Originally posted by Christopher M.W.T But is the K-5's [low light, ed.] video mode majorly improved or not?
I guess this'll be moved into the video section soon
The real question is about the kind of subsampling done in video mode. I'm one of a few who tried to figure that out using zone plate test charts.
Once you know the subsampling factor you can derive all other properties from it.
The 5DmkII seems to do a kind of "read every 3rd line" skipping but reads all rows. That's a subsampling factor of 3. According to my own research, the K-7 has a subsampling factor of 6. However, the K-x seems to have an even higher subsampling factor giving it a worse low light performance than the K-7 despite its good low light capabilities in still mode. My guess would be 9.
In still mode and at ISO 1600, the 5DmkII has about 1 stop advantage over a K-x, and 2 stops over a K-7. Add the subsampling factor 3/6 and you end up with a low light advantage of 3 stops in video mode for the 5DmkII over a K-7.
As for the K-5, I expect the performance at ISO 1600 to be only 1 stop behind a 5DmkII (in still mode). In video, it depends on the subsampling factor (9, 6, 4 or 3?). There are rumors that Sony worked on this but who knows ...
Will be interesting to see real world footage of the same scenes.
BTW, even a subsampling factor of 3 let's you loose 1.5 stops over still performance, i.e., even a 5DmkII isn't a good performer in video (for a given lens surface). And a Bayer filter compared to a trichroic prism in a 3 CCD camcorder let's you loose another 1-2 stops...
[so, it looses 3 stops compared to the feasible. But the large chip being much bigger than only 8x the surface makes more than up for this]
UPDATE:
Just one more info:
Some where speculating Sony uses a lower subsampling factor because using Exmor HD in a dedicated video camera, the NEX-VG10. That doesn't seem to be the case. I.e., the Caprock filter (
http://www.caprockdev.com/antimoire.htm) still provides the usual level of moiré and aliasing reduction, search vimeo for it. With a subsampling factor of 1, the Caprock filter would have no effect.