Machine Design
8 November 2017 | By Casey Evans
“There's no I in try, but there is a why.” -Casey Evans and Jon Schiller
![[sm_parts.png]](img/sm_parts.png)
Assignment
Class Notes
Other People
Admin note: manage image sizes and video sizes. Also manage time well. Don’t come in late.
Erin Genia seems to have a page style I can relate to. Rotating kaleidoscope in the shape of a morning star. DC motor appears as a coil. Only need a MOSFET if you’re only going in one direction. H-bridge required for current going in two ways. Kaleidoscopes don’t need much torque though (10,000 rpm is a lot). Gear head motors for torque (bring down maximum rpm but keeps torque). You can just use teeny pulses to keep something moving. Not a stepper motor (which has specified speed steps) but it’s like poking something. It doesn’t maintain much of the torque though, which is fine for a kaleidoscope. Don’t let out the black smoke or the magic stops working! “Quick term development” is the point of this class. To be able to make new rather than trying to maintain an old part. That kind of goes against my usual mentality but that may be the point. Thick tracers are better because they have less resistance. They’re only limiting at high frequencies and narrow routing. Morning star is a traditional Dakota design. Anpa O Wicahnpi. Erin is an artist. Can’t all be on one page. Need an index. Her planning was pretty good. She wrote out all the different weekly projects and their contribution to the final project. Super bright LEDs. Driven with a MOSFET on the other side. Ohhh—this is what I will have to do. Keep current limiting resistor. PWM LED for brightness control but not current control. Okay, glad I came to class today. LEDs can be put everywhere in the spectrum but not equally well, but it’s okay since our eyes don’t see well enough to notice. Digikey may also have good LEDs. They sort by brightness and such. Daniel Marshall. Go measure 3D (microscribe portable CMMs). I’m curious about the artistry lost/gained by pairing humans with machines. Adaptive printing, adaptive cutting, digitizing. Stefanie Mueller. Software messages tend to say fail but it’s at the boundary software to hardware, which makes it really a hardware problem that’s ‘hard’ to solve. Haha. BLVCs? Driving servos. Delay us is also a thing. I’d like to look into the delay routines. Kyung Yun. She’s making a mask to make bubbles instead of speech. “I can say with confidence that that’s never been done as a final project before” Neil. Kyung made cool molding stuff. Oh. I was supposed to fully design the boards. Okay. “In the assignment the words are carefully chosen” – Neil. This makes sense. There aren’t many of them so they should all be important. I didn’t really catch that but I will 100% be doing that for my final project which will have outputs and inputs so I can maybe wait on that. Or after I finish debugging the metronome/music box I can reprint it as an OCARINA!! Woah, I just thought of that and it sounds awesome. I could have buttons for the notes. Hot glue plus buttons plus a cast ocarina…wow, this just keeps getting better and better. Chastity Li. Adversarial maze (can’t just do it once and be done with it). Depending on how you turn it you get a different maze. “Electrical tape is relatively evil. Don’t use it for anything. Between two weeks and a year it starts to turn into a gooey mess.”-Neil. Tin thin braided wire and then create two loops to hook to each other and use heat shrink to finish. Hardware description language to describe the board. Not really recommended but Neil likes it. I think it’s like VHDL from ECE281 (undergrad course). Resonator is for faster 20MHz stuff. Music isn’t a science, it’s math and art so as long as the meter is close, I think my metronome will be fine. $1-$100 servos. HobbyKing, Jameco. Small enough to turn a speedometer to large enough to rotate a table or something. Sam Chin. Not here. Jiabao Li. “Sorry is this a final project or just this week.”-Neil. My thoughts exactly. It’s okay to make one board and reuse it and make daughter boards with peripherals as needed. Goal to make an eye tracker? No that’s just the input device. “Ambitious agenda for a well funded startup.”-Neil. Meta vision. Lots of money just to get to a working stage. Priyanka Chatterjee. Absent. Loads of absent people today. Akshata Krishnamurthy. Temperature controlled drink heating devices. Sands Alden beautiful interactive table. Triage and focus. Simple resistive heating is fine. Anything extra is extra credit. I think I need a few days of just playing around to make this class great. Maybe I need to dedicate Sunday afternoons to this class? McMaster-Carr heat cables, Nickel chromium heating wires. You’ll need way more current than you can get from a battery. My guess is my LED room light will also need more current. Maybe I’ll just go HomeDepot shopping one day. Don’t use .mov – integrate in html5 native video. WebM, Ogg, mp4. Use mp4. -1 it picks one dimension if you specify the other. Cast resistive coil in drystone, run a current through it and you’ve got a heated tile. Richard Liu. Alex Kaspar. BLDC motors look like Y connected resistors driven by a triple half bridge. FLIR camera. Agnes Cameron. “Oh this is adversarial.” -Neil on Agnes’ page. Nrd52 has a built in radio. Sam Calisch. BC832.
This Week
Jake demo. This week is a group project. At a minimum reproduce the demo machine. Linear or rotary 4 degrees of freedom. One person two weeks to design and make. 2 days to redo. 1 day no sleep. Each group is 16 people. “So a few hours”-Neil “You multiply, not divide, multiply.”-Student. Haha. Machine design. Jake built a page. Git lab has loads of great collaborated tools. This is meant to be layered, not integrated (unlike final project). Stages. Check out machine building class webpage. Bantam tools, formlabs, handibots. Nadia and Jonathan Ward. Modello. Anyone can make it if you’re Nadia and Jonathan. Others had many troubles. Zahner, James. Curvy parts of gear-y buildings. Object oriented hardware. Start with mechanisms. Annotated tour through McMaster Carr page. Guide shafts. Ball bearings hold loads. We will make parts. Alex Slocum, design guru. Precision machine design. Many interfaces to close loop between end effector and work piece and each one adds error. Fablight. 4/1000ths of an inch precision. Stacking of error budget. It’s like an electrical circuit. It must be closed on both sides. Basis of this week is Jens Dyvik. Gave Obama’s grandmother socks. Pick up Nadia’s idea of a modular machine. mtm.cba.mit.edu/machines/science. Drive with lead screw and a nut. Rack and pinion is a gear on a tooth rail. Sawtooth no good (off-slam-off-slam)(discontinuous loading). Rolling motion is better. “Fairly subtle tooth design” will work. 3D print magnetic filament to make his own motor. High density polyethaline (HDPE). It’s pretty sturdy but you’d want something a little denser for a real product. Mills rack halfway through – it’s part of the structure. Delrin has some elasticity, fairly low friction. Hoover dam 100 ton turbine, thrust bearing (can’t use ball bearings, too heavy). Linear to non linear. Two similar materials get spalling but two carefully selected dissimilar materials can slide past each other well. Good to make the whole design parametric. Humphrey is a ShopBot sized machine. Ultimate test: can a machine make more of itself? Run a job that machines the surface of the sacrificial layer so that it’s all even in the coordinate system of the machine. Trochoidal milling to cut steel with aluminum. Aluminum milling on a ShopBot type machine. There’s a lot of faceters, in the future it will be flexural snaps to make it more parametric. Don’t use Jake’s machine for heavy loads. Light show machine, cake frosting machine, cutting machine, whatever you want. We’ll be using the TinyG. It’s only worth about $10 but it costs $100. Next week is sensors. Then I should really be starting on my final project. Shaper tools founded by Alan. It’s really cool! 3 ball bearings in a kinematic mount (from machine in a briefcase). V grooves only all lined up in one way so you can get micron tolerance. That one machine can use different end factors to do 3D printing or cutting. Open loop you send instructions. But if you’re near the stall speed of the motor it can get messed up. Closed loop measures distances. Linear encoders on rails or something. Usually a bad idea to trust the stepper to do the proper number of steps (open loop). Huge chunk of steel for stability. Waterjet curves around corners. Head changes shape based on how the tool is moving. Bang bang control. Turn on motor to get it moving and turn it off when it arrives. Generally a terrible idea. Stepper’s positioning can’t be trusted under heavy loads. Different from DC motor. PID control (proportional integral derivative – just sounds like words to me). Feed in error terms, derivative of error, integral of error to get a good correction system. Three delays with three coefficients. Many generations have gone into tuning. Derivative too low it’s slow, too fast it rings. You can find coefficients somewhere. Better than PID is PIDA (adding in acceleration to smooth acceleration too – decrease jerk). Think of slowing down before a stoplight. Which means you need to anticipate things. G codes. TinyG, G for G codes. Photoplotters and sewing machines. The G code interpreter needs to know a lot about the machine to work though (G code interpreter). Networking is becoming more common than interpreters though. Han use handshakes to get nodes to talk to each other. Motors can say “I’m working, I’m done.” Can broadcast, can use time stamps—and other things. Gestalt Framework for Virtual Machine Control Ilan Moyer thesis. Driver – edge finding, toolpath calculating, and all that. Sounds challenging. User interface for machine. Ex. Chilipeppr. Motion systems, motion control, path planning, user interface. Most conservative thing for this week would be to copy Jake. Or one linear and one rotary axis (laser cutter). And many other things, I can’t keep up. Moving mirror for light shows. Goal is a hello world—“something that does something.” Meant to be a real quick put all the parts together. Neil says it’s fine if we have 4 pen plotters though. Wait—like calligraphy? That’d be pretty sweet!
Boards and labels. Label tasks. Kanban sticky notes on boards now visualizing what’s left to do. Milestones make dates and assign responsibilities. Make a group page for the whole assignment and on your individual page describe your individual contribution. Don’t overshoot it. Keep it bounded to finish it in a week. Bring machines for a machine show next week. To Jake: “Do you want to add anything?” Jake: “Uh, no.” Honesty is fun. Learn about group planning, rapid machine building, fab lab working together, machines making machines confidence boost.
Training Summary
No training. We met with Jake on Thursday and he somehow convinced us to make a delta picker. Other TAs helped us along the way. I went to the welding demo and the water jet/sheet metal recitation. Pretty cool. I saw spot welding, TIG, WIG and stick welding. I got to try MIG welding.
Assignments
Individual Assignments
None.
Group Assignments
For the group page see here: link.
In my individual participation I began by participating in voting on the idea. My first choice of pen plotter was not selected but we ended up doing a pen plotter anyways just on a delta picker instead of the given design. Meep. I then joined the end effector team (hence the pen plotting). Others wanted to do a claw or something but I felt we should get a moving working machine before worrying about what we put on the end of it so I switched to manufacturing. I worked with Xin Wen to redo the design work so that flaws were fixed and the design was flattened and ready for making toolpaths. That came to a standstill when some of the first pieces we pushed for manufacturing got stuck with issues and we had to wait a while for Gavin. Gavin wanted to approve our design before taking us all the way down to the architecture shop where we could machine our HDPE. So I became the self appointed webmaster and set up the web page and started collecting/soliciting information and pictures from people. That is still in progress. Or it was when I wrote this. Now it's finalized. I've also helped out with sanding of parts 9 and 10 to fit in the press-fit, analyzing problems and coming up with solutions to mechanical problems (like how to connect pieces), locating tools, and helping in finding food for hungry students. I also helped assemble some of the arms, fed design parameters to the coding team and assembled the end effector with some rubber bands and a Sharpie. Later I helped threading holes for screws and stringing the arms.