To begin with, the most common mistake I see when people are trying to ETTR
(expose to the right) is that they are applying WB gain to the equation and
stop when one of the channels appears to be clipping based on the embedded
JPEG. Usually this is the red channel, and you read it all the time when
someone warns "watch the red channel on the histogram". If you are shooting
JPEG, then this is good advice; but ETTR is for shooting in RAW and not
suitable for shooting JPEG.
To better understand how your camera is actually exposing the scene, WB gain
needs to be ignored. Consider the BFA (Bayer filter array), with its 2 green
photosites for every blue and red photosite; this means the sensor is twice as
sensitive to the green channel as it is to the red and blue channels.
Therefore, the red and blue channels have to be amplified by the camera's ASIC
for OOC (out of camera) JPEGs or by software as a part of converting the RAW
data. This is why noise comes primarily from the red and blue channels,
because they are often underexposed.
One solution is to only look at the green channel in the histogram being read
from the embedded JPEG. The problem is that sometimes you will end up
overexposing one of the channels and not even realize it. Now there's
something I bet you weren't expecting. As many of you already know, BFA is a
composite of the light reaching the photosite filtered to respond to one
color, but it is then blended with light from the neighboring photosites to
arrive at a value for that pixel that has all three color channels. Take a
picture of a deep blue sky in the afternoon and the composite value that makes
up the blue channel in the embedded JPEG will look good; but if you only look
at the blue channel without applying any red channel gain to the composite
pixels in the sky, you will see that the blue channel there has been
overexposed. Overexposing a color channel means that even though the
histogram says you have an accurate exposure, you have actually ended up with
a color shift (not enough blue in those pixels relative to red and green).
The problem is that there is no way to view the data accurately without
removing the WB gain.
To accurately see what your camera is doing therefore requires that you start
with a neutral WB, which is called a uni-WB. I am indebted to Julia Borg for
here graciousness in making the uni-WB available to me and everyone else who
has asked for it. I have some versions of the uni-WB for the D200 and D70
here: http://imageevent.com/tonybeach/mypicturesfolder/junk and they can be
installed directly onto those cameras by saving the original file and copying
it to a folder being used by the camera (e.g., DCIM > 105ND200) and then
accessing the image through the custom WB setting in the camera and applying
it. For those with other cameras that can use Nikon Capture, if you have
Nikon Capture, you can get the custom WB here:
http://www.pochtar.com/UniWB.zip and install it using Nikon Capture.
Some other image optimization settings and the selected colorspace and color
mode will also affect the embedded JPEG. Therefore, the most accurate display
needs to have the following settings applied:
aRGB
Color Mode II
Linear custom contrast curve
Saturation Normal
Sharpening Off
Once you are seeing what your camera is really doing, it becomes apparent that
some attenuation of the color channels is desirable for optimal exposure, that
way you keep the more sensitive color channels from overexposing while you get
adequate saturation in the less sensitive ones. This is accomplished by using
a CC (color correction) filter in front of your lens. The amount of color
correction is largely dependent on the camera and the lighting, but for a D2x
Julia Borg recommends CC40M and for a D200 she recommends CC30M.
Here's the problem, and the reason I mentioned Nikon's new cameras in the
subject line of this thread -- Nikon's new cameras cannot have custom WB
installed. Camera Control Pro is now required to access and upload settings
to any camera after the D200, and it doesn't do what could be done with Nikon
Capture's camera control module. This means that the D2xs, D80, D40, and D40x
do not have this capability. It gets worse, since Nikon is not supporting
NC's camera control module anymore, the feature is lost in Vista and any
future OS upgrades as well.
Recently I was unfairly called a "Nikon fanboy" for not conceeding that my
D200 was a piece of junk. Well, I think Nikon could improve their products
and I'm confident that they will, that seems like a reasonable stance on my
part. However, I am very upset with what Nikon is doing right now with NX and
Camera Control Pro; they have taken a $100 item that worked reasonably well
and replaced it with a $140 and a $70 item that for my purposes don't work as
well. I don't think it's unreasonable of my to implore Nikon to restore
custom WB settings to Camera Control Pro so that we can benefit from using uni-
WB in the future and make accurate ETTRs.