I read an article about UniWB: http://www.guillermoluijk.com/tutorial/uniwb/index_en.htm
and played around a bit with my SX230.
The problem RAW shooters are facing is that the histogram of the camera shows only the information about the resulting JPEG (when the RAW file is already developed in the cam), but the information about the RAW image is not displayed at all.
Unfortunately white balancing is the morst critical step during developing the image from RAW to JPEG since it influences the histogram very much. White Balancing is nothing than amplifying RGB channels in a certain way. E.g. these are the factors for white balancing daylight of a 350D:
Daylight: multipliers 2.132483 1.000000 1.480864
A factor of 2 is round about -1EV which means that the raw sensor could have much more exposed than the resulting JPEG.
First I tried to get multipliers of "1 1 1" for "R G B" by covering the lense and performed a manual white balance. I have taken any picture and double checked the RGB multipliers by dcraw:http://www.heliconsoft.com/dcraw.html
After several tries I got following deviation: 2 % in G and exactly 0% in R and B. Most of the tries it was around 10%, which is also pretty fair.
After that I mounted the cam on my tripod and double checked how far I can get:
1. Taken an image with UniWB (without overexposing the image)
2. Taken the same image (with same exposure as 1.) with AWB (I got a lot of blown fields now from chdks zebra-funcionality (dont know the english word, sorry)
3. Taken the same image with AWB, but decreased the exposure to get no blown zebra fields (it was -2/3 EV)
After that I checked the images in lightroom and voila: the second image looked pretty good with respect to the RAW histogram. Image 1 looked also very good when the white balance was corrected. In fact there was almost no difference between 1. and 2. image.
The 3. images histogram had a lot of room for improvement in the first decade of the historgram (the first decade covers the most brightness steps of the entire histogram).
Conclusion: It is possible to transport the histogram information of the underlying RAW to JPEGs histrogram by using UniWB. By doing so you know the limitations of the sensor much better and you are able to expose more to the right resulting in improved image quality.
Often CHDK-historgram differs a lot from Canons histogram when using UniWB - I dont know why. Furthermore the zebra functionalty dont works correcly when using UniWB. Therefore I have to double check also Canons histogram and Canons Zebra to find overexposure.
Furthermore I had several manual WB tries to get reasonable multipliers.
Where CHDK can help:
Normally you can take the image with UniWB and correct the White Balance afterwards in Lightroom. This works pretty well. Nevertheless the JPEG file is totally useless. Furthermore I would like to have Canons AWB suggestion stored in the DNG to compare it against Lightroom suggestion. Therefore I prefer Image number 2.
To get this measuring more convenient with CHDK I need to modify the White Balance to manual White Balance and back to AWB in an easy way. I was wondering if it is possible to use a key (lets say display key) to do the UniWB measuring and adjustments (in fact it is the same measuring as I would do by half pressing the shutter but only with manual White Balance) and use normal shutter to take the image.