> If you were look at the same pixel in a 16bit TIFF file rendering of > the 12bit RAW data, the relative values of the channel assignments > would be the same, but you have 16x as many numbers available to > describe the values in the original RAW file which is then scaled to > a representation in a discrete numeric space 8x larger (0-4095 in the > RAW data, 0-32767 in the 16bit RGB channel (the topmost bit is not > used so it's actually 15 bits of data)). Only 4096 of the values in > the 16bit space are actual photosite RAW values, they're fitted into > the larger space because current computing machinery manages 16 bit > numbers with greater efficiency than 12 bit numbers, in general, > *and* because as you perform Real or Discrete valued operations on > these numbers, there are more numbers to represent the results, thus > greater precision and less likelihood of clipping or round off errors.
While almost everything has been said on the subject, I would like to mention one important issue, so that a possible confusion is avoided. The text above, while being correct, can cause a confusion. It might seem that since 16-bit TIFF has more bits to operate, it is a "reacher" format compared to RAW, so no information is lost in the conversion RAW->TIFF. (I've seen it people thinking this way, and was under a similar impression in the very beginning of understanding RAW vs TIFF differences). One has to remember that information in RAW is stored differently. The jpeg/tiff you get from DS or after conversion in PS is 3008x2008. This size corresponds to the 6MP size of the sensor, i.e. to ~ 3008x2008 sensor sites. (I leave aside the question of a few "extra" pixels which are not being used in the conversion in PS, but can be used by some other software). However, in the JPEG/TIFF, - each pixel has information about 3 colors (RGB), while in RAW - you have a "mosaic": out of each 4 sensor sites 2 have information about G(reen), and 1 - R(ed), and 1 - B(lue): RGRG GBGB RGRG GBGB (4x4 part of the sensor ) Converter from RAW does what is called demosaicing, by mixing values from the mosaic to assign values to the pixel in between. While the description below might not be exact, I think it is presents a picture that is easy to understand. If you were to put the mosaic described above on a square-ruled paper (matrix), then we can put "pixels" P on the nodes (corners of the squares): R_G_R_G _P_P_P_ G_B_G_B _P_P_P_ R_G_R_G _P_P_P_ G_B_G_B In the demosaicing process, essentiall the value of each pixel "P" is assigned based on the values of the four surrounding sensor sites. While in a small image above it looks like you have fewer "P"'s than R's, G's, and B's altogether, but when you have a large matrix, the difference in number is negligible (in the scenario described here it would be ~ 5000 pixels for a ~2000x3000 matrix which is ~0.1%, but in reality it could be a bit more (no pun intended), depending on the implementation) It is this demosaicing process when the white balance, sharpness, gamma correction, ISO values, etc are being used in the conversion. Since the process of demosaicing is fully mathematically reversible, the loss of the initial information happens at this point. That's why you have better control over the image when it is in RAW format even compared to when it is in 16-bit TIFF. I think this rather simple documents explains some of these issues in more detail: http://www.adobe.com/digitalimag/pdfs/understanding_digitalrawcapture.pdf I hope I didn't cause too much confusion, and my explanation was helpful to somebody. Igor -- PDML Pentax-Discuss Mail List [email protected] http://pdml.net/mailman/listinfo/pdml_pdml.net
