Computer-Controlled Machining

11 October 2017 | By Casey Evans

“Everybody wants to be a cat. Because a cat's the only cat who knows where it's at." -Aristocats, Disney

ShopBot in the IDC Shop


  • Group Assignment: None.
  • Individual Assignment: Make something big.
  • Class Notes

    Other People Hasso-Plattner-Institut visiting professor. “Xerox Park is a great place to have been at.” Mobile devices->hardware. Stephanie Mueller student, laser cutting. Dr. Patrick Baudisch (?). Ars electronica-humans from users to used. Interesting ethical consideration. Democratization of fabrication.

    General programming in recitation. Not sure if I should go. Diana Zwetzich. Common mistake she didn’t make: letting the header rest on the board. I’m not sure if mine is like that but I should be able to tell with the prototype board. Also make sure to check for shorts. Pins fight – “who can make the most current to win the war” -Neil on shorted pins trying to determine the mystery voltage. ATTiny44a, check electrical characteristics. You can blow the processor if you go half a volt above the voltage supply. You can name nets instead of drawing them. Diana is doing wearables for her final project. Neil did work with Levi’s to sew circuitry in clothing (E-broidery: Design and fabrication of textile-based). With Steve Mann and Thad Starner. Tons of nerds. Nervous System’s way of producing fabrics. Based on Fuse 1 from formlabs. Selective laser “sinering.” Unfused powder is support. Can you reuse that powder? Because that would be great. Test prints with formlabs (they’re 0.5mi down the road. Neil uses MyPaint. Flexures from laser cutting. Seammetry. Dr. Baudisch is working on software to automatically compute flexures. Otherlab Experiments in Collaborative Design. “Beautiful and expressive.” Flexural clothing? Cylindrically bendable, not spherically bendable. That’s just a technical issue, you could make multiply bendable materials. But how transparent would they be, lol.

    Ben Yuan. Ahhh, I know this guy. He’s in our section. He used KiCAD. Usually people think it’s excruciating but Ben liked it. It’s apparently not well integrated. KiCAD is free. KiCAD has a design rule checker. Ben used a ground fill, which is pretty cool (filled the board almost everywhere as a ground plane). Only downside would be at high frequencies where the big board would create some capacitance (I think). Right now mods won’t do the insides of things. That is, if you made an O it would be one circle, not two consecutive circles. Neil said he’s going to change that. Ben said placement of objects was a bit less forgiving. Solder paste! Toaster ovens and frying pans are great but then don’t use it again for food. Just google “reflow frying pan.” Resonator – lower quality crystal with capacitors built in. RC clock in ATTiny is temperature dependent, only good to about 10%. Ben is doing a hardware password storage unit that emulates a keyboard. Cryptographically strong randomness is nontrivial. Amplify Johnson noise. Microsecond timing between button presses. Using low order bits (like how quantization noise can be modelled as random noise). Width and height in thousandths of an inch (ex 1206). KiCAD has a 3D tool. Human entropy. Measuring capacitive charges (touchpads). Ben is a 4th year grad student working with autonomous vehicles.

    Diana Yan. “PCBs don’t have to be squares and lines don’t have to be straight.”-Neil. I agree. This board is very cool. 10 mil lines work great. She used default and it worked. Skinnier then resistance goes up though. Cut traces and solder jumpers is perfectly normal. She drew by hand. It would be better to do in Paint or something I think. “That was fun, I don’t think I’ll ever do it again.” Ooh, like what I was thinking you could get the Eagle footprints of everything and port them into Paint and make the connections there (being sure to not resize). Her final project: Matt’s thesis in prefix-whatever that is. She wants to make a tent. Smart Citizen is a similar project (logs environmental data). Intentionally jamming cell phone signals is illegal but not really that hard to do.

    Kate Weishaar. Autorouting liked vias. They aren’t very good. Use the class aliases—I’m not sure what that means. Vinyl cutting can be much faster. Shears and zip saws work to cut the outline if you want. Final project: mood controlled light. Kokopelli is like openSCAD. Nema stepper motor. Nxp tiny logic gates. Safina did a lion head shape. Bead of glue gun glue on heavy components on vinyl cut circuits will stabilize the thing. Tomas “LUFA, it was super easy.” Tomas is too smart.

    I should look through my digital logic lab booklet to learn more about using the buzzer. Will I have to program a frequency? I’ll have to find the thing first.

    Burhanazeem made a 4 layer board. Super cool. Copper has an electrically conducting glue.

    This Week

    Large format machining. “Make something big.” One of the best assignments. Yamnitsky made a table. Larry Sass MoMA house is neat. ShopBot, Ted Hall made these and they’re what we’ll use. 20K$. PCB but big. Onsrud, stiffer 200K$ but aren’t that much different than the ShopBot. Tormach, not as big but stiff. Make it yourself machines are on the rise. Rigid foam insulation from Home Depot. Architectural molding. Special foams for milling too but they’re expensive. Veneer plywood ($100 for 4x8), MDF is wood fiber in a binder (homogeneous, heavy, bends, bad bindings make people sick). MDO has ply core and MDF outsides. “Hot dogs are made of you don’t really want to know all stuck together. This is the hot dog of wood.” -Neil. Sand, paint and seal still possible with oriented strand board (hot dog wood). We each get a 4x8 sheet of OSB. HDPE what cutting boards are made out of. “It mills beautifully.” Quite expensive. Between acrylic and foam. Can also mill aluminum. McMaster-Carr is expensive but they have materials. Cycle Start, DSH and Proto Labs are like PCB shops—send them the design and they’ll cut it and send it back. Nice. Makes me think of CustomInk. Endmills move in three dimensions. Tools may or may not be center cutting. Center cutting can plunge, you can use it like a drill but it’s not as good at removing material. Flutes are the spirals on an end mill. Can go counter clockwise or clockwise. It will either push material up or down. Down cut pushes down, clean top. Up cut for clean top cut, down cut for better removal in deep cut. Don’t need a ball end mill for a curved surface. Flat end mill cuts create steps, ball mill creates a smooth edge. But a computer can do that with a flat end mill and has better chip clearing too. Hot wood chips can cause a fire in the dust collector. Ball end mills can sometimes get a tighter radius of curvature. The Machinery’s Handbook has speeds and feeds information. Depends on material, machine, etc. Not fast enough you rub rather than cut. Too fast and you stress the machine. Chip load – feed rate/rev per min *flutes. Each pass 1-10mil removed. Step over – passes next to each other can leave a ridge. Lubricants for machines. Heat transfer help and moving the mills. Ceramic and glass use abrasive slurries. Computer controlled grinders are also a thing. Fixturing – double sided tape in PCB. Vices here. Or C-clamps, woodworking clamps (not as strong forces though), wood screws and cordless screwdriver (not where you’re machining!). Expensive ones have vacuums. Gravity/weights for waterjet cutters (not much shear force to fight). Carpet tape. Encapsulating metal in a low temp melt metal (interesting). ShopBots have three layers – underlay, sacrificial layer, cutting surface. Dust is bad for you, not as bad as carbon fibers but still. Maybe do flexures or kerfing for the box. That’d be sleek. But you need nicer wood apparently. Darn. I’m not about that. Kirsten Petersen. Kamakura fab lab. Japanese joinery. Runout is the uncertainty in the tool’s expected origin over time. Climb machining puts more force on machine but clearner surface. Conventional machining is less clean but puts less stress on the machine. Trochoidal milling, each flute spends less time in the material, smaller angles, can mill harder materials faster. Coolant creates thermal shock, trochoidal milling allows limiting cooling and helps tools last longer. Rough cutting, 3 axes with crude stepover, 3 axes with fine stepover, picture on main course site. T boning is an imperfect corner, so you compensate for that by making bubbles at the corners. Leave little tabs at the bottom to hold it together until you’re done cutting. Cut air first. Then do sample cuts. Try laser cutting in cardboard first. VCarve works. Fusion has HSMWorks. Rough cutting and then fine cutting to save time. HPGL Rolands. ShopBot is pretty open. G code works on just about everything but dates back to sewing machines. I’d be interested to see how mass production in clothing works nowadays. Safety serious this week. Student died at Yale. Hair caught in lathe. EHS may have required training. This is all so much more intense than Habitat for Humanity. Fingers cut off, cuts, burns—be careful. These things can launch projectiles. No loose clothes, no loose hair. Never reach into a powered tool. “The pretty looking thing is a hot razor blade” -Neil on the metal spindles that come off a thing. Find the emergency stop. Never use these alone. Don’t do these things when you’re tired—but there are undergrads here, how is that possible? “The machine will make happy little grinding sounds or painful screaming.”-Neil. Hold down screws nowhere near the tool path, but wear safety glasses anyways. Recitation later on welding. If I have extra wood I will certainly try some flexible structures.

    Training Summary

    Abbreviated comments transcribed from handwritten notes: join vectors, merge, toolpaths, pocket=eats the wood, cut depth, stick to conventional because the ShopBot doesn't like climb mill, select end mill, down cut bit 1/4" probably, calculate, preview toolpath, create profile toolpath, add tabs if you want, save toolpath to send to ShotBot postprocessor (in inches), turn the machine on, press the blue reset button, screw it down, air cut or do holes in design first to guess where to put screws if needed in center of design, brass screws-not steel, nail gun and plastic nails are another alternative for 1/8" stuff, use wood screws and watch for pulling the piece up instead of holding it against the sacrificial layer, you could use an edge finder, with spiral up there will be some splintering and tear up but more chip and saw dust clearance compared to spiral down, compression bits are great (best of both worlds), use two wrenches to tighten the bit (crank down pretty good, especially with a spiral down), page up and down and the arrow keys control the bit head, using the bottom left corner origin is good practice, zero your axes, job size, z standard is at the top of the material, use the plate and clip on to zero the z, it measures twice, kill in task manager if the software crashes, file->parts to file, start, start on machine, hit okay, sawdust holds okay in larger pieces - you maybe don't need tabs, use sand paper and rasps to clean up, flexures create horrible torture splinter chairs, the IDC shop has orbital sanders (which are the bomb), don't bring in computers because they will eat sawdust and die.


    Individual Assignments

    Okay. The assignment we've all been waiting for: make something big. I'd originally intended to just make my box but since our lab time is so limited this week I will try to be efficient and try a few things. I want to test out flexures, even though they aren't supposed to work, a box design, and a bookcase, since I want one near my desk (which is currently just a table). I plan to model it all in cardboard. I originally planned to make a bookcase that looks like this image from Google:


    But then I decided it wouldn't fit nicely into my room. I thought about making a jellyfish bookcase since I love jellies but that also isn't very easy with wood and books. I'm determined not to make something perfectly rectangular though. Just to branch out a bit. I thought of making something like this to hold books:


    But then I felt I would feel bad making it carry things. But I should probably not care about that since it's not alive. But it would still possibly creep me out since it would have to be child size and children make me uncomfortable. Maybe a dragon to carry books? Or a lion? That could also help since four legs would definitely have better stability than two. Or maybe a cat with a top hat. I think a cat with a top hat would like to be carrying a bunch of books. Perhaps she is a librarian. That may be a better project for casting/molding. Oh! To not feel so bad about whatever I make carrying my books I will make it be more of a "World's Strongest ___" and look like they're supporting it on their back like a weight lifter. Or I could just keep that story and make the simplest/most logical design. Cats are actually a bit too small to carry the amound of books I want without looking a bit off. I think I'll do it anyways. I found this source picture on


    The full set of silhouettes is here. Using my iPhone level and a heavy textbook on my thumb and forefingers (mimicking cat arms), I found that 30 degrees is a good angle for the arms to feel stable. Furthermore I predict that the bookcase will be stable if the cat's arms are at least as long as my hands. So I drew it up in CAD. Or tried to. Instead I kept tracing the wrong parts of the cat and had to go into PowerPoint to create a more personalized reference image:


    Eventually I was able to get a respectable CAD model:


    And even a cardboard laser cut version:


    Which helped me to realize that I'd made my supports too small. When I got to the shop I was further advised to make them the exact same size as the holes so that I could sand down as necessary and avoid loose joints. I'd seen from my cardboard model that loose joints created quite a bit of instability. I felt that the rigidity of the wood would help to counter that on its own as well though. In the shop I followed along with Gavin and Vibha Agarwal. We needed to do much of the items discussed in training like join vectors, rescale drawings, rearrange components, and set toolpaths in the proper order. On of our toolpaths was previewed as the following:


    We zeroed our axes and did some air cutting before running a short test run with a few of each type of cut we planned. Then we executed the rest of our cuts. After cutting, sanding and malleting into place I was pleased with the resulting cat bookcase:


    That was basically it. I didn't get to work on my final project like I wanted to this lesson but after a good conversation with my roommates, we thought of a great option for my project: a fancy lamp for our dim living room. More to follow on the final tracking page: Link. Also, to further illustrate my reasoning for making this bookcase, here is a before and after of my desk. Notice all the stacks of papers that finally find a home in the cat's welcoming arms.


    Group Assignments