Why very high white balance temperature for some cameras?


Starting from the camera setting for white balance, RT gives some files very high temps, 40000K or 50000K. The same file opened in other raw developers might come out at 8000K. Casual observation suggests this happens when the sky is blown, for example. A change from Camera WB to Auto will sometimes bring the temp to a reasonable value.

Why is RT so different with some cameras? I’ve seen this on Sony and on Leica SL. Very high temps are not meaningful to me and therefore difficult to adjust to taste.

[Question]Why very high white balance temperature for some cameras?
(Morgan Hardwood) #2

Please upload a sample file.


Here is a sample at “only” 30856K, tint 0.970. By comparison, in a trial version of Silkypix Developer 8, the camera setting is given as 4998K, tint 3 magenta. The shot has blown sky (deliberately allowed), but other shots with no blown pixels may still come out at tens of thousands degrees Kelvin. Sometimes RT Auto white balance computes a more typical value, sometimes it too says tens of thousands degrees.

leaves_temp_example.DNG (42.3 MB)

(Ilias Giarimis) #4

Usually RT shows abnormal WB values for raws of non fully supported cameras (missing colormatrix data) or cameras with different than the expected colormatrix2 at D65 illumination in DNG exif.

This last is what happens with this DNG sample … it reports a D50 colormatrix2 … RT supposes a D65 and the calculations fail. If we change the matrix to D65 (copy from Adobe dcp or just convert the DNG with DNG converter the values calculated are temp: 5222 tint: 1.030


Thanks for a clear, concise explanation and a good workaround!

Is it appropriate that RT check for the type of colormatrix and calculate accordingly? Should I post the suggestion somewhere? (No more question - I wrote up the issue.)


I got this feedback on a Leica forum: DNG’s typically (but not always) have two color matrixes, one at D50, and one at D65. The issue with RT is that RT assumes that typical == always. The correct way to read a DNG is to read the tags to find out what color temperature each matrix is, and process accordingly. (CalibrationIlluminant1 and CalibrationIlluminant2). That’s why other raw processors don’t have that issue.

(Morgan Hardwood) #7

DNG’s typically (but not always) have two color matrixes, one at D50, and one at D65

@RTCharles that part is incorrect. Most of the time the StdA illuminant is set as illuminant 1 and D65 as illuminant 2. While is it legal according to the DNG specification (v1.4.0.0, page 31) for D50 and D65 to be specified, I have never seen that, and it would not make much sense.

Agreed that a raw converter should check and not assume - I don’t currently know whether RT does that.