Rationale for Dynamic Range Table

[Jonathan Wienke]

[font color=\'#000000\']If we were digital bit limited wouldn't it make more sense to convert the linear measure into a log measure (ev units) and output brightness in constant ev increments?[/font]

[font color=\'#000000\']That's what every digital camera in the world does when creating and saving JPEGs.[/font]

[MMuddler]

[font color=\'#000000\']Just read a piece by thom and understand that Nikon's NEF compression takes advantage of the fact that we really don't need the hyperfine shading in the brightest tonal ranges. Indeed, he says our eyes work in a non-linear fashion with brightness. That fits with my hypothesis of where designers would want to go when faced with the observation about so much data range  being allocated to the top two f stops in the captured sensor range. Cool!

IMHO, the Luminous Landscape articles "Expose to the Right" and "Understanding Histograms" ought to be adjusted to explain the linear vs log brightness reflected in the digital data.

BTW in response to earlier post, it's my impression that jpeg compression is more related to interpretation of adjacent pixel data to be able to portray the image in fewer pixels rather than to compress brightness shades.[/font]

[Jonathan Wienke]

[font color=\'#000000\']Digital cameras throw away 1-2 stops of the sensor's dynamic range when generating in-camera JPEGs. Shooting RAW will let you use the sensor data that otherwise would be thrown away.[/font]

[MMuddler]

[font color=\'#000000\']If CMOS and CCD are linear translators of incident light within a range of capability and the A-D converter also has a dynamic range to cover the sensor outputs linearly then certainly we'll get output as shown in table. Last bit of 12 binary bits can be related to the range of the brightest f stop... and has  2 to the 12th  minus 2 to the 11th = 2048 discrete luminosity steps. Step down 5 f stops ( binary bits) and only have 7 bits to play with and 128 discrete brightness levels.  I doubt that we have the ability to see the 2048 steps in the brightest f stop range and wonder why we "waste" the resolution up there and pack so much info into that range. If we were digital bit limited wouldn't it make more sense to convert the linear measure into a log measure (ev units) and output brightness in constant ev increments?[/font]

[BJL]

[font color=\'#000000\']To expand on the last comment; common formats like JPEG use "gamma compression", which is roughly raising the luminosity value to the power 1/2.2. Not exactly logarithmic, and the precise details were chosen for obsolete reasons to do with people watching TV in darkened rooms (true!), but it seems to get the job done.[/font]

[Ray]

[font color=\'#000000\']Just read a piece by thom and understand that Nikon's NEF compression takes advantage of the fact that we really don't need the hyperfine shading in the brightest tonal ranges.[/font]

[font color=\'#000000\']This might have a lot to do with the nature of the subject. My own experience tells me, if shooting waterfalls, avoid JPEG like the plague. The same would apply to detail in clouds.

In general, low dynamic range subjects are okay for jpeg. For high dynamic range subjects, shoot RAW.[/font]