Photoclubalpha

Iliah Borg ( http://www.libraw.org/" onclick="window.open(this.href);return false;) discusses UniWB in RAW on the A700/900 on DPR forum http://forums.dpreview.com/forums/read. ... e=30781366" onclick="window.open(this.href);return false; with some of his input quoted below.

As I understand it this method shows us the true histograms in RAW on the camera LCD so as to allow you to monitor no channel is blown out, the consistent use of filters and to ETTR (expose to the right) with confidence. The default histogram is based on JPEG.

To achieve this set the camera to manual,100 / 200 ISO, take a fully blown out image, say max. lens aperture at 30secs, the camera will give an error warning, but save as a custom WB setting, you cannot do this by setting WB in menu. Exif tool should show (256,256,256,256) for RGGB), all sensor elements are saturated, so turning camera into accurate colour meter and not changing recorded data. You will end up with green tinted LCD image but accurate exposures in RAW.

"White balance is multiplication of linear raw data to equalize sensitivity of sensels of different colour groups to render greys as greys.
UniWB is setting WB multipliers to all 1's and is meant to be a workaround for 2 major problems:
1. Histogram distortion with white balance application.
Cameras show a histogram based on JPEG preview, while JPEGs have white balance applied to raw data. If you want to use histogram to judge exposure, this white balance application prevents from the proper evaluation of the real exposure problems.
2. Many raw processors have problems scaling data due to incremental white balance application. Say, white balance is applied causing overflow in a channel. When we start adjusting white balance the lost information in the highlights can't be recovered as it is lost during white balance application at the first stage. Same goes to applying negative "exposure compensation" in a converter to a channel that is already blown out due to white balance (sequence of operations error, negative "eV" in a converter should be applied before white balance, but to speed things up and to use less memory it is not always the case, same as incremental application of white balance saves memory but causes irrecoverable overflows)."

"Each "colour" channel of the sensor has its own sensitivity. For A900 those are 3, R, G, and B (because G1 and G2 are essentially similar - not always the case with other cameras). You can think of those as having different base ISO. That is, R has ISOr, G is having ISOg, B - ISOb. The "normalised" ISO the camera reports and measures for is ISOg. Under normal daylight ISOb and ISOr are lower than ISOg. Moreover, for most of the light types most of the cameras have ISOg being the highest one. What white balance is doing it is digitally bumping (through multiplication) the data on R and B channels to reach the sensitivity of the green channel. "
Stephen

Sonolta wrote:Some of us have been following uinWB techniques for years...I don't use uniWB as I find it to be a bit of a pain in the butt for my style of shooting.

http://www.photoclubalpha.com/forum/vie ... iwb#p10108" onclick="window.open(this.href);return false;

The original work referenced in that link has been moved/removed...You will need to tun it through a translator as it is now part of a tut.

http://www.guillermoluijk.com/tutorial/uniwb/index.htm" onclick="window.open(this.href);return false;

-Sonolta

Thanks for pointing to previous posts, but if UniWB gives you an accurate histogram in RAW and you ETTR I thought it would be an automatic choice, so what are the drawbacks that it is " bit of a pain in the butt for my style of shooting"
Stephen

snn wrote: ...
As I understand it this method shows us the true histograms in RAW
...

So many people had been killed in the name of true -any true...
I think it is a good technique and, as I do with other techniques, I apply it if needed. But, as Don said, is a pain to use it as standard.

Regards

As far as I can tell, uniWB has no effect on the raw file except (possibly) some adjustment may happen to overall exposure - what it does to make the sensor into an equi-energy colorimeter with the RGB histogram as a simple readout, so it really shows you how much red, blue and green level is being recorded by the sensor. It does this using the JPEG or embedded JPEG, because there is no other way of seeing the histogram in-camera without relying on a converted version of the raw file. But by using uniWB, you ensure that the histogram matches the actual RGB sensitivity of the sensor itself, before any JPEG conversions are applied.

WB settings appear to have no effect on actual exposure - if you meter a scene as 1/60th at f/8 using daylight WB, it stays as 1/60th at f/8 using tungsten, fluorescent, auto, or uniWB or anything else. The raw file data itself remains identical too. But with uniWB, the JPEG may appear differently exposed on the review screen, as well as being different in colour, and the histogram will always look quite different. So if you then adjust your exposure manually to make the uniWB previewed result avoid histogram clipping but be the maximum (to the right) you get what should be the best possible raw file to work on - one in which no channels are clipped, and the highest channel determines exposure (instead of the green channel).

I reckon that bracketing your shot +/- 1 stop in most conditions is a quicker way to ensure you get one raw file which matches this standard. Certainly, I would not want to lose a usable JPEG preview, and uniWB guarantees you DO lose your usable preview.

David

I think you're right on, David. UniWB has done a lot for me in showing how much more I can ETTR than I previously had thought (resulting in lower noise,) because, in nearly every light other than tungsten, the green channel blows much sooner than the red/blue. I think Sonolta is correct that it can be an annoyance, and I don't always use it. Sometimes, if my light is remaining consistent, I'll meter with UniWB, and then change back to regular WB methods...or I'll just use the green histogram channel with regular WB and that is pretty close. All of this being said, in-camera jpeg settings can be just as crucial as uniWB for nailing perfect exposure (when shooting RAW,) because your camera's LCD histogram reflects those settings. For the A900, these settings would be somewhat like this:
---------------------------
Neutral Creative Style
-1 saturation
-3 brightness
-1 ZONE
Adobe RGB color space
---------------------------

If you use the above settings, I believe that's about as close as you're gonna get to the actual RAW histogram. One important thing to remember is to change the initial curve of your RAW converter** to linear when you import with the above settings, or your shots will look blown out when you initially look at them in your converter. Don't worry, it's just the converter adding special sauce like a tone curve and brightness, which needs to be adjusted.


**I would imagine that this isnt a good technique with IDC, because it uses the jpeg metatags.