I've added support for specifying the media size, which should help with A3 sheets or similar (sorry about the inches-based units though). Often after cutting a small stencil there will be an odd-sized piece of media left over, so those pieces will become usable. I think the media size is used for the initial slewing of the media all the way forward and back, possibly for registration-mark detection.
Also, cutting a border outline can now be disabled by specifying a border of 0,0.
[quote author="teletypeguy"]One option to minimize the travel time after training cuts, and to speed up in general, is to scatter training cuts around the perimeter of the board, just outside the final boundary. Make a training cut, then cut a half-dozen nearby cuts in that orientation, train another orientation, keep cutting, and just work around the board. [/quote]
Yes, I'd like to do something like this for the next version. Seeking all the way back to y=0 results in running time quadratic in max(y), which is not acceptable, and it can potentially result in larger accumulated errors, though in general I'm surprised by how good the retracing in the y direction is. Errors don't accumulate in x, but one would think any slippage in the paper rollers would start to degrade the positioning in y.
Perhaps I'll just leave a trail of training areas along the edges, spaced by an inch or two. That way each two-inch block along the y axis will be completely cut, never to be revisited, and errors will not accumulate.
I've added the GUI code from jesuscf to my github repository. Also the default cutting mode now uses simple raster scanning, according to Andre's suggestion, which apparently results in better quality for some types of stencils.
I've tagged this as version 0.2; github provides a download link here:
Future versions will get their own tags, of course.
Please let me know if this change results in a decrease in quality for any types of stencils. In general, if there are any Gerber files with features at or above 0.4 mm pitch that result in poor quality, those would be interesting to debug with.
Would you mind if I added g2g_gui.py to my Github repository (suitably credited)? Or you can fork the project and I can pull from your repo if you prefer.
Also I'll add the cutting mode from Andre that doesn't retrain for every group of backlash-free strokes. I'm a little puzzled at the rough patches several people are seeing, apparently at the extreme edges of the artwork. Could those folks who have encountered this please post/attach the Gerber files? I'd like to try them on my cutter and maybe understand better what's going on.
