Hi,
Interesting discussion here, what's your opinion of pro photo RGB Pxl8?
Over the summer I had a client that specifically asked for 16 bit, Adobe RGB in camera colour space, working profile pro photo RGB, I submitted the images as a 16 bit TIFF in pro photo RGB which they later coverted to CMYK for press publishing.
T
In camera colour space is pointless if you're shooting raw, the output from Lightroom, etc. is what matters which in your example ProPhoto.
If your outputting to a high end system for press, etc. then using a larger colour space makes perfect sense and the system can handle the colour depth and range. That's where the larger colour spaces really come into their own and has lead to the idea that they "give you more colours and are better" as, after all, more is better isn't it?
In an 8bit bandwidth a red value of 255 is full saturation of red. Now if you compare red values between sRGB and a larger space such as AdobeRGB or ProPhoto to get the same saturation you need a much lower value, maybe 230 and 200 respectively. Now if the saturation in the image only needs the 255 level of sRGB then using a larger colour space means you're throwing away data. The larger space doesn't have values above 230 or 200 but still needs to represent all the levels up to that point so instead of 0-255 you've only got 0-230 or 0-200 and that can produce banding in gradients - you've got fewer values to represent the transition from zero saturation to full saturation.
Note this is exactly why images posted online in AdobeRGB look dull and lose saturation - a given RGB value represents less saturation which would normally be converted by the colour management system to the correct value for your display profile.
So what happens when the image does need more saturation that the large space offers? 255 in AdobeRGB might need a value of around 280 in sRGB to display the right amount of saturation but of course in 8bit you can't go that high so instead all the values are clipped to 255 (actually it depends on what mode you set the colour management to but for simplicity I'll stick with clipping). So now you're gradient in sRGB loses some tonality at the top end but still has the smooth gradient to that point. In AdobeRGB you have an evenly space gradient but the banding is still there - remember AdobeRGB is going further in the saturation but still only has 256 steps to get there.
The working space should always be the smallest one you can get away with for the image data. If you put the image into a larger space you don't get more colours, you get the same colours but with less tonality because the working space has to save headroom for more saturation. It's like having a room 10ft tall, you might get a few people who are approaching 7ft in there but you've still got 3ft of space that's never going to be used. Better to have a room only 6.5ft tall and have a few people duck (clipped colours). In some situations the clipped colours might be vital and a larger space is needed but when that happens you really need to think about using 16bit colour with 16bit output devices to prevent the banding issues, something your client was able to deal with. For most of us, printing on inkjets or using photobox, etc. we're stuck at 8bit and images that don't need the larger space anyway so all we're doing is trading tonality for satuation values that aren't being used.
I shoot in Adobe 1998, process in Adobe 1998 / 16bit, then convert to custom profiles for printing. I don't see why you would choose to throw away information from the beginning, never know when you may want it.
See above, having the data in 16bit doesn't solve the wasted space issue, it just means you can measure it more accurately.