GR User Forum

The spot for all Ricoh GR camera users

Register for free, meet other Ricoh GR users, share your images, help others, have fun!

Tell your friends about us!

Incorrect use of LinearResponseLimit tag in DNG files

simon_t

New Member
The DNG files produced by the GRDII (and GX100) for that matter contain 12-bit data (numbers between 0 and 4095). This data is mostly linear, but less so in the highlights, which can lead to a magenta cast in extreme highlights.

The DNG standard allows manufacturers to indicate the fraction of the range in which the response can be considered linear using the LinearResponseLimit tag. In Ricoh's cameras, this is set to the default value of 1.0 (fully linear response). A value like 0.9 (exact value to be determined by Ricoh engineers) is more appropriate and really cleans up the highlights.

For testing purposes, I have used ExifTool (http://www.sno.phy.queensu.ca/~phil/exiftool/) to change this tag on the overexposed doorway shot that I downloaded from this site (thank you for making the raw files available!). For the results, and a more detailed explanation, see http://forum.getdpi.com/forum/showthread.php?t=757.

I hope that this will be fixed in a future firmware version. It should be an easy fix.

Cheers,
Simon
 
Hi Simon and welcome here! A very interesting find! I will try to pass this info to Ricoh developers. This magenta cast is definitely not nice behavior. It would be great to see such simple fix in Exif. Thanks for report!
 
Just one question. What RAW developer do you use? I just tried dcraw, and its highlight recovery feature gave me exactly the same results with LinearResponseLimit set to 1.0, 0.9, or even 0.4? So the use of this tag must be RAW tool dependent? The dcraw highlight recovery command used in my test was H 2 (2=blend).

Here is my comparison:


As you can see, the amount of recovered highlights is more or less the same as in your case, but with very low magenta cast and with significantly brighter shadows? And Silkypix highlight recovery seems provide no magenta cast even with LinearResponseLimit = 1.0 (see this post).
 

Attachments

  • highlights.jpg
    EXIF
    highlights.jpg
    538.1 KB · Views: 1,024
I'm sorry, I forgot to specify that I used Adobe Lightroom (=ACR). It makes sense that the tag works in Lightroom, because Adobe modeled the DNG standard on the way ACR works internally. So it could well be that (most) other applications don't support this tag.

An alternative approach to fix the highlights would be to change the WhiteLevel tag from 4095 to something like 3700, essentially discarding the data from the brightest regions. Unfortunately, ExifTool does not allow you to write the WhiteLevel tag, but it should be possible to modify it to make it work. Also, it's less elegant than using the LinearResponseLimit tag - if that tag is supported, that is.

If Silkypix does not produce a magenta cast, it probably means that it uses some extra data that wasn't included in the DNG file (a hard-coded LinearResponseLimit, so to say).

Simon

PS - The fact that my shadows are darker is because I simply used the exposure slider (set to -4) instead of the recovery slider.
 
Back
Top