colour depth: 3x 22 bits / 3 x 12 bits

I was reading the technical information of my K10D again. About the colour depth I read it is 3x 22 bits internal, and 3 x 12 bits output (in Raw).
What does this mean? Why is there a difference between the internal colour depth and the final output? Why don't we get Raw pics with a 22 bits colour depth, or at least with 16 bits colour depth?

Several issues need to be considered here. The chip used in the K10D as the analogue to digital interface between the image sensor and the digital image processor is a very specific ADC with digital signal preconditioning. It samples the analogue output of the CCD with 22bit precision then applies mathematical transforms to the data in order to optimize the RAW output.

The output of the ADC/pre-processor chip is available in differet hardware versions at 16, 14 and 12bits/pixel. Pentax chose the 12bit/pixel option for the K10D which it appears is a very good match for the dynamic capture range of the sensor.

The 12bit/pixel RAW output produces very good quality de-mosaiced 16bit composite RGB images. There is no certain advantage to using a RAW file with greater bit depth in the case of the K10D because most of the bits of least significance would likely only contain noise due to the noise to saturation limitations of the sensor.

More information on the theory and operation of the Nucore NDX-2240 used in K10D can be found in the following PDF (~6.5MB):

http://www.nucoretech.com/nu3/images/80_downloads/pres_technology.ppt.us.pdf

Cheers,

---

Rob Studdert
HURSTVILLE AUSTRALIA
http://home.swiftdsl.com.au/~distudio/publications/
Pentax user since 1986, PDMLer since 1998

Ok,

I 'm beginning to understand it. Thanks!

Myrdinn

But if you check the Exif info for the K10D you'll find its 16bit

---

Old wood best to burn, old wine to drink, old friends to trust, and old authors to read.

The value reported depends on the application reading the EXIF data I guess:

Extract of a K10D Photo interrogated via PhotoME:

[Image]
Image width: 3936 px
Image height: 2624 px
Number of bits per component: 12
Compression scheme: Pentax PEF Compressed
Pixel scheme: CFA (Color Filter Matrix)
Manufacturer: PENTAX Corporation
Image input equipment model: PENTAX K10D
Image data location: 0x00014D0C
Orientation of image: 90° CW (left/bottom)

The same file viewed via ThumbsPlus:

3892 x 2612 36-bit color (uses PolyImage Pro
Library for RAW decoding)

Which application/s reported that the K10D RAW files were 16bit/pixel?

Cheers,

---

Rob Studdert
HURSTVILLE AUSTRALIA
http://home.swiftdsl.com.au/~distudio/publications/
Pentax user since 1986, PDMLer since 1998

Adobe bridge CS3 reports 16bit

I also found it in the manual (somewhere) and on the Pentax web site specs.

Try this thread - it goes into good detail on bit depth
https://www.pentaxuser.com/forum/viewtopic.php?t=5036&postdays=0&postord...

---

Old wood best to burn, old wine to drink, old friends to trust, and old authors to read.

Thanks for the link Rob I was hoping that the 22-bit output was available, and that it didn't have a gain-control. It would've been great to get a RAW output that encompassed the complete ISO range in one hit (great for HDR stuff). Hey ho!

Rodger - are you talking about the "Workflow Options" in the RAW conversion dialog box? It defaults to 16 Bits/Channel (and the only other option is 8-bits). This is only an output bit depth for the conversion from RAW to PSD.

Photoshop, AFAIK, doesn't have a straight EXIF display (because once you've run it through the RAW converter it'll show the setting you set in the RAW converter, and not the original EXIF data of the PEF file).

FWIW, Adobe Bridge shows "16" as the bit depth for the *ist-D, *ist-Ds and K10D. I can tell you 100% that the *ist-Ds has only 12-bit output (it stores the data as a byte+nibble per sensor site). The *ist-D output a RAW format that was actually 16-bits per site, but on examination the lower 4-bits were always zeroes - effectively wasting 25% space (that's why the *ist-Ds RAW files are correspondingly smaller than the *ist-D - even though both use the same Sony CCD).

Bottom line - the Pentax RAW files are definitely 12-bit.

Hope that helps!
Matt

---

http://www.mattmatic.co.uk
(For gallery, tips and links)

As per Matts posting in the link above the conflict that you perceive is likely due to the confusion between RAW and RGB composite data depths.

RAW data is captured using a Bayer R-G x G-B matrix of sensors (call them pixels or sensels or whatever), each pixel is representative of only one colour so a 6MP sensor would have 3M green pixels and 1.5M each of blue and red pixels. Each of these pixels are recorded using a 12 bit encoder so the RAW file is 12bits/pixel.

The RAW file has to be processed by a complex de-mosaicking algorithm in order to synthesize the missing colour channel information for each RGB output pixel.

An 8bit RGB file contains 3 channels of 8bits or 24 bits/pixel, a 16bit RGB file contains 48bits/pixel so you can see that in either case their relationship to the source RAW file of 12bit/pixel isn't quite so direct.

Cheers,

---

Rob Studdert
HURSTVILLE AUSTRALIA
http://home.swiftdsl.com.au/~distudio/publications/
Pentax user since 1986, PDMLer since 1998