"Am-thinking of an AVR circuit to measure the characteristics... Which, then, is just a fancy logging ohm-meter with a synchronized "charger"... Heat the bulb, measure the resistance (inductance? thermocouple voltage?!) as it cools." 

If you're just interested in the changes of actual value of resistance (and not the speed of cooling when turned off) wouldn't it be quite easy to just first measure the fully cold resistance with an ohm-meter.  Then switch over to current mode on the meter and hook it up to a variable PSU and start measuring the current thru the bulb at 0.1 volts - followed by incrementing in 0.1 volts up to the nominal bulb voltage?

Easy to plot a graph of that.

Hey Matseng. It sounds like a good solution to a different problem...

Here, for my incandescent-DRAM design, the idea is to read/refresh the resistance of one bulb (indicating its binary value) at a time, sequencing through many bulbs. This means that once a bulb is "written", it will only be refreshed, say, 1/5th of the time. Then it will be isolated from all power, while the read/refresh circuitry handles each of the other bulbs sequentially, until the cycle repeats.
This is similar to capacitor-based DRAM, except that where a capacitor holds a charge that slowly leaks when not powered, here a lightbulb holds a resistance that slowly decreases as the heat stored in the bulb's filament slowly cools when unpowered. 

My plan, here, was to determine how long I could go between refreshing the bulb's (hot=1) value before it cools too much that a hot bulb can no longer be measurably-discernable from a cold bulb. This would, essentially tell me *if* it's possible (with these bulbs) and how many bulbs I could refresh sequentially.

The idea, then, is to heat a bulb and measure its resistance as it cools, thus determining how long the bulb can go unpowered and still read "1." Which, of course, is a bit tricky since resistance-measurement usually implies applying some current, which would keep the bulb warmer longer. 

But, here, it's mostly just a matter of proof-of-concept... if my refresh circuitry is too slow, remove a few bulbs so it has fewer to cycle through...

There are, however, a few other characteristics I'm not sure about... I ran this experiment, by-hand, with a larger, much-higher-wattage bulb, which literally took about three seconds to cool to the point that it was no longer clear, by measuring its resistance, whether it had been powered three seconds prior. 

Handy for proof-of-concept, as I could literally power it off, then fumble with my meter's leads, and still get some useful data-points. 

However, a very unexpected discovery: The bulb's resistance went *up* briefly, after power was removed, when it clearly should've begun cooling. I don't know why, maybe the filament *resistance* went down as expected, but the filament may've also acted as an inductor, forcing current into my meter. Or, maybe the dissimilar metal interface with the bulb's metal base caused a thermocouple-effect, injecting voltage into my meter. And, of course, there's also the possibility the meter's current had some effect (on an 8W bulb?).

Regardless of the cause, this or other/similar effect may exist in my new, much smaller bulbs--whose heat/whatever capacity is far too small/short-in-duration for me to measure with equipment on-hand.

As, presently, I don't have the equipment, either, to actually build prototype[s] of my relay-based read/write/refresh circuit-design, I'm working with simulation-models which don't seem to include these unexpected effects... So, it stands to reason real-world bulbs may differ slightly from ideal models, just as real-world capacitors leak differently or have different series resistances. The thing about lightbulbs, however, is that as far as I'm aware they've seldom any reason to be specified (in datasheets, etc.) in such detail. (similar problem with relays, but we're not there, yet). 

Thus, I thought maybe it would be useful to do some characterization of the off/cooling-time effects, e.g. with an AVR.

But, the fact is, the measurement circuit, itself, likely will play a major part in all of this. E.G. my relay-based measurement circuit injects some current (enough to power a relay!) through the bulb, albeit *very* briefly, which will cause even an OFF=0 bulb to warm up a bit. Plausibly more than whatever an AVR/ADC-based resistance-meter might inject constantly throughout a much longer series of measurements.

And, regardless, the measurement techniques, currents, durations, periodicity, etc. are so different that, realistically, I think it's really just a matter of giving it a try, and modifying things like threshold/calibration resistances, refresh frequency/duration, and even the number of bulbs being sequenced, based on how it functions in "the real world." Maybe a tiny "glitch" occurs immediately after the bulb's power is cut, but e.g. by the time three other bulbs have been refreshed, that glitch may've long-since dissipated. Or maybe the measurement-current applied three refreshes later will be *slowed* by inductive-effects (just imagined), which wouldn't've been accounted-for by my original plans for this AVR-meter.

I'd say it's a bit of a crapshoot, I might be able to simulate the relay-circuit's loading/injection characteristics, somewhat, with an AVR-based circuit... Or just do some worst-case measurements to make sure it *can* work... 

Or I can just put 'er on the ol' backburner 'till building (and testing) prototypes is more feasible. 

I got here as a tangent to a tangent to a pre-necessity to continuing-on with #Floppy-bird, which is just waiting for my return!

But, I'm always interested in stirring back-burner pots! Thanks for the ideas!

Had another thought, largely-inspired by your comment, thanks...

If the "glitch" is inductive or thermocouple "injection", the effect on a series 5mA=on relay-coil would probably be negligible, whereas the effect on a resistance-meter using a 1uA measurement-current would be tremendous.

Good call.

And great to hear from yah, been a while!