Originally Posted by jonoslack
With the caveat that I'm a relative newcomer to using UniWB, I'd also like to chime in on the UniWB issue since Jono's and Edward's comments suggest that they are perceiving the issue in terms of color and white balance whereas UniWB is primarily about exposure and dynamic range (although these factors can obviously be impacted by one's choice of white balance).
Originally Posted by edwardkaraa
As I understand it, the main problem is that -- even if you shoot RAW -- the luminance and RGB histograms displayed on the LCD screen are based not on the actual RAW data but rather on the in-camera JPEG which has been created by applying your white balance, saturation, sharpness, contrast, white balance, etc settings to the RAW data. Moreover, because a Bayer array has two green sensels for each red and blue sensel, a gain factor (whose value depends partly on the selected white balance) is applied to both the red and blue channels to "equalize" the channels. This means that the histograms will be inaccurate for any images that contain significant areas of bright red or blue because those values will have been "amped up" in the JPEG to the point where it appears they are being clipped whereas, in reality, they could get more exposure without any danger of blown highlights.
Since digital sensors capture linear data, half of the 4096 levels in a 12-bit capture are devoted to the brightest stop. Underexposing -- which is what the in-camera histograms frequently encourage us to do -- leads to throwing away a lot of the information that the sensor is actually capable of recording. This post to a UniWB thread at DPreview contains a lucid explanation as to why you might want to use UniWB:
...uniWB gives you a histogram that more accurately portrays the end points of the RAW data, allowing you to more accurately see when your image could take more exposure or when you've blown highlights. So, if you aren't having exposure issues today, why would you want to consider using uniWB? Besides being able to conclusively see if any color channel is really blown or not, what uniWB really allows you to do is to achieve the maximum exposure possible by increasing exposure until the histogram is maximally to the right without blowing any highlights. So, why would you want to do that. The answer is that this "expose-to-the-right" (ETTR for short) technique allows you to get the maximum dynamic range out of your sensor and your image and the lowest noise in the shadows. Anytime you don't use the full highlight range of the sensor, you are not taking advantage of the full range of the sensor and you then can't record as much dynamic range as if you had used the full range. So if you have a sunny shot with bright highlights and dark shadows, uniWB allows you to set the exposure for the most dynamic range possible without risk of blowing highlights. An inaccurate JPEG-derived histogram does not give you as accurate information and thus you might either not expose as much as you could or you might accidentally overexpose and blow some highlights.However if you're shooting landscapes in relatively even light, where large areas of the image are already bright green and there aren't extremes of contrast, then UniWB is not as useful because the JPEG green channel histogram is providing you with all the information you need to make an accurate exposure.
To set up UniWB on a Nikon camera, the gain of each color channel is set to a factor of 1 by creating a custom white balance based on a specially created image, all of the in-camera settings that are used to produce the in-camera JPEG are zeroed out, and a linear Tone Compensation curve is selected. Douglas's settings for the A900 would appear to emulate this procedure.