1) Kodak/TrueSense and Teledyne-Dalsa have been working on improving their CCDs for a very long time and still have far higher read noise than CMOS sensors, so I doubt it is easy as you think.
I never said it is easy, on the contrary I said it increases the price considerably.
2) Active pixel CMOS sensors are coming to astronomy now: Canon has make some, and Teledyne-Dalsa now makes very large and custom-sized CMOS sensors for such uses. Meanwhile, the CCDs used in telescopes do not have inherently lower noise or inherently higher DR; quite the contrary. Telescope sensors reduce dark noise by active cooling, and achieve very high per pixel DR by having huge photo-sites; both equally possible with CMOS sensors. The persistence of CCDs in astronomy is probably for the same reasons as in MF and medical imaging: the initial cost of changing to CMOS is a great barrier for lower volume products, and the simpler full frame transfer CCD design has lower overheads when designing and putting a new model into production, and so are more cost effective to make in small quantities. The recent rise of lower-volume active pixel CMOS sensors (like the CMOSIS one of the Leica M 240 and now this one for Hasselblad) suggests to me that the cost barriers are coming down, and if so, CMOS will soon spread into many of the areas where CCDs survive for now.
Possibly, but if there were so much to be gained, it would have been made earlier. I think I did not express what I meant clearly enough: there certainly are possible theoretical gains in having the converters as near to the sensels as possible and cmos sensors try to capitalise on these gains. It is just that these gains are not as large as people fantasise about. Here people talk about MF backs being only good at base iso 50 or 100 and hope for "clean iso 1600", a four stop increase. On dpreview people write about new cameras being as good at iso 6400 as old ones were at 800, a three stop increase. I think that this is wishful thinking.
My estimate between the 2007 ccd H3D-31 and the 2012 D800 is, considering the smaller pixels, that we had maybe a one stop increase. It is remarkable and certainly better than nothing, but not groundbreaking.
3) The active pixel CMOS sensor in the D3X still uses on-sensor change gain as described in my previous post, and this is a major factor in the low noise advantage of active pixel CMOS sensors. if it does anything off-boardm it is only ADC (which Canon still does off-board with its CMOS sensors.) Anyway, AFAIK that "off-board ADC in the D3X" is only a rumor; another explanation is that the D3X simply adds the option of a lower ADC ramping speed in its column-parallel ADCs to get its 14-bit output at a lower frame rate.
Yep, Erik suggested so much. That would be a very clever trick indeed, but would only provide real gains if the ramp is accurate enough.
