« on: December 31, 2014, 10:12:43 AM »
The 5D4 should also include a bump in FPS and buffer performance. I don't see this happening with 52 MP. All I see with 52 MP is HUGE RAW files, maybe 90-100 MB each. That's a lot for in-camera crunching and recording to cards.
I think you're grossly overestimating the size. On a 6D, the average image size ranges from about 21–25 MB for 20.2 MP (with occasional spikes up to 30 MB). The reason for the variation is the embedded JPEG. The RAW data takes about 1 MB per MP. So I'd expect a 52 MP camera's files to be 52-57 MB apiece, assuming they do nothing whatsoever to improve their compression ratios, and assuming they keep the embedded JPEG images at the current resolution/quality rather than scaling them up for no apparent reason.
He does - a lot.
This is what you get when stretching the Nikon 810 to the max with 14 bit lossless:
14 bit lossless file size: 42 MB
RAW file converted in DNG 36MB
So 52 - 57 MB a piece sounds right to me.
Also, while many here complain about the "monster" file sizes a 50MB camera would produce I have lots of Photoshop files >100MB and quite a few >250MB.
Personally, I'd be happy with a 36-40 MB 5DIV if it came with significantly better dynamic range and high iso (and not another incremental crawl such as what we got with 5DIII).
Going on what I've gleaned from many many debates on here and elsewhere, there won't be a big leap in high ISO, because we're already reaching the limit of what the current sensor technology can do (cue jrista and quantum efficiency). I think people have to accept that if you want to go a lot further with low light, you have to start looking at dedicated solutions - like the amazing video sensor Canon showed off a while back, or a bigger sensor, like the 645z. Standard DSLRs can't make leaps in every area forever (until and unless someone invents a new type of sensor).
I have an idea how to increase sensitivity by a factor of three: put an optical grating in front of each pixel and use 10 or 20 photodiodes per pixel to make a rough spectral scan. No loss in a color filter array. Different angles of incidence (of different optics) can be corrected by software ("which photodiode corresponds to which wavelength). No longer 3 color channels but the chance to track colors more detailed.
The only problem: I have no machine at home to produce that sensor and I am shure no one has. But perhaps just a matter of time because a 52 MPixel DPAF sensor has 100 Mio photodiodes to read and this seems possible!