One last thing (actually a few things). Do you know if anyone has tried using these supercooled astro cameras for earthly subjects? Are they related to so-called 'starlight cameras' used for night time ambient light capture of wildlife? And in the long run, given all you've said, would a move to larger sensors (medium format, for want of a better term) be feasible for this sort of thing? I appreciate that requires larger lenses, but new materials and designs seem to be able to shave off weight, perhaps that could help?
Incidentally, I've been fascinated by this whole discussion, once we got past people moaning. So many fascinating concepts!
AstroCams are cooled and haver very good Q.E....around 70-80% for the good ones. They are CCD type cameras, and part of their low-noise is due to very slow readout. On most astrocams, readout rate is specified in megapixels per second. So, if you get a 5mp astrocam that reads out at a rate of 1.2mp/sec it actually takes four seconds to read one single image off the sensor. An 11mp astrocam might have a readout rate of 2mp/sec, so it takes more than five seconds to read out a single image. That's really slow, excessively slow by modern DSLR standards (which can read out as many as 14 frames per second, which on an 18mp body is a readout rate of 270mp/sec.) I suspect that requiring a slow readout rate is a limitation of CCD technology, and CMOS technology probably wouldn't need such a limiting factor.
Starlight cameras are different than astrocams. Older starlight cams are usually security slow-exposure video cameras, and use VERY LARGE pixels with VERY LONG exposure times. They basically take continuous frames at say 2-3 second exposure times. They often only have a couple megapixels at most, and can see down to 0.001 lux. There is newer starlight technology that can literally see "normally" under nothing but actual starlight (0.0001 lux), and have normal frame rates with normal pixel sizes (i.e. they can have many megapixels in small form factor sensor sizes.) Newer ultra-sensitive sensor technology is using materials, rather than pixel size, to increase sensitivity. SiOnyx recently purchased the beginnings of technology for black silicon sensor design, and they have turned it into sensors that can see exceptionally well in light levels that would render most things black to the unadjusted human eye (given about an hour and a half in starlight levels, and the human eye will fully dilate and we can actually see about as well under nothing but the illumination of stars.)
Black silicon is not really all that complex. It employs the same general concept as nanocoating on modern lenses. Multicoating uses multiple layers of reflective coating in order to cause negating waveform interference with light. Multicoating does not prevent reflection, only results in reflections largely being canceled out. Hence the reason why lenses with multicoating can still lose several percent to as much as 30% transmission in the worst use case. Nanocoating, on the other hand, prevents reflection entirely by eliminating hard transition points. Modeled after moth eyes, cones or rods of varying height (but usually no larger than about 200nm) are used to create a smooth transition zone that guides light in, avoiding reflection entirely. A lens that used only nanocoatings (as of yet, do not believe any such thing yet exists...nanocoating is currently employed on the most critical and largest internal elements, and never on external elements) would have grand total transmission loss of maybe 0.1% at worst, and maybe 0.05% on average.
Black silicon employs the same general concept...it is comprised of nanoscale rods of silicon that barely reflect any incident light at all, guiding the rest through into the substrate. This eliminates reflection off the sensor itself, greatly increasing the rate at which incident photons are able to actually reach a photodiode, thereby increasing Q.E. Technically speaking, the use of black silicon, along with properly designed microlenses to capture more high incident angle light, would increase the rate of photon strikes by so much that much larger photodiodes and even layered photodiodes would be essential in order to convert all those photons into free electrons and increase full well capacity. With black silicon, though, since it guides light through, layering photodiodes should be easier, as photons could penetrate much deeper into the sensor than normally. With multiple layers of photodiodes and little photon loss as they travel deeper into the substrate, full well capacities for small pixels could, theoretically be increased SIGNIFICANTLY...doubled, tripled, maybe even more. A sensor like the 7D's could be doubled from 21ke- FWC to 42ke- FWC, or maybe even achieve parity with the likes of the 5D III at 64ke- FWC.
That is, assuming someone like Canon picks up the technology and employs it. I don't think anyone is looking to black silicon yet for DSLR sensors, and who knows how long it might be before they do.
