FXD to FUD Comparison - Locomotive Shell

[C855B]

Executive summary first: FXD makes a difference and is worth it.

This comparison is with James Norris' great DD35 shell providing the opportunity for A-B comparison. I had an unfinished (...heck... unstarted) FUD version on my bench, and just received the FXD version this morning. The FXD has about four hours in Bestine with a little bit of stirring and a quick hit with a toothbrush. Photos are with an iPhone, which did sort of OK. I left the images at full resolution to download for zoom viewing. FUD on left, FXD on right:



Detail is MUCH sharper, especially noticeable around the hinges and hood latches.



We still have a bit of surface fuzziness with FXD, but notice the sharpness around the headlight and especially the access hatches at the bottom. FUD version is blurry, and the two smaller hatches have nearly disappeared. The printing artifacts on the FUD version under the headlight and ladder are just that, printing artifacts. I was more or less OK with them and would have sanded them off since I was planning to use wire grabs, anyway.



Again, notice the sharpness around the hinges. I couldn't get an in-focus A-B comparison around the brake wheel, but FXD was much, much better. With the FUD I was thinking I would have to cut the wheel out and find a suitable brass version.



This is a nice shell.



Excellent rendering of roof detail. I couldn't get the light right to show differences in comparison, but FXD made a difference.

There were striation artifacts on the roof with FXD more noticeable than with FUD, and do not show in the pics. They were not objectionable and will probably disappear after the first coat of paint.

Final judgment really ought to remain until they're painted and detailed, but I think FXD is a good process for N shells. With this level of detail more or less in the bag, I am greatly anticipating James' release of the Alco Century 855 artwork, expected in the next few days.

[chicken45]

Cool! Were there any weird streaking parts around rivets?

[Chris333]

Whoa, that is great! Gives me hope.

[tom mann]

Pretty nice.  Any dimensional differences?

[sirenwerks]

Quite noticable.  Like that plate of steel under the headlight, it doesn't even look like it exists on the FUD version.

[C855B]

No streaking I can see around the rivets. I think that problem is similar to the printing artifacts I noted on the end.

Tom, there is a slight difference in length, with the FXD about .050" longer. When I get back to the bench later tonight I'll see which one is closer to "right".

[C855B]

I'm back... the FXD version is more accurate - roughly 2 scale inches short of prototype. The FUD version of the same 3D digital artwork is 8 scale inches too short.

[Mark W]

Excellent comparison shots!  It looks like the rounded part next to the brake wheel has very little stepping also.

If you don't mind me asking, what was the total cost for each?

[peteski]

Shells print too short?  I honestly wonder if the design is incorrect rather then such a large printing error.

[C855B]

Artwork was the same file, Pete - just different output processes. My hunch is a cumulative increment error on FUD. I have heard mention of "shrinkage" with FUD, and if the step increment calibration is off by a couple of microns, it would show up in long pieces like this. The FUD shell was printed several months ago, and... easily... in the interim they could have learned more about calibration. In the early 2D high-res days, we found out about cumulative increment errors the hard way, and instituted quality checks.

Cost? $50 for FUD, $75 for FXD.

[peteski]

I would have expected the 3D printers to use stepper motors. Those normally have a defined number of steps per revolutions (and they don.t skip unless something is seriously broken).

[C855B]

Well, so did our 2D printers. It was the variability of the physical characteristics of the equipment. The output difference between (random number for example) 17.35 and 17.35001 as a step size calibration factor for raster conversion would only be seen when accumulated over thousands of steps. It took a few "discussions" with a couple of clients complaining about dimensional accuracy on certain large pieces for us to realize the out-of-the-box settings for the imager were close but not close enough for big distances.

What's not considered here is the possibility that the FUD test piece shrunk in the 8 months it's been sitting on my bench. Now wouldn't that be a pill?

[bbussey]

The ProJet output doesn't shrink. Your first hypothesis is the correct one. If the machine isn't calibrated properly, the parts will not be sized accurately. When you have control of the orientation, you can compensate by scaling your 3D model accordingly. But that isn't possible with Shapeways because the build orientation can change from run to run.

What I have done in the past is output a 1" cube and measure all three dimensions to determine what scale adjustments were necessary to compensate, when working with other contractors.

I will say though that there appear to be fewer if any "rough" surfaces, on the model faces that were against the wax support structure, than on their FUD-building machines. Probably because the equipment is newer and in register. So the FXD process is viable for smaller detail parts. For the moment. Hopefully they can maintain the quality as the months progress.