i drafted a mesh tesselator to generate triangles for stls from parametric surfaces!!!! it was super hard to do, so inhibited. to make it easier it was made on an offline computer. it’s like 1989, making 3d triangle heightfield meshes from coordinate functions with hand point projection by writing bytes to a framebuffer device (oops) it turned out it’s easier if i don’t work with the visuals at all, less inhibition maybe less clear what i am troubleshooting (more likely reduces power) or doing

i/we tried to make something real before the day was over its midnight 30. i will be tired tomorrow. i didn't quite finish but it may be usable. extant issues: - holes in the model may be generated if there are aliasing artefacts in the approximation function. using a euclidean approximation function removes this. the tesselation should be reviewed for correctness and upgraded to inform neighboring triangles, this would double the speed. - the surface normals may not be consistent and correct across chunks yet. i used smooth shading in the attached image to depict this. - it is possible to do more efficient tesselations. it at least may be fine to subdivide some triangles to only 3 rather than 4. - we think the code would be more organised if the two subdivision approaches were normalized as maybe the commented interface near top of file - a little more organisation etc needed in general celebrations: - it makes a solid helical gear now! the example is triple-helical. - i fudged in automatic detection of degenerate points to the tesselation so it can handle poles as well as the rest of a surface - it outputs a .stl file!

_it doesn’t work at all yet_ the code got messy trying to push thru to making a physical object quickly i think i need to understand gear teeth better so one gear drives the other it seems to get stuck at the peaks and troughs which maybe aligns with me not designing them but just connecting the curves :) it’s cool to print the helical shape. i used a library makerspace i hadnt used before so it wasnt the same printer that made the working bearing. the planets are stuck into the ring and sun. the lasercut is a much easier and faster test, it cuts in 29 seconds compared to printing in an hour.

0922 looks like i’m in edt-1, aug 3 2024 earlier figured the key to gears not binding would be getting the rotation per tooth to be greater than the angular tooth width so that another tooth would come in contact wrote a little code, found the derivative of the involute function (tan^2) and by substitution the inverse (1/dinvolute(involute^1(angle))), it looked at first like if i brought my teeth lines down to the base circle that i’d get more rotation than tooth width, but - i have to figure out how to align gears now with profile shifted teeth (first) - when the teeth are farther down there is no room for the opposing tooth so i get to draw the spaces between teeth (as well as figuring out how much space is needed) which will make code more general (second) - separating the teeth means less involute curve to rotate the gear so back to the drawing board a little after that it looks to me like the gear may indeed work if printed with parts free to move (unsure), but that the lasercut approach does need this fix. the thought is that the fix would make the printed part stronger by giving more contact surface. 0928 1548 aug 8th oklahoma or west ummm so much changed may be hard to finish take careful