I agree that Canon lenses are very good. It's one of the reasons I've stuck with the brand as long as I have. For most astrophotography, though, I think that it's the sensor were after...and these days, Canon sensors (and more than that, their entire readout pipeline, including the off-die ADC units and DSP) are rapidly falling behind the competition.
I think most competitors have moved to non-die ADC units, highly parallelized if not per-column parallelized, are using much smaller transistors (some, like Samsung, are as small as 65nm!!!) and many are even using stacked or 3D fabrication to package all the image processing in the same die package as well. Canon is still using the same fundamental sensor fabrication technology (500nm transistors) and same fundamental system design (off-die ADC units and DSPs) that they have since the 10D (at least)!!! O_o
The problem with Canon's system design is the very long paths between pixel and ADC unit (since the units are off the sensor die on the PCB, the paths are VERY long indeed), and the high frequency of the ADC units (Canon cameras have 4, 8, or 16 ADC units, so each one has to process many multiples of columns worth of pixels). Long traces and high frequency components increase noise, and can lead to more patterned noise (rather than just random noise, which is easier to deal with.) Sony's Exmor sensor moved to column-parallel ADC units on the sensor die. So if there are 6000 columns, instead of each ADC having to process 6000/8 * 4000 or 6000/16 * 4000 pixels, each CP-ADC unit simply processes 4000 pixels. That allows lower operating frequency, which reduces noise. Since the ADC units are right at the bottom of each column, the path length between pixel and ADC is much lower as well, also reducing noise. Sony also reduced transistor size, to 180nm and I think even 90nm in their newest parts, which are significantly smaller than Canon's 500nm. That allows more die space to be contributed to light sensitive photodiode area, while concurrently allowing more on-die electronics.
Moving more logic onto the sensor die opens up the doors to a lot of other innovations as well, and Sony has employed a few in their Exmor sensor design, which is why their read noise curve was flattened (it's ~3e- across the board in most Exmor based cameras), and dynamic range at lower ISO increased by several fold.
One thing I do have to give Canon a hand for...they have done some amazing things with their old technology. They aren't exactly competitive today with their sensors, but they have somehow managed to keep their high ISO performance of their cameras overall competitive. I think with the 7D II that is more thanks to the DIGIC 6 processor, which does hardware NR on RAW pixel data, than to the sensor design itself. Still, it's surprising to me that Canon has remained that competitive with late 1990's/early 2000's sensor fabrication technology...technology that is over a decade old and which consumes a considerable amount of die space for transistors (lower fill factor than the competition.) I would love to see Canon move to a 180nm process...if they can eek as much performance out of that as they have with a 500nm process, they would be a force to be reckoned with...they just have to actually DO it, and stop all the posturing.
In addition to Sony, I think Samsung's new Backside Illuminated APS-C sensor that they have used in the NX1 mirrorless is something we should keep an eye on. That sensor has already tested very well, and technologically it is one of the most advanced sensor's out there. With 65nm transistors and the BSI design, it is putting far more light-sensitive space into the path of the light, so it's fill factor, and thus overall Q.E., should be higher than any other APS-C sensor on the market (including Sony's). I'd love to see Mike tear one of those apart and see what it is capable of. I would readily replace my 7D with an NX1, rather than a 7D II, if it is astro moddable and has the kind of sensitivity I think it does.
