I am glad Tom took the time to write all this up. I have 3d printed hundreds
of items, and used a couple different printers. Sharing files between
different printers does not work, in fact I have had to do complete redesigns
because of a software upgrade on a printer. It was worth it, the software
upgrade made the finished product that much better. Any files shared would
still need to be edited to fit particular printers. It can be done, but it
is not as easy as just grabbing a bunch of part files and assuming they will
work in any given printer. The newest SLA printer from Kudo 3D has a
resolution of 37 microns, with the possibility of adjusting it down to 25
microns. If it shows up on time, I will have examples in clinics at the
Amherst Show. Kent Hurley nvrr49.blogspot.com ---In STMFC@...,
<pullmanboss@...> wrote : Some notes from the real world…

When you get to your late 70s your list of “not in my lifetime” possibilities
is considerably longer than it is for you youngsters in your 60s. That
certainly colors my view of 3D printing and libraries of CAD files. So what
follows is the view of someone more likely to receive a pacemaker than a 3D
printed Pacemaker boxcar in the foreseeable future.

I won’t discuss 3D printers except for how they affect the design process. We
had an extended thread on 3D printers in July 2013 which may be worth
reviewing.

The two printing processes I use the most are stereolithography and mutltijet
modeling. Ross (no last name) mentioned 3D Systems’ Viper high resolution SLA
(stereolithography) machine. Shapeways’ FD and FUD (frosted detail and
frosted ultra detail) parts are created on 3D Systems’ even higher resolution
MJM (multijet modeling) printers. SLA builds by drawing the image of each
layer on the surface of a UV curable liquid resin. It’s a vector process. In
hi-res mode the laser beam is 0.002” dia, the layers are 0.002” thick, and
the surfaces are the smoothest of any 3D printing process, but it can’t do
parts with overhanging features. MJM printing is a raster process (think 3D
inkjet printing), with resolution of 600 x 600 DPI (and increasing) and
layers less than 0.001” thick. It’s faster and less expensive than SLA, the
surfaces are slightly textured and it can handle overhanging features.
Surface texture is no big deal for 1:1 objects, but it can be for our
miniatures.

A good designer has to be familiar with the capabilities of the manufacturing
process he’s designing for. By their nature, stereolithography underbuilds
features slightly, and MJM machines overbuild by a similar amount. Not a
factor unless you’re designing things like ¾” dia rivet heads in HO. For SLA
designs I increase the head diameter by 1/8”; for MJM designs I decrease it
by the same amount. It may seem like no big deal, 1/8” in HO is just under
0.0015”. But if I use the same design file to print an uncompensated ¾” rivet
head in each process, the difference in the printed rivets (5/8” SLA vs.
7/8” MJM) is 40% and is very noticeable. Prototype rivet heads are not full
hemispheres, but to get all the rivets to print properly and be visible
through a couple of coats of paint, I make them full hemispheric domes and
put each on a 1/8” riser. If a car uses several different sizes of rivets,
and you want to represent the visual differences, it may be necessary to
design the smallest ones so they print reliably, even if that makes them
larger that they should be, and increase the other sizes accordingly. In HO I
know if I want to represent a series of surfaces that are offset from one
another, like a window frame, or a rivet batten on a car side, the layers
must be separated by at least 3/8” for the layering to be noticeable. In
other words, it helps if you can apply the precision touch of an engineering
designer with the eye of an artist.

So there we require two different design files for a single part, depending
on what process will be used to print it. For HO. But those HO rivets won’t
print if reduced to N scale, and they are probably too large if printed in O
scale. The differences in the level of detail required for the various scales
is what tempers my enthusiasm for a CAD library. Ignoring the design audit
issue, a printer will try to print everything in the design file, but parts
smaller than the resolution limit of the printer will be blobs on the
surface. So a highly detailed design file that prints gangbuster parts in O
scale will have to be dumbed down considerably for the smaller scales. And
vice-versa. Too many variables for this old mind to process.

Finally, Dennis mentioned that I “gamed” Shapeways’ process to get good parts
by ganging three parts facing in different directions. It’s actually five
different directions, as shown here: www.pullmanproject.com/Gaming.jpg
http://www.pullmanproject.com/Gaming.jpg

Those are left- and right-handed versions of compressor boxes used on
passenger cars with B&O-style York air conditioning. The plate is
approximately 3” x 3” and cost $54 from Shapeways. Too expensive for parts to
use as-is, but just fine for resin casting masters. Shapeways built the plate
on edge and I did get two good parts. But only two.

Nothing is as simple or as straightforward as it seems.

Tom Madden

Some notes from the real world…

When you get to your late 70s your list of “not in my lifetime” possibilities is considerably longer than it is for you youngsters in your 60s. That certainly colors my view of 3D printing and libraries of CAD files. So what follows is the view of someone more likely to receive a pacemaker than a 3D printed Pacemaker boxcar in the foreseeable future.

I won’t discuss 3D printers except for how they affect the design process. We had an extended thread on 3D printers in July 2013 which may be worth reviewing.

The two printing processes I use the most are stereolithography and mutltijet modeling. Ross (no last name) mentioned 3D Systems’ Viper high resolution SLA (stereolithography) machine. Shapeways’ FD and FUD (frosted detail and frosted ultra detail) parts are created on 3D Systems’ even higher resolution MJM (multijet modeling) printers. SLA builds by drawing the image of each layer on the surface of a UV curable liquid resin. It’s a vector process. In hi-res mode the laser beam is 0.002” dia, the layers are 0.002” thick, and the surfaces are the smoothest of any 3D printing process, but it can’t do parts with overhanging features. MJM printing is a raster process (think 3D inkjet printing), with resolution of 600 x 600 DPI  (and increasing) and layers less than 0.001” thick. It’s faster and less expensive than SLA, the surfaces are slightly textured and it can handle overhanging features. Surface texture is no big deal for 1:1 objects, but it can be for our miniatures.

A good designer has to be familiar with the capabilities of the manufacturing process he’s designing for. By their nature, stereolithography underbuilds features slightly, and MJM machines overbuild by a similar amount. Not a factor unless you’re designing things like ¾” dia rivet heads in HO. For SLA designs I increase the head diameter by 1/8”; for MJM designs I decrease it by the same amount. It may seem like no big deal, 1/8” in HO is just under 0.0015”. But if I use the same design file to print an uncompensated ¾” rivet head in each process,  the difference in the printed rivets (5/8” SLA vs. 7/8” MJM) is 40% and is very noticeable. Prototype rivet heads are not full hemispheres, but to get all the rivets to print properly and be visible through a couple of coats of paint, I make them full hemispheric domes and put each on a 1/8” riser. If a car uses several different sizes of rivets, and you want to represent the visual differences, it may be necessary to design the smallest ones so they print reliably, even if that makes them larger that they should be, and increase the other sizes accordingly. In HO I know if I want to represent a series of surfaces that are offset from one another, like a window frame, or a rivet batten on a car side, the layers must be separated by at least 3/8” for the layering to be noticeable. In other words, it helps if you can apply the precision touch of an engineering designer with the eye of an artist.

So there we require two different design files for a single part, depending on what process will be used to print it. For HO. But those HO rivets won’t print if reduced to N scale, and they are probably too large if printed in O scale. The differences in the level of detail required for the various scales is what tempers my enthusiasm for a CAD library. Ignoring the design audit issue, a printer will try to print everything in the design file, but parts smaller than the resolution limit of the printer will be blobs on the surface. So a highly detailed design file that prints gangbuster parts in O scale will have to be dumbed down considerably for the smaller scales. And vice-versa. Too many variables for this old mind to process.

Finally, Dennis mentioned that I “gamed” Shapeways’ process to get good parts by ganging three parts facing in different directions. It’s actually five different directions, as shown here:

www.pullmanproject.com/Gaming.jpg

Those are left- and right-handed versions of compressor boxes used on passenger cars with B&O-style York air conditioning. The plate is approximately 3” x 3” and cost $54 from Shapeways. Too expensive for parts to use as-is, but just fine for resin casting masters. Shapeways built the plate on edge and I did get two good parts. But only two.

Nothing is as simple or as straightforward as it seems.

Tom Madden