Re: Another Shapeways report


--- In STMFC@..., "soolinehistory" <destorzek@...> wrote:

Tom's recent round of test projects prompted me to send my L-2 passenger car triple valve out again...

Tom and I seem to be taking opposite tacks on this... Tom is looking to learn the limitations, so as to be able to design usable parts/patterns within those limitations. I think with Shapeways, he's found that build orientation is important, and if you can connive to have them use your preferred orientation, you can get usable flat parts. The problem is, there's lots of ways to get flat parts, photo etching, for one, and photo etching brings the added advantage of doing some creative folding to produce 3-D objects. Of course, Shapeways can be considerably cheaper than etching, especially for one-off parts.

I, on the other hand, want to test the usefulness of the process as a sort of "universal modeling medium" to produce those needed parts that just can't be made easily by any process... that's essentially why they aren't made. My triple valve fits this description; it's small, no, downright tiny, has almost no flat surfaces, and has mechanical features in multiple orientations, which would preclude simple injection molding even if I wanted to cut a simple cavity. This is not a simple part, but it shares these restraints with a variety of hardware parts; door hardware, hand brake housings, truck journals and spring packages... all the little details, many of which we would like to combine with a larger object, such as the door hardware integral with the door, or carside.

Anyway, here's a link to a set of photos of the project to this point; a photo of the prototype, some screen shots of my solid model, and photos of the part I had printed a couple years ago, and now from Shapeways.

This result is better than last time, to the point where I'll probably use them if nothing better comes along before I get back to the project they're intended for. I'm not happy with the surface, but have seen enough dirt caked underbody equipment to justify the lack of precise surfaces in this application. The problem is, once you start working with computer modeling, you realize how simplified the other parts are. I would love to model a better detailed, correctly sized brake cylinder to go along with this valve, and it occurs to me that it would be really cool to also include the branch pipe with its dirt collector and cut-out cock. Problem is, I'm not willing to build models that look like they were done by Georges Seurat, where solid surfaces dissolve into thousands of little dots.

Several people have been a bit down on Shapeways, mentioning broken parts. Meanwhile, Shapeways has apparently just revised their minimum cross section requirements upward, obviously in an attempt not be responsible for failures when designers attempt to push the envelope. The problem I'm seeing, however, doesn't relate to minimum cross section (OK, so I can't include the brake pipe) but rather to layer thickness and surface resolution. The various machines I've researched that have a build envelope the size that Shapeways offers have a minimum layer thickness of 50 microns... essentially .002". Some systems are now claiming 27 microns, a bit over .001", but have build envelopes of only about one cubic inch. The problem is, .002" is clearly visible on an HO scale part; I've used that dimension for lap seems (although I prefer .003", to be sure it doesn't get filled with paint). I'm not sure that .001" is going to be much of an improvement. When the systems get down to 10 micron layers, I think we'll have something. I don't see why it isn't possible; I doubt 2-D printers lay down much more ink than that, but build time, and cost, is going to go up by a corresponding amount.

It will be interesting to see what the next couple of years brings.


Many thanks for the easy to understand report!

Looking at the shapeway part on flikr, I confess to being a little confused. On Tom's parts, I felt one could detect the orientation of the build layers. But on your triple valve, I can not spot a layering effect, instead the surface just seems to be "noisy".

I am wondering if a barrel tumbler/polisher technique might help smooth the surfaces of a three dimensional part? Start out with surface features a little oversized, and let the polishing medium remove the "noisy" surface features. In place of a tumbler, I wonder if a medium-thick paste of baking soda and water, in a small tight container with the part, might be effective if it was immersed in an ultrasonic cleaner? Kind of a mini-polisher.

Probably a crazy idea, but for 3D parts with all of the detail, it may not be vitally important that the desired features be perfectly replicated - as you noted about underbody grime. Outside of very close examination, what is more important - the suggestion of six hex nuts on the valve cap with the adjacent surfaces looking relatively smooth, or six perfectly formed hex nuts with the adjacent surface covered in warts?

But a very impressive effort and improvement over the old part - looks like a little more patience may be in order.

Dave Evans

Join { to automatically receive all group messages.