Other Folks Doing 3d Printing

I'm not the only person doing 3d printing of HO models. Check out Kent Hurley's 3d model of a hexagonal gas station.

Printing a Freight Car: Southern Pacific CS-35A Flat Car

I’ll admit it. I’m not a freight car modeler. Sure, I’ve had an appreciation for appropriate freight cars from my era, and I’ve built a lot of resin kits for the Vasona Branch. But I’m not one of those folks who could argue the difference between the different patent pressed-car ends, or could tell you what percentage of B-50-14 boxcars had Bosendorfer side doors, or would argue about the appropriate underframe for a stock car. But, hey, all that goes out the door when I decide to print freight cars on the 3d printer.

My list of potential 3d-printer projects has always included a few freight cars. As popular as the Southern Pacific may be, modeling the 1930’s means building a lot of resin kits, searching eBay for old kits, or finding a lot of stand-ins. One example would be the Hart convertible gondolas, frequently used for dumping ballast on railroad tracks. Another would be flat cars - Southern Pacific’s 1930-era flat cars look much more fragile than the typical 1950’s plastic flat car kit that’s available, and tended to have a very specific look. More importantly, I needed flat cars for the Vasona Branch - lumber coming from the sawmill at Laurel needed to come down the hill to the lumberyard at Auzerais St.

I’d hoped my first-generation Makerbot was going to help me with freight cars, but it couldn’t print something that looked good at HO scale - the layers in the plastic were too obvious. The Makerbot could only print within a 4 inch cube - not big enough for even a smaller flatcar. The Form One, though, was getting big enough. At 125 mm x 125 mm x 165 mm, the printer could print a 36 foot freight car diagonally, or a longer car if I tilted it up on end. So, soon after I’d gotten the Form One up and printing, I tried a couple flat cars. Here’s the details on that first flat car.

The Prototype: Southern Pacific CS-35A Flat Car

In 1903, the Southern Pacific decided to experiment with freight cars built completely of steel. Two plans appeared that year: the CS-35 car (36 feet long with 40 ton capacity) and the CS-35A car (40 feet long, 40 tons). Both these CS (Common Standard) cars represented first attempts at a flat car with a steel under frame: a pair of 15 inch I-beams along the center, and pressed steel sides tied together by a 19 inch deep C-channel. The two designs were apparently in a bake-off; both designs date from October 1903, but only the CS-35A got the green light. SP had 1,000 of the 40 foot cars built by the Pressed Steel Car Co. in Chicago.

The CS-35A, designed during the Harriman era, differed from later flatcars such as the F-50-2 by having only nine stake pockets on each side compared to eleven or twelve on later cars. Other spotting features was the lack of rivets on the bottom edge (unnecessary in the stamped forms of the early cars, but needed in the fabricated side frames of later cars) and twelve rivets on the side at each truck bolster.

These flat cars definitely looked different from the other standard SP freight cars - relatively shallow sides, no truss rods, and longer than most of SP’s flat cars. Compared with modern freight cars, they look positively wispy and fragile. They also lasted a long time, with Tony Thompson’s SP Freight Cars book showing the cars still in common use in the 1920’s, both for revenue service and for maintenance.

Of course, there’s no models of the CS-35A available, whether as a resin kit or plastic model, so it was a great first try for 3d printing.

The Model

The first question was how to build the model - do I try to print the whole freight car in a single piece? I’d seen models on Shapeways where only the metal parts were 3d printed; the wooden deck for the flat car was done with laser cut wood. That seemed great; the wood deck would be more realistic, and I could print the model upside down on the build platform. (That didn’t work; more later.) I tried doing a couple models where only the frame of the flat car was printed, with big openings below. That didn’t work - the printed piece ended up being too thin to print, and wasn’t strong enough to survive removal. I ended up building the model so the top was solid and 6mm thick; the wood deck would be glued onto that surface.

I did the initial drawing for the CS-35 in SketchUp. I started out sketching the rough shape, extruded it as a 6 scale inch thick slab, then started adding details - first the frame and ends, then more and more detail. I scaled the sides so they were 6 inches thick and printed more reliably. For details - stake pockets and patterns of rivets - I often drew the objects in a separate SketchUp model, then made the object into a group and pasted copies into the final model. I learned that 1” diameter cylinders, 1 inch high, made great rivets. I also marked the holes for grab irons by drawing a 1.5 inch diameter hole.

There were also details I didn’t model. I didn’t add brake gear to my initial model. I planned on using wire grab irons, and grabbed brake wheels from my scrap box. I did need to model the brake cylinder for the car; it’s obvious on all the real photos I’ve seen. I started using the Tichy Train Group’s K Brake kit, and bought a half dozen at my local model railroad shop, but found I was only using the brake cylinder. I ended up building a 3d model for the brake cylinder, and printing the brake cylinder myself.

One of my other surprises when building this was the question of what exactly to build. Any model has compromises - I might not want to add all details, or I might need to change a dimension, or I might tweak the model to fit a model railroad coupler. Figuring out these differences between the model and the real thing is the hard part of making a model, whether by hand, with a 3d printer, or when designing an injection-molded kit. For the CS-35A, I had to decide whether to add some of the underbody beams and braces (no), or what detail to add to the ends (minimal.) I also had to choose between scale dimensions and a reliably-printing model. I ended up making the center I-beam much thicker than in real life so it would print more reliably. Instead of 1 inch, I made all the walls 2 inches thick. I also drew it with slopes inside the I so there wasn’t a difficult transition from thin to full-width.

Flat car and support structure. Support structure holds car at 30 degree angle to avoid any horizontal slice through the middle being too large.

I also had hiccups. I had problems printing the final model because of extra hidden faces; I used Netfabb Basic to clean up the models and prepare them for printing. Even that didn’t work reliably. Because of the Form One’s resolution, making a “watertight” model (no holes or gaps between faces) was critical. On the Makerbot, the low resolution meant holes would usually be ignored when breaking the model into layers. With the tinier resolution, even a small hole could convince the Form One software that you intended the model to actually be hollow. I also found that pieces thinner than 2 scale inches (0.020 inches, or 0.5mm) didn’t print reliably. (The design rules from Shapeways for their frosted ultra detail material have similar limitations.) I found I couldn’t print the freight cars flat on the build surface either without layers failing to bond. Formlabs, the maker of the printer, suggests that large flat prints be done at an angle so that any individual printing layer doesn’t cover a large area. Doing so gave me more reliable printing, but doubled print time. I also found I needed to be careful about cleaning the model - I’d been leaving the finished prints in direct sun to cure, but had problems with warping. Keeping time in the sun short, and making sure both sides were exposed to the sun helped limit warping.

Initial model. Note messed up corner caused by deformation when printing near edge. The other cars weren't this bad, but all cars printed with the long edge tipped had a bit of deformation.

The overall process for printing the flat cars was something like this. First, I’d print a couple models over a day or two. Like all the 3d models, I’d pull them out of the printers, remove them from the build platform, and soak them in isopropyl alcohol to wash away the extra resin. I’d remove the support structure added to print the model, then leave the models in the sun for 10-20 minutes on a side. Another day, I’d spray paint the models (Scalecoat boxcar red in a spray can). I’d drill out holes for the screws holding on the trucks and couplers, then attach both with plastic screws, and cut the screws off flush with the top surface. I’d then add a wood deck using individual 2x8 wood boards stained with india ink and alcohol, gluing them on with contact cement. I decaled the cars, then then drilled holes for wires representing grab irons and superglue them in place and touched up the paint on the grab irons. A final spray of Dullcote sealed the decals onto the model.

I made a couple embarrassing mistakes with the model. On my first try, I used the plans in Tony Thompson’s “Southern Pacific Freight Cars” book to build the CS-35 36 foot car. That model turned out well (and the 3d model is available on Thingiverse). When I went to decal the model, I looked for the car numbers for the 36 foot cars… and realized there were none listed - SP never built any. I'd been printing the Loch Ness Monster of SP flatcars.

To correct my mistake, I stretched the design out to make a 40 foot car. In SketchUp, this wasn't too challenging; I'd select the portions of the flat car from the bolster out, then use the move tool to drag these out the appropriate distance. After I printed a few of the longer cars, I compared them against photos, and realized the model didn’t look right - the wheels were set closer to the center of the cars. The CS-35A looked a bit like the ends of the cars were unsupported. I went back to SketchUp, and moved the truck bolsters in.

The final model requires about 15 cubic centimeters of resin and prints in three and a half hours. The resin cost about 15 cents a cubic centimeter, and about half gets lost during printing and cleaning, so we’re talking about $3 in resin per freight car. Compared to Shapeways, which would charge around $30 for a similar part, it’s not a bad deal. On the other hand, I went through a lot of prints for testing - probably 10 to 12 pieces. Once I got the design and printing settings done, it was easy to print more flat car castings on demand.

I still have some minor problems with warping, and the arrangement of the model usually causes one corner to be bent in a bit. Both they’re still fine models,and I’ll be able to make some reasonable 1930’s flat cars for my layout quickly. One potentially big problem is that the models are very light - only about 1.5 ounces with trucks and couplers, much lighter than the recommended 3.5 ounces. I’m planning to either add weight to the loads on the flatcars, or add some lead sheeting in parts of the model.

The Decals I printed and decaled one of the “wrong truck spacing” cars using a mix of leftover details; that process was amazingly painful. I tried finding some commercial decals that had the needed freight car numbers and tiny lettering, but didn’t find any good sheets at my hobby shop. I ended up deciding that custom decals might be reasonable, especially if I was going to build ten cars.

I drew up the artwork in Intaglio (a vector drawing program like Adobe Illustrator). I tried using Ben Coifman’s Railroad Roman font which I’d bought years ago, but its lettering looked a bit thinner than the stock SP lettering style. I ended up doing half the sheet using the Baskerville font available on the Mac which resembled the SP lettering, but had thicker letters. The bolder glyphs should make it easier to read the cars in my dark garage.

I sent the decals off to Rail Graphics; my 1 inch by 4 inch decal sheet cost $50 for a run of 50 decals, each with enough lettering to cover a few cars. The finest 2 inch tall lettering isn’t legible, but all the other lettering was readable and fine for my cars.

I’d never seen any hints on making custom decals, but designing my first sheet gave me a few lessons. First and most important, I made sure that the car numbers included a bunch of optional numbers that matched the real cars. For my CS-35A cars, car numbers would have been between 78500 and 79499. I added several numbers starting with 78/79, and also made sure that numbers starting with 48, 49, and 40 existed in case I wanted to do some of the other SP flat cars. I added several different car lengths and weight capacities so I could do some of the alternates. I also added the SPMW lettering in case I wanted to model any of the maintenance of way cars seen in photos.

I didn’t think about the difficulty of cutting the decal apart. The lettering on these decals was remarkably tiny; I had to borrow my wife’s Optivisor to be able to cut them apart. I put the “feature” lettering - “dirt collector”, “metal brake beam”, etc. on a single line, so it was easy to cut a strip out, then break out the individual words. For the other lettering such as the reweigh information, I did a lot of cutting out individual words. I wish I’d put most of the data for a single kind of car on a single line so I could cut that line off the decal sheet, then cut the individual pieces off separately.

Oh, and if you decal one car by hand before doing the custom decals, don’t use that number on the decal sheet.

The Final Models

So far, I’ve built six of these cars - two of the non-existent 36’ CS-35 cars, one of the CS-35A cars with the wrong truck spacing, and three of the 40’ CS-35A cars with the correct truck spacing. I also have one model that I had printed by Shapeways in Frosted Detail (the lower-quality.) They’re fine cars for my layout. They look positively fragile next to the chunkier Athearn 40 foot flat cars, and I also know that they’re historically accurate.

What’s next? I also like the look of the SP F-50-4 cars, with their narrow sides but prominent center beam. I’ll show more of my progress on those later. I’m also seeing about building models of the W-50-3 Hart convertible gondolas used in ballast service. Keep your fingers crossed that I can print those reliably.

And finally, let's check out the cars in action:

A Most Excellent Detail Casting

I'm still working towards that article on the 3d-printed flat cars. I've printed a bunch of the flat cars over the last month, and currently have seven on the layout. There's a couple of the "extremely rare" and in fact never-built CS-35 36' flat cars, a couple of 40' cars that I built from a stretched model. As I was decaling one, I realized I'd messed up the truck spacing; the true 40' cars set the wheels closer to the center than I'd modeled. After another tweak to the design, I've got two with the correct truck spacing too.

The printed models aren't necessarily a big win over buying commercial cars, nor are they museum-quality; I'm not looking for perfection, just some recognizable models that I can't get elsewhere. Each one of the flat cars requires some finishing and painting. Then the wood deck needs to be stained and glued on, details like the grab irons and brake wheels need to be added. Finally, the cars need to be decaled.

Today's story, though, is about the detailing part. I haven't been putting elaborate brake detail on these cars, but the old-fashioned K-style brake cylinder is a pretty visible detail and worth adding. I had been buying the Tichy plastic brake gear detail kits, but (1) I kept buying out the Train Shop's stock, and (2) at some point paying $3 to grab one little casting seemed wasteful, especially if I want a bunch of cars.

Hey, wait, don't I have a 3d printer?

So, I tried it - took some guesses at measurements, drew something up over a couple hours in SketchUp, and printed it. And it worked - the Form One did a most excellent job on a tiny brake cylinder.

So I upped the detail, adding bolt heads and even the very fine piston rod and clevis. I printed all these at the 0.025" setting on the Form One. A few hours later, I had a dozen brake cylinders, ready for the next set of cars. The photos are worth sharing; if I ever had any doubts about whether the Form One would help me in modeling, this particular piece convinced me. Zoom in to see that the clevis fork printed (though only half printed fully), and notice the bolt detail both around the top of the cylinder, and on the mounting plate at the back of the model.

As a modeler that buys a fair number of detail parts, these brake cylinders completely change the game - I'm no longer dependent on what's available from the store or what's in stock, but instead on what I can manage to draw in a 3d program.

For the folks who are curious how I drew the brake cylinder:

The first step was to get some rough measurements off one of the commercial parts and off official drawings. Once I had rough shapes, I drew a 2d cross-section of the piece's shape, and drew a circle the size of the cylinder bottom in SketchUp. Once I had these, SketchUp's "Follow Me" tool allowed me to drag that cross section around the circle, making what the high school geometry teachers might call a surface of rotation. That gave me the rough 3d object; I extruded 1" diameter, 1" tall cylinders wherever I wanted bolts. The clevis fork was similarly extruding a circle from the cylinder top, then sketching the rectangular shapes on top of that to make the clevis.

The hardest part was the mounting plate on the top of the cylinder. I did this by drawing a 10x16" x 2" thick block, and moving it so it sat on top of the cylinder. Now, this doesn't work so hot because SketchUp doesn't like figuring out the intersections between curved and flat surfaces, so it just leaves the top half of the cylinder inside the block. To make a 3d-printable shape, I deleted the top face of the mounting plate and drew extra edges at the intersection between the round and square faces, then deleted the parts of the cylinder that were stuck inside the plate.

That's pretty sweet work on a part that's less than 3/4" long, and only around 1/8" in diameter. And that part is starting to get to injection molded quality. Tichy's part is a little finer,includes the Westinghouse logo and cylinder size cast right on the side, and its ABS probably can take a bit more abuse than the Form Labs resin. On the negative side, when I look at the commercial part with a magnifying glass (and I'm using that magnifying glass a lot these days), it's easy to see that the mold was misaligned, and the two halves of the part didn't match up.

3d Models Available

And if any of you out in blog-reader land are curious about what a 3d model is like, you've now got a chance. I've put some of the 3d designs for things I've been describing here up on the Thingiverse web site. In addition to that 1920's strip mall I built a couple years back, there's now a 3d model for the CS-35 36 foot flat car and for the drying flats.

Each model also has the SketchUp file, so you can modify the design to your heart's content.

Have fun with the models, and let me know if you make anything interesting based on these designs!

SP Concrete Telephone Booths: The Results

I decided against redoing the design of the telephone booths - the existing models were quite good enough for my layout. Even the ones where the individual flat faces were visible weren't that bad. And, thanks to 3d printing, I've also got around 15 telephone booths finished and painted when I only needed... oh, maybe two. So, no work on a better door, no improvements to hinges. But I've got finished models. One's already in at the far end of the Glenwood siding and looks great!

SP modelers, expect me to be handing out phone booths as party favors for the next couple months.

I'm Un-Friending Burrito Justice

Normally, I'm a big fan of Burrito Justice, a San Francisco blogger who spends equal time talking about burritos, fog, Sutro Tower, Mission District happenings, and San Francisco history. His posts on the SP's route through the Mission are definitely worth a read. But then his big-city parochialism appears. In the latest article on the stringing of the first telegraph line from San Francisco to San Jose, he makes the sarcastic comment:

Another way of looking at it — the people of San Francisco were so isolated they actually got excited about talking to people IN SAN JOSE. I mean, that’s pretty isolated.

I could make a smart-ass comment about that's just the sort of comment you'd get from folks who live in fog and don't eat enough prunes, but that would just be sinking to his level.

3d Printing: Not for the Claustrophobic

When I was a kid, I loved exploring - climbing the big cypress near our house and finding paths to the top, or figuring out ways to reach the roof of our house. I even crawled under our house a couple times, and only once because my dad needed help getting the TV antenna wire under the house. That kind of fun stuff isn't just for kids. I remember when my aunt first started using a computer, she was playing a 3d game and managed to get into a place that was *just outside* the virtual world. Suddenly, she was seeing the back side of all the scenery and models, and could see how the scenery is constructed.

I'm getting to do that crawling around and exploring with 3d modeling too. SketchUp, my favorite tool for 3d modeling, started as an architectural sketching tool. Although it can do three-dee models, one weakness is that it doesn't always make perfect models suitable for printing. Most 3d printers expect the model to be watertight: no holes in the faces making up the model, or gaps between faces. Most 3d printers also expect that there's no hidden faces embedded inside the model. These two requirements are needed so the various algorithms for figuring out the 3d shape can figure out what counts as inside the model, and what counts as outside. One of the simplest ways to determine inside/outside is to draw a line through the model, and count the walls you encounter. If you find an odd number, then your end point is inside the model; even, and you're outside. Extra faces and holes mean the software loses track of the model shape, and you end up with a model with either odd faces, or even odder parts missing. SketchUp doesn't enforce either of those rules, and while there's software to check and fix models for you, it's a pain to have to use the external tools to do a sanity check after each bit of creative drawing.

That means my three-d modeling habit involves a lot of what I did as a kid: crawling around in strange places looking for signs of extra faces that should have been deleted. The picture at the top shows the 3d model of the telephone booth model in SketchUp. I've deleted the faces making up the bottom surface of the model, but you can see the outer wall, the hollow inner surface (light gray), as well as all the inner surfaces marking the limits of where the printing material goes (in light blue). On the left is the reverse image of the doorway; the four cylinders sticking out are actually the ventilation holes on the right side of the model. Although the original model only took about 90 minutes to draw, probably half that time was trying to fix faces that were either on the hollow inside, or were incorrect faces on the inside of the model.

So if you ask me what 3d printing means to me, it means an awful lot of crawling around in confined spaces. Good thing I'm not claustrophobic.