What I'd expect - or at least what I would do in his place - is to use this technique for building test prototypes. (And that's prototypes in the electronic sense, not the model railroad sense!) After all the bugs are ironed out then I'd go ahead and put the effort into building proper scale representations.
There are some references to use of Tomix 3, 4, and 5-color signals in the programs, and a picture on one of the Japanese pages, so he may be planning to use those (or just building in compatibility for them).
He also has
a page on converting the Kato 3-color signal to DCC operation, and this includes a reference to the Tomix 4-color signal.
Did anyone see what he's using for train detection? I couldn't find anything specific, but I haven't gone all the way through the details yet.
There's a "train detection" element on board "A", which is simply a TLP521 photocoupler. If I'm reading the diagram and program right, it pulls input RB4 on the DCC decoder PIC low when current is flowing to the rails (it's labeled "stop trigger in" on some diagrams, and "s-in" elsewhere). See the diagram labeled "1: Type-A element" just below the "How to assemble signal board elements" heading on the English page: the circuit labeled TPS521 is the photocoupler that detects current. It looks like there are multiple generations of his board described on these pages, so that circuit may have changed, and I'm not sure why he'd referencing a TPS521 on the English page (which is some kind of microcontroller, not a photocoupler, as far as I can tell).
He says in one place the type A board can have up to three onboard detector circuits (one is typical). However, the board also has inputs for external detectors to supplement the onboard ones (needed with some of his designs for multiple blocks controlling a stop relay) and for a photodiode input (for precise position detection when stopping trains). He mentions use of the Digitrax BDL16 as a good supplemental detector.
I may have misinterpreted some of that, as I'm no electronics whiz, and he didn't document all of the elements clearly (machine-translations of his Japanese pages don't help me either). But I think I understand the rough outline of what he's doing. Not well enough to build one, but I'm more concerned with understanding how he wants it to work, than with making one.
