And THAT is why I’d rather have a 12MP Nikon D3S than a 24MP most anything else – because the D3S has a sufficient but not overly large # of pixels spread across a BIG HONKIN sensor.
This is why I was vexed to see Red spec out Scarlet at 5K for S35 sized sensor – I’d MUCH rather have 3K Bayer (to generate a sharp 1080p) with more dynamic range, better low light performance, and less rolling shutter (aka image skew when panning) than a 5K similar sensor.
EDIT – see Graeme’s response in the comments – it is more subtle than just these issues. Rolling shutter is read/reset, not total pixel count. More pixels may make it more difficult to reduce the read/reset time (I am GUESSING), but higher resolution doesn’t HAVE to mean lower read/reset. My bad. Higher resolution also offers the benefit of reduced aliasing, and cleaner signal/noise once scale back down to target size.
I’d been meaning to write an article about this, but like most things, if you wait a bit, somebody will write a better one than you were going to.
I am, apparently, quite the lazy writer.
My backup on this? Arri went with 3K for their next gen digital camera – dynamic range and better high ISO performance beats “high resolution” any day. EDIT – again, see the comments.
The human eye sees contrast and color before it sees detail. That’s why I’d rather have better dynamic range and high ISO performance (with low noise, please) rather than a massive megapixel count on my cameras, be they still or motion. EDIT – results vary – see comments again.