Lots of Holes

[jdacal]

I'm giving the printer a bit of a torture test tonight. Making a tray to store Dremel bits. Measures about 7" x 9" and consists of over 600 3.3mm holes spaced about 8mm apart.

Total height of the tray is 18mm, hole depth is 15mm. 0.25 layer height for 72 layers. 20% infill.

Curious about one thing. Rather than do all the holes in sequence the print is going to all kinds of different points to do the holes. Seems like its wasting a lot of effort with all that head travel, rather than doing the circles in sequence or by rows.

Just wondering if there is a setting to correct that behavior?

Also interesting is all those holes gave S3D quite a workout. First time i hear my workstation fan go to 100%. Also first time my CPU ever pegged at around 100% for a solid 15 minutes. It generated a huge GCODE file for those 72 layers. Its printing fine, just curious if theres a way to optimize all that head travel.

[insta]

Oh God that looks like a MF'er of a print (just watched the live stream).

I don't have more to contribute since I use slic3r

[jsc]

I don't know about the travel behavior, I haven't seen that myself. But for the slicing, what did you model it in? How many facets does each hole have? You will get faster slicing if you reduce the number of facets. And in fact, 3D printed holes will be more dimensionally accurate if you use low edge-count polygons (http://hydraraptor.blogspot.com/2011/02/polyholes.html).

[jdacal]

I don't know about the travel behavior, I haven't seen that myself. But for the slicing, what did you model it in? How many facets does each hole have? You will get faster slicing if you reduce the number of facets. And in fact, 3D printed holes will be more dimensionally accurate if you use low edge-count polygons (http://hydraraptor.blogspot.com/2011/02/polyholes.html).

Hi jsc, thanks for that nice link. I use Cubify Design which is a derivative of GeoMagic, but am not experienced enough with the CAD yet to tell how many facets I'm using. Pretty much working with the default settings at this point. So far I've been pretty good at guesstimating how much bigger to make the hole so that it works ok for the application. But will definitely look into it further based on your article.

[Tim]

3D printed holes will be more dimensionally accurate if you use low edge-count polygons

That's counter-intuitive when taken out of context, and is quite likely to be taken out of context by anyone who doesn't read the article thoroughly. You should point out that low edge-count polygons are very poor representations of circles if you just plug in the inside diameter you're looking for and the vertex count. Because the design software will put the vertexes on the circle, then the edges between the vertices are always cutting into the circle, making it smaller, but in a deterministic way. For N vertices, divide the diameter by cos(pi / N) to get the number you want to plug in for the diameter so that the inside of the polygon is the same size as the circle you're looking for.

Otherwise, the article cites several reasons that high-edge-count polygons make poor circles, which taken together are difficult to model and can be dependent on the machine and the type of filament.

Really interesting article. Thanks for posting the link to it.

[jdacal]

I just realized what its doing with all that head travel to do the circles. Apparently even with all those circles spaced closely together there are some in-fill lines that stretch pretty far. So its not wasting its time traveling to do a far away circle, its actually laying infill and when it runs into a circle it does it. Makes more sense now and pretty neat!

[jdacal]

Just to finalize the thread I did finish the project, and as luck would have it I mentioned previously that up to now I had guessed hole diameters properly on previous projects. As it turned out I was off on, of all projects this one with over 600 holes! So my Dremel drill press got a workout expanding all these holes. By the way, Dremel totally sucks on drilling PLA. Totally bogs down. I have the battery powered Dremel and the top of the line AC model - terrible torque for drilling, seizes up completely. Ended up doing most of them with my hand driver visible in one of the pictures.

But all's well that ends well...

2014-09-23 22.52.26_1.jpg (234.56 KiB) Viewed 10670 times

2014-09-23 22.52.58_1.jpg (240.18 KiB) Viewed 10670 times

[Dale Reed]

This whole (hole? ) discussion reminds me that S3D REMOVED support for G2/G3 optimization a couple releases ago. When I was working on settings going through the MAKE Magazine torture test object ( http://www.thingiverse.com/thing:33902 ), I tried turning on the G2/G3 setting in S3D and got the BEST columns ever. I believe the G2/G3 setting in S3D would look for sequences of small lines that defined an arc/circle and replace each such sequence with a G2 or G3 G-code command to "draw" the arc/circle in one shot, which the Marlin firmware supports.

I wish they hadn't removed it. I suspect it would have detected what was going on with the holes and done proper circles. (Not sure if they detected ID or OD and whether it assumed the points were on the circle or if the n-gon circumscribed the circle. But you could probably figure it out with some quick test prints and measurements...)

I'll have to see if I have one of those older versions hanging around and install it on another PC and try something like this. I don't have that many Dremel bits, but I also have two or three other boxes of drill bits. It would be awesome to have a customized holder to organize the whole lot of them.

Dale

[jsc]

I think the issue was that the Marlin arc detection/generation code is experimental and could result in massive slowdowns trying to do all that math on the poor ATMega. I'm fairly certain that all it does is detect what appear to be circular arcs and redraw them with lots of linear segments. You can get the same results by upping the arc resolution on whatever modeler used to generate the STL files, although in cases where you only have the STL that is not possible.

The torture test STL file looks like it has very finely faceted arcs, so I'm surprised that using the arc support improved matters. You could try it with Slic3r, which still has support for G2/G3.