How about it? Experiments of the third kind , take 999999. Update Hotplate design.
After entering the results of the heating experiment into the neural net (mine), it became clear to that neural net that things should be done differently than a simple resistor heating the air in a box as a means to keep component temperature stable.
So then, it was shown to me, that if we mount all the components that we want to be at a fixed temperature on a 'hotplate', in this case a small aluminum plate, and heat it with a nice TO220 type transistor, then if we keep 'hotplate' free from the box, we can make a nice even temperature with very little input power for the heater. So after some sawing and drilling I changed pictures of solution presented by the neural net into something real in the material world:
Anyways, below the hotplate is the hotplate heater:
So next is superglue, or / and maybe 2 components glue, we have that at hand, There are of course more components, but mainly some small capacitors and some very small SMD resistors that can easily fit between IC pins etc.. I will not make a guess how much less power we will need in this setup to keep say the tritium tube at 40C, but ... OK. You can do and place bets of course, if somebody takes them. I think this is a cool circuit.
[1] The big thermal leakage is now likely to come from the wires going in and out of the black box. we need: ground power +4V (heater) power +3V (chips) I2C data perhaps I2C clock perhaps RS232 logic level in (MAX232 is on main board RS232 logic level out Temperature warning LED out (can be done with I/O expander on main board via same I2C). So basically if I put the EEPROMs in the black box, I can save 1 copper wire. And of course what I forgot, but I do not know that yet :-)