If you thought my four-bit decoder and seven-segment-display controller were pointlessly convoluted, you'd be quite right, but this was nowhere near the maximum insanity in this design. I was unhappy with wasting so many mechanisms linking the thing up, considering it's totally possible to drag four distinct paths over a single pressure plate. My seven-segment display codeset called for a grand total of 34 signals. One segment gets switched off by three, one by four hexadecimal numbers, three by five and two by six. So i sat down with a piece of paper, sketched out the basic direction of operations and went mining.
The result is impractically big and slow, but the cart has enough speed to get through any reading path and results in the correct displayed symbols, with no more than seven pressure plates (and seven connections) for seven segments.
Paths Pressure plates and bridges
The test run (going through all sixteen possible inputs) took almost two weeks.
Parts:
decoder - four input-operated bridges, seven pressure plates; one pressure plate, one hatch, one grate and one bridge for regulation/latch reset. One minecart to run the circuit.
latches - fourteen hatch covers; seven linked to the reset pressure plate in the decoder, seven to set, each run by one pressure plate in the decoder. Seven pressure plates to run the display segments. Seven minecarts.
display - seven bridges
operation - four input levers, one operation lever
58 mechanisms for the needed 29 trigger/building links. Zero power.
EDIT: i shortened the path of the seven a bit, but couldn't find anything else that i could easily improve on. The two longest paths, 1 and c (using "c" instead of "C" to better fit between "b" and "d"), are 72 (i think) and 85 tiles from the last bridge passage to the last pressure plate activated. "c" takes ~480 game steps running the proper circuit until hitting that pressure plate, but another 180 to get back to the starting hatch from there, and the regulator forces a start delay of a bit over 200 steps to actually allow working with bridges for display; so in this case, one read comes out at ~900 steps. I wasn't exagerrating when i said "impractically slow".
PS: Hooray! I got the start to last plate period for the c down a full thirty steps, and another thirty by straightening a bit of path (not shown). Now it only takes 420 (or slightly over 800 for the full read). I also managed to reduce the overall width by another three tiles, from 25 to 22. The decoder/display controller still stretches 34 tiles north to south, the display-control pathing grid itself takes up 18x22 tiles (396 area, 350 of which actual used track). I switched the pictures to this slightly improved version as of today, 3rd of May.