The Story of Hyde Yard
Part 2 of 4: The Concept
Automating a staging yard doesn't sound all that difficult: toss some switch machines, some sensors, and some electronics into a blender and there you go. This might be fine when a modeler's locomotives, rolling stock and turnouts all conform to exacting standards. But not everyone lives in such an ideal world, and since Rick has been modeling in N almost since the birth of the scale, he had an enormous roster that would be impractical to fine-tune to precise standards. Furthermore, there's no such thing as a "perfect" turnout, meaning the chances of derailments could never be zero.
Thus, the only way for an automated staging yard to operate reliably was to eliminate the turnouts. And I had an idea how to do that. It took quite a while for me to articulate my idea to Rick, but he eventually caught on, especially after I coined a very odd phrase: "half-turntable multi-stub switch machine."
This was the idea: take a length of flex track, secure one end to the layout, and attach the other to the end of an arm that swings in an arc across the stub ends of the yard tracks. Done right, the rail ends would align precisely, in theory eliminating any chance of derailment. At the same time, a moving arm would be simpler to control than a gang of twin-coil switch machines.
I returned a few nights later with a proof-of-concept device made mostly from styrene. Although it didn't function very well because the styrene was too flexible, it at least demonstrated the principle. I made the next machine from clear Plexiglas: it's more rigid than styrene, and relatively easy to work; plus, it provides a full view of all parts from any angle.
Aligning the track ends precisely relies on a cam and roller system: a spring-loaded roller on the swing arm drops into detents in the cam and locks the arm in place. The trick to achieving such precision is to build the entire mechanism first, then attach the ends of the storage tracks to the mechanism as it moves to each position on its own.
One other trick required to ensure precision is to deliberately make the motor drive "sloppy" in order to allow the roller and cam to control the final resting position of the arm. I did this by designing a strange device based very loosely on the principle of a "Geneva drive." This is a self-indexing mechanism that doesn't move an object continuously, as gears would; instead, it moves an object incrementally, pausing after each rotation. By making sure the drive parts have considerable play, the drive then disengages at each index position, allowing the roller and cam to properly engage.
All of the above assures virtually perfect track alignment without the need for any precisely-machined parts; indeed, I built the mechanisms entirely using hand tools. It's also worth pointing out that any derailments occurring over the machines' lifetimes have been due to operator error.
I built the first control system using TTL-logic chips. Although the prototype worked flawlessly on the workbench, once installed on the layout, it went absolutely haywire. (Owing to this initially inexplicable split-personality, the controller was appropriately dubbed "Doctor Jekyll.") Ultimately we determined that the cause was insufficient noise filtering, the noise being generated by passing locomotive motors. By then, however, I'd rebuilt the brain using relays exclusively, since they're immune to most any form of noise. Plus, relays allow one to see when they're energized at what times, whereas IC chips constitute a black box. The relay-version of Doctor Jekyll still operates the yard today, more than 45 years later.
Train detection is handled by infrared LEDs and sensors. It didn't dawn on us, however, that the yard would still be operating past the lifetime of the IR LEDs. And so they've had to be replaced—not a big deal, except that when they first began to fail, it wasn't immediately apparent what was causing spurious faults.