Monday, February 22, 2010


The layout has to live in a small space (ideally 12"x32"). The list of gauges which can make a 180-degree turn in 1' is really short. Z gauge can do it with Marklin 8510 (5 3/4" radius) track. I can probably get a locomotive that'll handle that turn, but ... Z gauge is really small. Really really small. Putting any electronics of my own design inside a Z gauge car is likely to be really difficult. Add to that the fact that Z gauge is really expensive. Locomotives and cars seem to be twice or more expensive than their N gauge counterparts. Finally, the Z gauge world is also small. There's Marklin, MTL (Micro Trains), and AZL (American Z Lines). Oh, and Marklin just declared bankruptcy. I had an Amiga once. I don't need another.

N gauge is bigger. Bigger minimum radii (well, except for a Japan-only manufacturer named Tomix which sells 4" curves), more manufacturers (Walthers lists 2,564 locomotives in N gauge versus 144 for Z gauge), and lower prices. It's also the second-most-popular gauge in the world, next to HO. But again, the problem is with the turning radius. The smallest radius sold is the Kato Unitrack 216mm (8 9/16"). It's unfortunate (we end up with a layout that's 18"x40"), but bearable.

I don't want to use the Kato Unitrack for the real design. First of all, it commits me to designs which can be implemented using their hardware. Second, the track has integrated ballast, which always looks weird. Assuming that Atlas Flextrack can accomodate the turn radii, it should be sufficient. Unfortunately, 8 1/2" isn't the tightest turn on the layout. The turns from the switches to the crossing are sharper. I still need to find precisely the right position (possibly the angle too) for the approaches to the crossing, but I know of at least one solution which gives greater than a 6" turning radius on all four corners.

The trick now is to find a locomotive and cars which can handle a 6" curve. An initial investigation suggested that the Kato EMD NW2 was the way to go, but then I found another page suggesting that the NW2 wasn't very good at all when it came to extreme radii. That second page said the Atlas GP-35 had no trouble with 6.4". Hopefully it'll work for me too.

The Big Project

So I've decided to teach myself enough EE to let me build circuits (and, gasp, understand what I'm building). This means working my way through The Art of Electronics. It's a really good book, but it's extremely information-dense -- especially the first chapter, Fundamentals. One nice side-effect is that I now understand how trig and complex numbers are actually used in the real world. I've seen trig before when doing 3D graphics, but I hadn't seen complex numbers since ... high school? It's all very exciting.

But what will I do with this new-found knowledge? I worked my way through Computer Organization and Design while I was at Sun, and found it fascinating. I never got very far into Computer Architecture: A Quantitative Approach, but I'd like to. Either way, I'd like to use my EE skillz to build some of the circuits described in those books. Eventually, I'd like to build a processor. Even if it doesn't have a head (though that would be nice), being able to telnet or ssh into hardware I designed would be quite cool.

That's the long-term plan. What about the short term?

I just finished reading Playing With Trains, which reawakened my love for model trains. Gigantic layouts just dripping with scenery and detail are great fun, of course, but I've neither the patience nor the artistic eye for the scenery. Instead, I like the complication of the track plan.

What if I could merge model railroading, digital electronics, and software?

I'd like to build a reasonably complicated train layout that is automatically controlled. The idea is this: Start with a figure eight superimposed on an oval. Add two trains. They'll run completely automatically, without human control. Having them run in the same direction is boring. Much more fun if the first train runs on the oval while the second one runs on the figure eight. A commodity microcontroller runs the show. Its job will be to coordinate switch positions and throttle settings to keep the trains doing their thing without colliding. The faster the trains can go, the better.

Here's a really ugly picture of the layout I'd like to build/automate:

Something about the idea of two trains whizzing around that layout without plowing into each other, speeding up and slowing down on their own, switches magically switching themselves ... so awesome.