Latest update, 6th Oct 2018 (scroll down)
Hi all. There is a lot of pixel peep flapping occurring (as usual) online surrounding the new Nikon Z7 camera.
As always whenever I get a new camera I compare the NEF compression options on a high dynamic range scene to make sure my current thinking holds. That thinking is I may as well shoot 12 bit compressed raw. The files are much smaller and you loose basically no quality even when pushing the raw files to levels you never would on a real image. I have checked and confirmed this many times and always come to the conclusion that 12bit, compressed is the way to go. My NAS server stays less full, I need fewer cards and the buffers on my cameras are subsequently much bigger in a burst sequence sense which is much more important that recording more sensor noise that you don’t see.
In testing the raw options (see my soon to start Nikon Z7 review diary for many examples) I observed that the mild banding issues the on chip phase detect sensors cause when pushing a non detailed shadowed area by 5 (yes, 5!) stops pretty much goes away when shooting in 12 bit mode. I presume the resampling of the sensors native readout to 12 bit does something to remove, or should I say mask it and it appears to have no impact on details in the image. See below.
Happy shooting all and if you disagree with me, that’s fine, let me know but please play nice.
Check back soon for some Z7 photos with actual animals in them (imagine that!) or take a peek at my Instagram.
Rich.

14 bit left, 12 bit right (both uncompressed).
Click to see the full file un resized.

Also note that for example, Arri Alexa saved raw files (that top end film and tv use extensively) store 12 bit files, it’s a lot of data. 12 bit for example has 16 times as much data as a 8bit (say jpeg) image.
14 bit uncompressed left, 12 bit compressed right.
Click to see the full file un resized.
UPDATE: 6th Oct 2018
I have been looking into this further and thanks to a user over on dpreview (see the thread here) there may be an issue in 12 bit with large shadow areas that contain little details (ie. large areas of a bland wall say) that have a green hue when really (really) pushed. I’ll run a few more tests, chat with Adobe and Nikon to see what’s what. Very curious!
I’ll also investigate how this effects real world images vs static tests. To that end until this is resolved I will be shooting 14 bit compressed just to be safe.