Here's another idea for the super-foamers out there. You know who you are...
Along with working air brakes, having a locomotive that is able to roll freely is an important part of being able to realistically model the way a real train is run. The difference between that and a worm-driven model is fundamental, and affects a good deal of how we control and operate our layouts.
Step back for a minute, and ask why anyone would care. What we have no works well enough, right? Well, yes and no. What we have works, and we accept it because that's all there is. But what we are really doing is modeling the conductor's half of railroad operations. We run the train where it needs to go, throw switches, spot cars, do paperwork, handle meets, etc. We simulate fairly well the job of the conductor (with brakeman, flagman, switchman, etc. thrown in, and occasionally yardmaster, dispatcher, and trainmaster as well). What we do not model at all, however, is the job of the engineer. There is no skill, no knowledge of the territory, no sliding in to a stop at the last possible second, no runaway trains, and no train handling or schedule keeping. A single track main line on a layout serves two functions - to separate layout design elements, and to have a pretty place to watch trains go by. That same track on a real railroad is every bit as interesting and challenging as a complex yard, from the engineer's seat.
Also, we have sound decoders which simulate the sounds of throttle notching, dynamic brakes, steam locomotives drifting or pulling hard, etc. The sounds are not completely appropriate, because they are interpreted from only one thing - motor voltage. There is in reality only one control on even the most complex of control systems - motor voltage. You show me the most sophisticated computer controlled system the world has ever known, complete with momentum controlled by the weight of the train and simulated automatic and emergency brakes. I'll pick up my home built transistor throttle, and I can do everything that system can do just with my direction switch and speed knob. If I want to simulate momentum, I turn the knob slowly. If I want brakes, I turn it fast. If you want to be an engineer (I know I always did, from the time I was a toddler), realistic operation would be a great thing.
Okay, you want to model the engineer's job, too. How do you do that? As I said earlier, the key (in my opinion) is a locomotive that can roll down the track freely with the power off. Because there's no worm gear to hold it, it must have brakes, which is why I say a working brake system is the first step - you can use brakes with the system we have today, but not a freewheeling locomotive. The challenge will be in packing a suitably powerful motor into the tight constraints of our models. Turning it lengthwise has some significant advantages, since most of our models are long and thin, and can hide a bigger motor better that way. This is probably the only real difficulty, and it may be a blessing in disguise. Our models can generally pull far more than their prototypes ever could, especially up staggeringly unrealistic grades. Scale traction motors (and I know of at least one company in the US working on such a thing for EMD modelers, and another that makes a setup which might work) would be less powerful, but that might force us to model things a little more realistically. Since this would really only appeal to those crazy guys for whom an out of scale crosstie causes indigestion, they would consider that a feature. One option I have considered is a three phase AC motor, if it could be made small enough. I suspect that it would give better low speed torque, but I don't know if it would be worth the effort. One advantage of a 3 phase motor is that shorting two of the phases will more or less lock the rotor, which might work nicely as a parking brake. Of course, that would absolutely require a new control - a DCC decoder could be made with a 3 phase inverter, but none exists as far as I know.
The other issue will be control. While you could continue to use the same voltage and polarity control (for DC traction motors, at least), the real advantage would come from a DCC decoder (or something comparable) which can handle things like regenerative (dynamic) braking. If the motor(s) had field windings instead of permanent magnets, you could simulate the use of variable cutoff and throttle in a steam loco. Diesels could of course have working dynamic brakes, and multiple traction motors could mean the need for making transition, just as we do full scale.
And then there are the sound effects. How would it be if, instead of turning a throttle to speed step 23 to get notch 2, if you simply put it in notch 2? The sound would go to notch 2, but the engine would not instantly go to "notch 2 speed" - it would start rolling, and go as fast as the grade and tonnage allowed. On a steam engine, imagine throwing the reverse into full forward (with power reverse sound, if appropriate) and cracking the throttle, then slowly working to full throttle and a short cutoff as the train picks up speed, all with the sound matching perfectly - long deep chuffs at full stroke, tightening to little puffs as the Johnson bar moved toward the center, and even next to nothing (or even drifting valves popping) when the engine moved with the throttle shot off. It could all be integrated as easily as today's sound and power decoders are, and the sound would not be a simulation - it would actually reflect what the engine was doing.
