lamp jack

Here is one of five lamp jacks for the passenger car. It’s simply constructed from a pin and a length of tube. I held them apart with a razor blade as I soldered them. It’s about 18 mm (3/4″) long, and most of it will get cut off shortly before installation.

You can probably buy something like this, and one of the questions that I get asked when demoing is how do I choose to make something rather than buy it. The answer I give is one that I remember reading thirty years ago when I started scratchbuilding. It is this: if I can make something that’s as good or better than the one I can buy, I’ll make it.

Following this rule has had a couple of benefits over the years. First, it has meant that I’ve spent a whole lot less on my hobby than I might have. Here, for example, I’d expect to spend a couple of dollars for a set of jacks that I made in less than half an hour for pennies. When I started out, this was my primary motivation for scratchbuilding, because, well I had no money. None. No really, none.

The second – unexpected – benefit of following this rule has been that my skills have improved with every part I’ve attempted. Often I set out to make a part, not knowing if I can or not, but knowing that the best I can find is not all that good. So, I attempt it once or twice, and sure enough, after a couple of tries I do get something that looks more like what I’m after than the best commercial offering. In the meantime, I’ve learned how to fabricate something that I didn’t know how to make or even know if I could make it.

Finally, and this was the motivation this evening, scratchbuilding actually saves me time. Sure, I might have been able to buy lamp jacks, but I would have spent as much time searching through Walthers, Cal Scale, Bowser, Detail Associates and every other manufacturer’s website, scrutinizing the terrible photography. Then there would have been the ordering at Central Hobbies or direct, a trip to the hobby store, and before you know it, these parts have cost me much more than half an hour.

And it’s fun to make stuff too.

Getting back to work on the passenger car

inally, after months of computer work, and Christmas, and finishing off the models for the train show, I am back to the passenger car. Amazingly, it’s been almost six months since I did anything physical on this model – most of the recent work has been on the computer, composing the interior, the clerestory lights, and improving the freight trucks.

Yesterday I ordered the interior and clerestory lights, and I’ve been working on the windows, which are going to get laser cut. I would have liked to have made them on the RP machine, but they don’t satisfy the minimum wall thickness at Shapeways. So, laser cut it is.

I cast about for a while looking for someone who will do a .010″ styrene through a service like Shapeways, but came up dry. So, I’m going to ask my friend, Brian, who happens to own a laser cutter, to do them for me.

First, I thought I’d better do a little proof of concept. Actually, I did two – one in brass to see if etching was an option and another in styrene. The styrene one works fine, and is shown here.

Notice I’ve sanded the jaggies off the roof, and lightly sanded the sides where they had faint horizontal ridges.

Waxing philosophical on the arrival of my models

Well, after three tries, my models were finally fit to print. They arrived about two weeks ago, and I’m delighted. The passenger car came out just about exactly as expected, and the truck, while not as detailed as the design, is completely functional and looks pretty good.

We’re at the dawn of a new era in model railroads. Just as the old kingdom is at its zenith – who would have believed we would have so many choices in injection molded models – we can see what will replace that when the market is no longer around to support the increasing costs of production. After all, we will not be able to count on cheap production in China forever: those people will demand increasingly high wages, and when they do, you will not be able to afford the same quality in injection molding and hand-assembly. As your ability to pay for the next amazing offering dries up, the quantity of new products on this technology has got to diminish, and we’ll have a much smaller hobby.

So many model railroaders realize this, or they are looking around at their graying peers, and are already predicting the end of the hobby. Nonsense. The hobby will always be here, but it will not always be the same as it is in 2009. Perhaps in 25 years, the market for the wonderful products we have today may be smaller, and the injection-molded, hand-assembled freight cars themselves might be priced out of reach, but that’s today’s technology.

In 25 years, the high quality models will not be hand-assembled, injection-molded, but built one at a time on high-throughput custom-manufacturing machines much like the ones at Shapeways that have produced this truck and passenger car (only better). Today, Shapeways is able to produce a cylinder 1 mm thick, which is really too thick for our purposes; however, we’ve seen a continuing improvement in resolution over the past twenty years, and I expect in a few years we’ll be able to have a car printed complete with grab irons and brake gear.

When that happens, say goodbye to your resin molders. Oh, they’ll hang around for a few years, but really, who wants to spend an evening drilling 40 holes in a piece of grey plastic, when you can snap your fingers and have the part pre-assembled for you? Sure, the resin parts will be less expensive than the rendered parts, but if I can casually go and create a computer model, then open a store on Shapeways and start selling it without even thinking about starting a business, well, who would get into the messy, nasty business of casting if you had that option.

Truck detail

At the very least, we will see an explosion in availability of interesting prototype-specific cars. Heck, there is no need to have two cars alike! As a modeler, you could easily dent each specific hopper as you know it should be for a particular day in its history; you could get a coach with two windows open, while your friend gets the same car with three windows open.

The truck is an example of this. I need this type of truck, which has not been manufactured before; I flirted briefly with the idea of spending $10,000 to get them injection molded, but never believed there would be much of a market. But getting them printed is actually competitive with retail prices for injection-molded trucks! So, while the detail level is not quite what I would like, it is acceptable and on par with the cast metal trucks that many modelers are still using. Once I’ve finished adding some parts like the spring beam and brake beams, I’ll make them available to other modelers to buy.

If there were a metal version of this truck, you would not buy it any more, because this printed version is equal, probably cheaper, and doesn’t require assembly. Goodbye cottage industries, hello den industries!

These den industries will require even less commitment than the cottage industries of today. A modeler will need something, they’ll design it, have it printed, and offer it to the community at perhaps a small margin.

Minimum Wall Thickness

I submitted my designs to Shapeways, and waited with baited breath for their realization in plastic. By the tenth day, I wanted to race home at lunch time to check the mail box. That day passed, as did the next and a few more; finally I got an email from Shapeways.

Apparently they do one last manual check before submitting the model to print, and in this manual check they found that I had violated the minimum wall thickness. Now, I knew I was violating this rule, but with a thorough read of their site, it seemed like I could do so as long as I kept the distances minimal. The way I read it, I should be able to have a short span of something finer than 1 mm diameter. However, this turns out to be false: you can have something that ends finer than 1 mm, like a knife edge, but you can’t have a barbell shape with a fine handle.

So much for window mullions, or printing ladders, at least for now. I’ve redesigned the passenger car without any window frames, and I guess I’ll get those laser cut in .015 material. The more challenging redesign was the truck, however. For this, I’ve made all the various straps in the double arch bar include a 0.7 mm (minimum wall thickness for white strong flexible) diameter rod that tapers to the front and to the back. It doesn’t look too much worse than the draft angle on some of the older injection molded arch bar trucks, at least on screen.

I’m going to wait a week before I submit the job so it doesn’t arrive when I’m on holidays. Hopefully this time, it passes muster and I actually get some plastic!

Water tight and ready to print!

Okay, it’s all cleaned up and uploaded to Shapeways. For only $100, I can get this printed, but I’m going to design a couple of small test pieces to try out as well. So, you’ll all have to wait a few more weeks for the print photos. Now that I’ve got the little dance with MeshLab and AccuTrans all figured out, getting a printable model is actually straightforward.

If there’s one thing I’ve learned in this drawing, it is that SketchUp! doesn’t like working in HO scale. It is much more comfortable in full size. One of the small pieces is a truck, and I’m designing that full size, so we’ll see how Accutrans does with the conversion to HO scale.

Finally a successful upload!

Alright! Finally, I’ve uploaded a model successfully. I continued to struggle with inverted normals until I figured out how to get Accutrans to fix the remaining normals. The do-it-yourself active worlds page came to my rescue when the Accutrans help files failed to load, and it turns out all you need to do is click on a couple of semaphore flags, then click on the model.

Once that’s done, you just have to save it as a Collada (.dae) file again. Unfortunately, somewhere along the line, the model got expanded a hundred times; it also seems to lose its units. As Shapeways interprets such files as cm, the resulting model was 778 cm long (about 25 feet for those of you who remember those). That’s a bit too big for the Shapeways printers, and a bit too big for HO too.

Fortunately, Accutrans knows all about scaling, and so, all I had to do was to save it at a scale of .0254, and I got a part that ought to work.

As you can see from the screenshot, it’s apparently only going to cost me $10 to print the side, which means that the whole model might come in well below my original $100 estimate.

Finally a Water-Tight Mesh!

Holy cow, this was harder than it should be! Shapeways keeps responding to numerous attempts to upload pieces of my model with an error stating that I have inverted normals. Yes, I understand why that might be a problem, but in a model with some 2000 faces (on one side of the car alone), some of which are extremely small, figuring out which ones are inverted is like looking for a needle in a haystack. In fact, if you look at the model with normals displayed in meshlab, it looks a little bit like a haystack!

It was time for a new free tool. So I downloaded AccuTrans 3D, which, while not free, is only $20 CAD for us Canadians, while the rest of you have to pay $20 USD (thank-you mouse guy from Regina!). We’re still in the virtually free world, but time is getting expensive. I still have inverted normals, but Accutrans enabled me to get a water-tight mesh, which I happen to know was going to be a problem at some time. Tracking down the holes in my mesh turned into the following little dance:

  • Export the model as a 3D model Google Earth (.kmz) file from SketchUp!
  • Rename the .kmz file to .zip and extract the .dae file inside
  • Open this .dae file in MeshLab and save it as .stl
  • Open the .stl in AccuTrans and choose “Tools/Check for Water-tight Meshes”
  • Now, the first hole I found was big enough that I could find it immediately. The others were tiny (but I guess still big enough to let water drip out), and the way to find them was to turn off the VCG layer, so I could see the holes, zoom in and turn on the VCG layer again so I could relate them to the model.

Greed versus SketchUp!

Ugh! I’ve spent most of the evening battling myself and Google SketchUp! The two of us are formidable foes.

The challenge is this: Shapeways wants a super-clean mesh before they’ll attempt to print. That means normals all have to point the same way, and no hidden faces or lines.

Sadly, SketchUp likes nothing better than creating hidden faces and lines — it works best like this. When you see the slick demos online, rest assured that the resulting model is unprintable.

Doubly sadly, when I started the model, I was seduced by the powerful gestural language that the tool provides, and quickly did things like extruding rounded shapes for the window frames (who knows how they’ll resolve on the printer!), and sloping the window sill so the rain doesn’t run into the car.

So now, we need to make those two surfaces meet cleanly. Should be a simple matter of using the “Intersect” tool in SketchUp, right? Well no, because SketchUp doesn’t like working at the small scale I’m working in at all. I should have realized this was going to be a problem when SketchUp wouldn’t even let me create an arc .015″ in diameter, and I had to draw segments myself.

I probably wasted an hour trying to get it to fill in the odd shape at the foot of the window frames. Finally I gave up, and here is my solution: I drew a rectangle at the foot of the window frame, then bent the bottom of the frame so it fills the rectangle. Voila, the shape on the window sill is easy to fill because it’s all big squares, and the inside will have no hanging surfaces. I haven’t tried to upload yet, but I have high hopes that this will work for me.

Sides and roof modeled in Sketchup

Well, I’ve learned a ton about Google SketchUp since starting on this project. Today’s lesson was about groups. Now in every other drawing program I’ve ever used, groups were of little actual utility because in order to edit the elements of the group, you had to break the group and recombine the elements. In SketchUp, on the other hand, you can edit the group in place, which is a huge boon when you’re dealing with groups of thousands of elements. For much of the past couple of weeks as I’ve been modeling, I’ve been struggling with SketchUp choosing the wrong element when I’m trying to pick things: it often wants to choose an element behind the one I’m interested in. I can now see that combining elements into groups and editing the group is the way to defeat this annoyance. The more I use it, the more I like this tool.

Anyway, as you can see, I’ve not modeled the ends yet, and there are a many details to go, but the car is starting to take shape.

Starting on a passenger car

Well, you may as well know I’ve started on my next Proto:87 project. It is going to be the passenger car here, which I sadly know little about except that it ran on the Pembroke Southern behind my locomotive. I believe it was a GTR car, about 60 feet long, but the number and all other information seem to be lost in time.

Now for the approach. When I was in Houten, I got a chance to handle some models made with 3d printing by Shapeways. This got me thinking about the best way to make the roof, which, as you can see, does not present any useful flat surfaces to make construction easy. So, I’ve been busy modeling this part in Google Sketchup with a mind to getting it made for me. Then, I thought, well why stop there? So the whole sides, ends and roof are going to get printed. I’m going to do the underframe in wood because it would have been unpainted wood on the prototype.

Exciting times! Here is the quarter roof after I figured out how to do the compound curve. I spent a number of evenings figuring out Google Sketchup before finally getting this together. The trick turned out (for me) to be to draw longitudinal contour lines in AutoCad. Then I transferred this drawing to Sketchup and moved the lines so they lined up with their locations on the transverse section (that’s what the ticks are on the far edge). There is a whole lot of work involved in stitching the adjacent lines together to form a surface, and I guess I could then go and smooth all the lines. I’m happy enough with the triangles; they should be small enough and even enough to be hard to discern in the finished product.