selmslie wrote:
The closest I could come with my D610 at ISO=100 and a full G=6 bias was at 4760K, red=1.2265625 and blue=1.1796875. At 4550K the red and blue coefficients are reversed implying that they would be close to the same value at 4655K if I could select it, which I can't. The ERADR I measured with this setup was about 1 full stop.
Maybe the problem is that you do not now how to properly measure ERADR. I suggest you review Dave's instructions and try again.
As I've previously demonstrated rather well, I can easily measure ERADR within less that 1/3 of an fstop for any RAW file.
Dave's measurement is not accurate.
selmslie wrote:
I tested the same scene using a daylight setting of 5880K and a neutral G/M bias and found an ERADR of over 1.3 by simply using the RGB histograms. This took a lot less effort.
A different scene or a different ISO might have yielded different ERADR values but there is no reason to expect that UniWB would turn the histogram of the JPEG into a representation of the full raw range.
You aren't yet understanding what UniWB does. When the multipliers are 1.000 for both red and blue, the JPEG histogram will look exactly like the RAW histogram.
selmslie wrote:
Based on everything I have seen so far, UniWB appears to have been a brief episode of lunacy in the evolution of digital imaging - long on promise but short on delivery. RIP.
It has no benefit for anyone that cannot understand how it works. And it is rather technical, so for many people it can't help. Even for those who understand and can use it properly, the major detraction of not being able to show a preview on the camera, because of the sickly green cast, makes it less than useful.
And last but not least, just how often does anyone need that kind of exposure accuracy?
selmslie wrote:
I have no way of setting or measuring exposure compensation in the camera in 1/10 stop increments and neither do you (±5 EV in increments of 1/3, 1/2 or 1 EV according to the specs) so the precision you claim sounds like puffing.
Still don't understand the technical in and out, eh?
Whether you can set EC within 1/3 of a stop isn't important. But that does mean you can get exposure to within 1/6th of a stop, but that is precision rather than accuracy. Given a histogram display that is accurate within 1/10 of a stop, that 1/6 is accurate too.
Guessing at what ERADR (or assuming it is fixed) puts your exposure accuracy at maybe 1/2 or 2/3 of a stop.