric, calculations done on the cloud, visualizing load paths in a structure. Promote to body, box was bounding envelope, mesh bodies are hollow, extrude through. Optimat design – stress/strain about equal throughout material. Xdesign – design guidance = simulation, apply clamps and forces; using symmetry makes the sim time faster, able to preserve or exclude certain regions; understand that things need to survive additional load cases too (but not necessarily at the same time), show granular results to see how it’s optimizing. Git.amandaghassaei.com/vortexshedding/. Simulation samples in fusion (they acquired nashtran solver).

Training Summary

Training by Mike Spector. The EDS lab has 2 types of printers: the Sindoh 3D Wox (two of those) and the uPrint SE. The uPrint is apparently "messy but awesome" because it creates supports in a secondary plastic that melts away when it takes a bath in a solution heated to 70 degrees Celcius. You put your print in a cage in the solution and you have to use heat resistant gloves to do so. You leave it in the solution for about half the print time, or even 1 hour for every 3 print hours.

Actual printing advice: export your file as a .stl then for uPrint Add file->Add to pack->General tab->Print. For the Wox you just load, arrange and hit print. It is recommended to print a small version first to check structural integrity before moving on to a larger design. The base plates are removeable so keep that in mind. The uPrint at least needs you to pull the plate out and reinsert it between cancelled jobs.

Clarification on additive versus subtractive: cannot be created with a chisel. I think that largely depends on the chisel and the sculptor but I'll at least try to think of something much easier to make additively than subtractively. Also, I think buildings are always made additively, right? And they have scaffolding, so I suppose that makes sense. Brunellesci didn't really though.

Assignments

Individual Assignments

It's hard to think of things that can't be created by subtractive manufacturing that can also be created by 3D printing (ie not having huge overhangs and the like). After looking at this I thought it might be cool to make a bird cage with a bird in it since I really like Robin in Batman and it could be a symbolic sort of deal. But then I realized how that wouldn't work with the overhangs unless I used the same printer Anozie used, which is apparently sticky and gross when getting rid of the supports. I want to start small and then go big. Small steps. Ambition is for the lazy, because if you dream to big you won't go anywhere. At least, that's what I'll keep telling myself until I find myself doing something worthwhile. But in typing this I thought of a decent idea: a Robin logo with circular columns at different and uneven heights acting as pixel like things. That would be hard to make with subtractive manufacturing but pretty straightforward for additive manufacturing (I expect). We shall see.

I am checking out shapeways for ideas. Looking at past years projects makes me feel like overhangs can be much bigger than I am thinking and you can maybe even have a roof. I can't picture the machine printing it but I can keep checking out examples to see what seems possible. A bird cage would be cooler than a Robin symbol. This site has some 3D printed bird cage examples. Again, I can't picture the printing but maybe I should just try it and see what happens. One cage showed its hand--it is made of multiple parts. Mm, many of them speak of excessive support material. I feel that defeats the purpose of additive over subtractive modeling though, so my goal is to minimize supports. There is an interesting "bird cage necklace" file by einglis up for grabs on TinkerCAD that I'm going to test out to see the support load it requires. It asks for about the same amount of support material as actual material. I'm thinking this may just be a fact of the business but I'm not impressed. Also, TA Gavin was in while I was setting up and he says that the bird would probably fail in the supportless printer but the cage itself is probably fine. So once I create the person in a cage design I will try it on both printers. Probably.

I actually ended up just making a queen chess piece. Shown here next to the bird cage from Yeggi:

I was working in Fusion. Making a person turned out to leave me pretty stumped and I realized making a good one would be a project of its own so I looked for ones online but none of them were the defeated droopy posture I was looking for. I thought of a chess piece because when I put the first cage in the printer the supports ended up looking like a chess piece so it was on my mind. For reference here is that first piece with supports:

To print a chess piece I brought a picture into Fusion and traced it with curves on one side then rotated it 360 degrees into a body, which I then placed in the cage. This is the picture I found:

In the morning I will give it a go on the printers. Gavin suggested I scan a figurine to put in the cage but I'd have to see if I have anything worthwhile. My flatmate gave me something else she wanted me to try scanning. Maybe I'll scan my class ring. But for now I'll stick to the basics. The birdcage actually printed fairly well but the bird got disconnected from the stand leaving me with two parts:

This is a close up of the bird itself. I was impressed with the detail:

The caged queen looked pretty cool. You already saw the supports from the uPrint (though I never saw the supports specifically on the caged queen since someone else put it in the bath while I was in another class). The 3D Wox supports were much simpler:

Overall I think that the Wox produced a better product but I can see how the uPrint could be useful for heavier designs. Below is a side to side comparison of the two printer output for the caged queen.

While the scanner is being finicky and Gavin is debugging it I'm going to comment some more on the printers. When I imported the caged bird file Gavin suggested I try printing it all on its side so that it would print more quickly. Height adds time but it needed way more support that way so we reverted it to upright. Later when I was printing the caged queen I had a 1.5" base that listed 5.74 cubic cm of material and 3.64 cubic cm of support material. I scaled it down 75% and it became 2.48 cubic cm of material and 2.25 cubic cm of support material. So the support:material ratio isn't a linear scale. Thinking back I probably should have made the base of the cage a square because that's where chess pieces go. But thinking again, it may be apt that it's on a circle because it wouldn't know where to go from a circle space. Even though it's a queen which should be able to go in any direction it chooses.

Finally go the scanner on the path to success using the Sense 1 (Sense 2 was out for the count). My first attempt was on my flatmate's object, which I named boingy. It's a Chinese toy that looks kind of like a badmitton birdy but it bounces. The scanner hated it and I didn't have the patience to try over and over again. Below is a picture of the actual object, the first attempt and the second attempt. The first attempt I tried just rotating boingy and leaving the scanner in place. That clearly was a terrible idea and it decided that boingy only had one side. For the second attempt I tried rotating around boingy but it still decided boingy only had one side. Still, the "solidify" for the second attempt worked much better and actually made a cylinder like shape instead of a Hans-Solo-in-carbonate like structure. I didn't bother to save the .stl files.

After that I decided to scan a plastic cupcake. The original cupcake is shown below.

The cupcake scan results were much better than those of boingy. In the first attempt I just waved it about since going all the way around didn't do much for boingy. It turned out as you might expect.

The second time I tried much harder and spent a while swinging it around, trying to go full circle without it losing tracking like it kept doing for almost every movement (and even non-movements back when it wasn't working at all). I was able to almost get a full 180 degrees of cupcake. Still nothing worth printing but a better concept of what scanning can do.

But if we want to talk real scanning, Lincoln Laboratory has got some really cool projects you can see at their website.

Group Assignments

Niki set up the printer for the test piece. It was able to print it without much trouble. You can see that there are some artifacts like the string at the peak of the pyramid/cone, which I also saw in different parts of my design. This was due to melted filament not entirely cutting off before moving to another piece and getting stuck to part of the build (or another build in the same set). You can also see where two close and thin extrusion lines start to join at the edge.

I also used my design to test differences between our uPrint and 3D Wox printers. What I noticed was that in our default settings the uPrint tended to use more support material and work more slowly than the 3D Wox. For my design the uPrint seemed to create lots of "fuzz" from support material even after the bath in the warm solution. The fuzz was easy enough to remove with a probe but the 3D Wox, with centralized and limited support, came out much more cleanly. From my design you can also tell that the Wox can tolerate pretty high angles of overhang (look back to where it put the supports for caged queen--it was nearly horizontal before it caved in and added support). Since the uPrint added more supprt I'm inclined to estimate that it isn't as self supportive but it could also very likely just be the algorithm used to determine where to place supports.