Warning - this is another loooooooong post. I've been thinking to split it into two parts, but since it all goes together so nicely I put it all into one post. I hope you'll enjoy it nevertheless and find it's worth your time. Here we go. :-)
One of the things that I noticed after a few days of using the D700 was how unreliable the exposure control (picture review on the camera's built in screen aka "chimping") was - I was seeing the blinking highlight warning (indicating overexposure, which is very undesirable in the digital domain 99% of the time) and made another exposure - just to find out at home that the raw data was perfectly ok and not overexposed.
The reason for that is (besides the fact that the picture control image is always the JPEG): the D700 has a rather steep and contrasty tone curve (see the review on dpreview.com for proof - notice the dramatic difference of dynamic range covered in the JPEG and the raw data - I don't know what Nikon was thinking here...). And just like on the S5pro, I've been using the "Neutral" image rendering and set both brightness and contrast all the way down, but I was still seeing the blinking warning where no highlights were actually lost/overexposed in the raw data.
I already mentioned that I am kinda "spoiled" from the S5pro and it's incredible dynamic range, but I can't help but notice that I am spoiled even more by the way I could handle the exposure control on the S5pro - I've simply been using 300% dynamic range all the time, and if I saw any blinking, I could be pretty sure that I was over the limit (using 300% dynamic range instead of the full 400% was simply to make sure that I'd be on the safe side, always). So, with the S5pro, I could accurately judge from the preview JPEG and the blinking highlight indicator if my exposure was good or not.
Not so with the D700, unfortunately. There is NO way to get an accurate highlight control (for ETTR) if you use only raw data. And it's not only the tone curve that influences this, but also the white balance, of course - the camera shifts the three color channels R/G/B around, while the raw data might be very different - that's mostly because the Bayer pattern consists of R/G/B/G (there's two green dots for each red and blue dot).
The solution is to a) load a linear tone curve into the camera (something that I found I could only do with Nikon's "Camera Control 2" software which has to be bought separately - it's NOT included with the camera! I really hope that I was just too dumb to find that option in the software that is supplied with the camera... but anyway - the good thing is that they offer a 30-day trial which has no other limitations) and b) use a technique called "UniWB" - simply said it's using a, sensor and exposure wise, absolutely neutral white balance (Lightroom says it is a color temperature of 2900/-131 - yes, very weird...). The resulting preview images look "interesting" because of the "double green" effect:
JPEG from unaltered raw data (NIKON D700, 1/160s @ ISO 1000; f/8, 300 mm)
...but this is an absolutely correctly exposed image (histogram-wise, with the "expose to the right" philosophy in mind). The effect of UniWB and the green color shift is quite clearly visible. :-P Now, why the hell would you do that to your photos?! Well, just try it for yourself in Lightroom/ACR: change the white balance, and notice how the exposure seems to change, too - sometimes so much that there will be highlight clipping. And that's exactly the problem (besides the tone curve) that makes the camera's highlight warning blink when it shouldn't.
I won't go into details on how to load a linear tone curve into the camera, or how to get UniWB - it's all over the internet (I found this article from roesand.com very helpful).
The problem with UniWB and the linear tone curve is: the blinking highlight display is rather useless now. If you look at the histograms, you'll see that it "kicks in" waaaay too late (I mean: it's NOT blinking now even if one channel is already blown out because the highlight warning uses the combined luminance histogram for the indication).
But the good thing is that the histograms are VERY accurate now thanks to the linear tone curve - you can precisely control if any of the three color channels is blown out or not. And that is VERY helpful. I've switched from the simple highlight warning picture control to the color histogram picture control and just check the histograms now. If only these histograms weren't so tiny and would have a better indicator for a blown out channel! In direct sunlight, it can be difficult to recognize a blown out channel.
And so far, I also noticed that the D700's exposure in Matrix metering is very accurate (at least for situations with a rather controlled contrast range). Most of the time, the histogram will be nicely on the right and not clip at all.
One final note: since both UniWB and the linear tone curve are user-presets, it's always possible to simply switch back and forth between those and one of the camera's built-in picture optimizations and normal white balance for the occasional JPEG-only snapshot session.
And last not least, here's one possible rendition of that photo - the result of using a Kodak Ektachrom 100VS cross-processing film simulation - yes, there's still a strong green cast over the image, but trust me, the white balance has been corrected. :)
Spring morning Wöhrsee detail (NIKON D700, 1/160s @ ISO 1000; f/8, 300 mm)
Closing words: I hope that camera vendors will think of the "raw only" photographers a little bit more in the future and provide them with more useful tools for exposure control that do not require the fiddling with UniWB and tone curves...
(Gee - my blogposts tend to become longer and longer...)