2 ms·

Using Python to vectorize artwork for PCBs

37 points by zdw 4y ago

thebeardisred 4y ago
For anyone interested in more details about the underlying CAM (computer aided/assisted machining) works I did a write up a couple of years ago - https://brianredbeard.com/blog/the-trouble-with-gerbers/ https://brianredbeard.com/blog/the-trouble-with-gerbers/
magicalhippo 4y ago
One might want something better than a fixed subdivision of the Bezier curves. Anti-Grain Geometry has a good writeup[1] on how to do that. A quick search revealed a JavaScript library[2] implementing the Anti-Grain solution, for reference. [1]: http://agg.sourceforge.net/antigrain.com/research/adaptive_bezier/index.html http://agg.sourceforge.net/antigrain.com/research/adaptive_b... [2]: https://github.com/mattdesl/adaptive-bezier-curve https://github.com/mattdesl/adaptive-bezier-curve
- phkahler 4y ago
  >> One might want something better than a fixed subdivision of the Bezier curves. In Solvespace we convert them to piecewise linear (PWL) approximations using a chord tolerance setting. If a straight segment deviates from the actual curve by more than the tolerance it is recursively subdivided. C++ code under GPLv3+ for that is available in my favorite file: https://github.com/solvespace/solvespace/blob/f7415048a5769c78fc4bb02afa426d101059374d/src/srf/ratpoly.cpp#L209 https://github.com/solvespace/solvespace/blob/f7415048a5769c...
stuntkite 4y ago
Someone once asked on twitter how to generate a PCB trace of the Big Wave Mt. Fuji print. I thought it was an interesting problem and wanted to solve it with Voronoi stippling, halftones, and some other things. I might have been drunk when I did it because some of what's in the readme doesn't make sense. I think possibly I was thinking that it would be cool to do something similar with point clouds created from photogrammatry or real or generated environments using Meshroom. I could also be just making excuses for why I don't understand why I put Meshroom in the README. Regardless, that would be pretty cool and I might try that idea later. I think it's a neat technique though. I made a repo here for anyone that's interested. https://github.com/millerhooks/jacktrace https://github.com/millerhooks/jacktrace It made a few really neat examples. I guess I didn't add my favorites to the repo. This one is pretty fun though. https://github.com/millerhooks/jacktrace/blob/main/Assets/Images/Test0001/bigwave-RedBG-MaskedExample.pdf https://github.com/millerhooks/jacktrace/blob/main/Assets/Im... I only spent a few hours on it but have recently been wanting to readress it. Especially grouping and connecting the vias (stipples) to traces and apply halftones for colors.

37 points by zdw 4y ago

For anyone interested in more details about the underlying CAM (computer aided/assisted machining) works I did a write up a couple of years ago - https://brianredbeard.com/blog/the-trouble-with-gerbers/ https://brianredbeard.com/blog/the-trouble-with-gerbers/

magicalhippo 4y ago

One might want something better than a fixed subdivision of the Bezier curves. Anti-Grain Geometry has a good writeup[1] on how to do that. A quick search revealed a JavaScript library[2] implementing the Anti-Grain solution, for reference. [1]: http://agg.sourceforge.net/antigrain.com/research/adaptive_bezier/index.html http://agg.sourceforge.net/antigrain.com/research/adaptive_b... [2]: https://github.com/mattdesl/adaptive-bezier-curve https://github.com/mattdesl/adaptive-bezier-curve

phkahler 4y ago
>> One might want something better than a fixed subdivision of the Bezier curves. In Solvespace we convert them to piecewise linear (PWL) approximations using a chord tolerance setting. If a straight segment deviates from the actual curve by more than the tolerance it is recursively subdivided. C++ code under GPLv3+ for that is available in my favorite file: https://github.com/solvespace/solvespace/blob/f7415048a5769c78fc4bb02afa426d101059374d/src/srf/ratpoly.cpp#L209 https://github.com/solvespace/solvespace/blob/f7415048a5769c...

stuntkite 4y ago

Someone once asked on twitter how to generate a PCB trace of the Big Wave Mt. Fuji print. I thought it was an interesting problem and wanted to solve it with Voronoi stippling, halftones, and some other things. I might have been drunk when I did it because some of what's in the readme doesn't make sense. I think possibly I was thinking that it would be cool to do something similar with point clouds created from photogrammatry or real or generated environments using Meshroom. I could also be just making excuses for why I don't understand why I put Meshroom in the README. Regardless, that would be pretty cool and I might try that idea later. I think it's a neat technique though. I made a repo here for anyone that's interested. https://github.com/millerhooks/jacktrace https://github.com/millerhooks/jacktrace It made a few really neat examples. I guess I didn't add my favorites to the repo. This one is pretty fun though. https://github.com/millerhooks/jacktrace/blob/main/Assets/Images/Test0001/bigwave-RedBG-MaskedExample.pdf https://github.com/millerhooks/jacktrace/blob/main/Assets/Im... I only spent a few hours on it but have recently been wanting to readress it. Especially grouping and connecting the vias (stipples) to traces and apply halftones for colors.