Originally posted by owl I haven't looked into this specific question regarding Pentax but I am an engineer who designs video cameras and here is how you can get there in my experience:
The "RAW" data is the computational result of the camera's image processing algorithms. These processes might include median noise filtering, applying gain and offset correction (for example to reduce vignetting and reduce speckle) and white balance scaling, gamma correction and the like. When you run these calculations, you end up with more bits than you started with, 14 sounds about right. It would be silly to throw away these additional bits since they contain useful information.
Good question.
These are differnt kind of bits
These are bits that correspond with colour.
It's like you can have a 8bit TIFF and a 16bit TIFF, it has nothing to do with processing or extra information squeeze in, its just colour.
Bit Depth Quote: Digital cameras actually record each color channel with more precision than the 8-bits (256 levels) per channel used for JPEG images (see "Understanding Bit Depth"). Most current cameras capture each color with 12-bits of precision (212 = 4096 levels) per color channel, providing several times more levels than could be achieved by using an in-camera JPEG. Higher bit depth decreases the susceptibility to posterization, and increases your flexibility when choosing a color space and in post-processing.
Normally 12bit is more then enough but because of the huge dynamic range of the K5 at lower ISO it goes past 12bit of information.