Lloyd Chambers has just added additional coverage on the RX1 including the color shift issue. Of greater significance to me was his section stating that the color bit depth was only 8 bits for the RX1. I have sent him the following inquiry and will post his answer when I receive it:
I just read your section on files sizes and bit depth on the RX1 where you state that the RX1 only uses 8 bits per byte; however, the Sony web site specification states:
"A whole new world of high-quality images are realized through the 24.3 MP effective 35 mm full-frame sensor, a normal sensor range of ISO 100 – 25600, and a sophisticated balance of high resolving power, gradation and low noise. The BIONZ® image processor enables up to 5 fps high-speed continuous shooting and 14-bit RAW image data recording."
Since it's ARW files are essentially the same as the ones produced by the A99 (just changing the EXIF descriptor allowed us to process the RX1 RAW files in LR as A99 files before RX1 support was released) it would appear both RAW files use the same compression scheme. Whether that scheme is losss-less or not really depends on the schema used. However, to capture and process 14 bit data, the internal pipeline must support 14 bits (probably 16 bits since that is easier to manufacture and work with in a binary world). Therefore, I don't know where you are deriving the 8 bit per byte data unless you are assuming that based on the file size. I don't think Sony would claim 14 bit raw files if it was only producing 8 bit raw files. Can you please explain where your information came from?
A check on the A99 on the sony web site states:
"Model Features: 24.3MP 35mm full-frame sensor, 14-bit RAW output, uncompressed Full 1080 HDMI® output in 60p/60i, up to 25600 ISO, Dual AF, continuous AF in all modes, up to 10 fps in Tele-zoom mode, 3" tiltable LCD"
We know from earlier experience that we could process the RX1 files as A99 files so they most likely use the same data storage method. Sony further claims to produce uncompressed files. Of course, this raises the question of how do you get 6,000 x 4000 x14 divided by 8 into a 24.3MB file without some type of compression. That should be a 42MB file uncompressed. There are lossless compression techniques that might do it.
I have now received a response back from Lloyd:
"Fact: the file size is always 8 bits per pixel.
Fact: the file size does not vary.
Fact: by the above, the camera thus stores 8-bit per pixel files.
Fact: to guarantee 8 bits per pixel, lossy compression must be employed.
Fact: I stated clearly that internal processing is certainly at a higher bit depth (apparently 14, but this does not work out properly according to RawDigger, which shows 12 bits when delinearized).
How can this be? Only by a scheme that limits storage to 8 bits per pixel by lossy compression. In Sony’s case, by a tone curve with a “sag” in the middle to high key area that effectively compresses the 14 bits down to 8 bits."
Followed by another email:
"This Sony claim is a fraudulent claim ("14 bit data recording"), just as Sony posts fraudulent MTF charts that are not possible by physics.
The files are by simple math, 8 bits per pixel. RawDigger developer confirms that they are 8 bits. This is a fact and not open to argument.
The files record a range wider than 8 bits by a tone curve that maps those 8 bits to a wider range. This is re-linearizing the data.
For example, 8 bits could record a range of 32 bits also. The question is how big the steps are within that range."
I'm not sure what to believe. It doesn't make sense to me that Sony would go to the expense to have a 14 bit processing pipeline only to throw it all away when recording the output. I experimented with converting to DNG and Tif, the ARW files. The DNG files varied in size from 26mb to 29.5mb and the tif files were huge at 144MB. Can anyone more knowledgable about this please provide some insight?