faster.mov

Update!

Brian yelled at me so I converted my autoplotting code to Python. Now it’s:

* Python for holding it all together
* openCV for grabbing image from the camera, applying Canny edge detection, saving bitmap
* autotrace for centerline tracing bitmap and converting to eps vector file
* pstoedit for converting eps vectors to hpgl
* Chiplotle for driving Roland DXY-1300 desktop plotter

As you can see in the video it all happens very quickly. You run Faster.py, point the camera at something, press ‘p’ for ‘plot it now, fool’, and the plotter starts cranking!

It’s like a very slow, low resolution, curve-inducing, monochromatic, photocopier hooked up to a webcam. Finally.

(Thanks to Sam Pluta for the expert camerawork)

WOW! In just a few hours of hacking I’ve got a fast, automatic camera to plotter system up and running. It’s a little Java app that grabs an image from a webcam, uses a Canny edge detector to make a bw bitmap of the edges, calls potrace to turn that into vectors (in an .eps file), calls pstoedit to convert the .eps to HPGL, then finally pipes the .hpgl file out to Chiplotle, which drives the DXY-1300 on my desk. The whole process, up to the actual plotting, takes just a couple seconds. Plotting takes several minutes, depending on the complexity of the plot.

It needs some controls and tuning to get a reasonable number of edges and to avoid lots of speckly noise. The images captured in the piece I’m working on will be in a controlled light environment, so once the basic tuning is done I think it’ll be able to run without much tweaking.

So now if I have something in my studio that I want a copy of I can just hold it up to the webcam and in 10s of minutes I can have a noisy, monochromatic pen drawing of it! It’s like a desktop version of Fotron2000!

It does not plot with mint (mmm) but it is a mint plotter, new in the box, acquired via the magic eBaysian time travel delivery service for a mere $100USD. It’s a Unisys GRP 2-PLT from 1990, which I think is a re-branded Houston Instruments model. It emulates an HP7475a, which means it’s running like a champ out of the box with Chiplotle!

movie!

Dissonant Counterspots

Larry Polansky and I have been geeking out over dissonant counterpoint. I made a quick Java applet visualization/sonification to help understand what’s happening:

Dissonant Counterspots

Update: I’ve done some more experiments and pulled a bunch of prints. The technique works really well!

I had one bad screen where I rastered the image too many times and it ended up sort of bleeding, opening up the lines too wide. This was on 250×250 mesh, which I thought would be too fine. I think that the clear coat started to melt away a bit as the screen heated up. The metal threads are really, really thin. So I have a hunch that they got a bit hotter than on lower count screens and that fuzzed out the image.

The bad image gave me a chance to try re-using a screen. I simply sprayed more clearcoat over the existing image. I could still see the outlines of the previous image, but I could tell that all of the mesh was sealed up again. So the parts that had no image before now had a double layer of clear coat. That turned out not to be a problem.

I decided to use a faster speed and more power this time, and to just do one pass. So I set the speed to 200 (about 1/3 max speed, don’t ask why 200 = 1/3 max speed, it just does) and 40% power. It worked perfectly on one pass. The image I used isn’t super detailed, but it has both sections with some small details and large open sections. Both look good! It’s the tape deck graphic above, which is one of Neg-Fi’s logos.

I added a couple new images of onesies with ArtBots and Neg-Fi logos on them to the gallery.

I’ve been experimenting with ways to combine the laxxxor and traditional printing techniques. The last week I’ve been playing with screen printing, and it’s working quite well. I’m using steel mesh that’s been sprayed with a clear coat. I use several passes with the laser to slowly and gently burn off the clearcoat where I want the image to come through.

I haven’t got the settings quite right yet, and they depend a lot on the exact thickness of the wire and the density of the weave in the mesh. So far I’ve found that < 100x100 mesh is too coarse. 100x100 works well although large open areas leave some texture from the mesh. Right now I'm lasering a screen with 150x150 mesh. See the Pioneer 10 plaque examples above for the results! 150x150 is RAD.

110 is a very common mesh size for normal photo-exposed screen printing using plastic screens. I don’t think my technique will work with plastic mesh, as the power needed to burn off the clearcoat is most likey more than enough to melt the plastic mesh. I’ve even managed to burn up some of the steel mesh. But so far I’m very happy with the steel results. I’m just using crappy homemade plywood frames, the wrong ink, a half-assed squeegie, etc., and I’m getting some nice prints. Yay! I haven’t tried re-using any of the screens yet. I guess I could just burn a 100% filled image to remove all of the clearcoat. Or maybe I can just spray more clearcoat onto the exposed parts. Will have to do more experiments.

Here’s a gallery of experiments with 70, 80, 100, and 150 screens. I used different images for each, so it’s not really a fair comparison. But so far 150 was the best for detail and consistency. 100 worked well for the ArtBots logo where there’s a lot of open space in the letters.

He’s soldering it up now, I’m curious to know if it’ll be any easier to solder. The powder was a little $, but each application only takes a minuscule amount, so the 4oz container I bought should last for hundreds of boards.

UPDATE: It worked great! Some pics: