Ran some sims... seems to work.

They're on the computer, kinda too beat to start it up.

Got a bit of feature-creep.

Reading, writing, and refreshing a single bulb requires a single relay, a resistor, pushbuttons, and eyeballs, or hands...

Adding multiplexing for three bulbs adds a rotary-switch for refreshing and another for selecting which to write. This topology allows refreshing in the background without interfering with write... so, throw a motor on the refresh-selector and have at it.

Presently the full refresh-cycle runs at 10Hz. Bulbs rated for Roff=10ohms and Ron=150ohms are asymptoing to around 45ohms off and 116ohms on.

....

Now I get a bit carried away...

wouldn't it be nice to automate the write process? Have it automatically stop writing once the bulb cools/warms enough? Bam: a couple additional relays and some boolean logic (XNOR!).

Ah, let's speed up the process, no need to write a zero when it already is! Sample-first-then-write. Uh-oh, bug. If you start a write of 1 to a 0 at just the right time after its previous 1->0, and at just the right time during refresh, it *just barely* crosses the 0->1 threshold, stops writing, and cools back to 0.

Part of the point of automated-write-testing/completion was to assure such things *wouldn't* happen, e.g. in extremely cold weather. And, ironically, maybe, attempting to avoid that scenario *caused* it.

So, now a further feature-creep... Hysteresis... Let's set a lower threshold for Roff and a higher one for Ron, but only when *writing*. No prob, we already have the relays... two more resistors and add another pole (3PDT, now!)

Or... yahknow, it worked before because it wrote for a while before testing... could just have an arbitrarily-long write-process with a single R/C circuit....

I think I've forgotten some intermediate feature-creeps... But, the point is, it's automated and simulated functional...

A rotary-switch, another on a motor (or a drum with magnets and reed switches), a few relays, and there yah have it, heh!

..

Oh, interesting observation: I used a decade-counter (4017?) to simulate our drum-switch... And using the reset terminal causes the first bulb's refresh-timing to be slightly shortened.

This is actually quite telling, as it reaches asymptotes of something like 10ohms lower than the others... Yet it remains stable and still far from the threshold (currently ~60ohms). 

Thus, this seems way more resilliant than I expected, and once that timing glitch is fixed and write hysteresis is functioning (heh!) it should be able to handle at least a few more bulbs. 5x7 may be realistic.

....

Oh, a few feature-creeps actually resulted in a simpler system... That was cool.

I should throw up screenshots and simulation files (TODO).

...

Actually, it seems resiliant-enough that I've pondered multiple values per bulb...

Presently the sims are for a 5V-only system... 5V bulbs (160mW seems reasonable), 5V relays (600ohm coils, similarly reasonable)... But what about going back to an earlier idea...? 120V bulbs, 5V measurement-system... And various DC voltage-sources (rather'n resistors/current sources) for refreshing various values... they can't get nearly as hot at, say, 60V as 120V... Pretty sure bulbs aren't a perfect thermos, and photon-ejection must require some power, otherwise 5V in a 120V bulb would eventually glow white-hot...

Relay-measurement, still? Also, 'spose similar could be done with 5V bulbs and a lot more precision.