Green tint in shadows (with DP Merrills) since 0.57

[geni1105]

0.56 works fine, but with 0.57 shadows have a strong green tint with DP Merrill X3Fs. The red
channel is underexposed according to the raw histogram in FastRawViewer (0.57 vs. 0.56):

This does not seem to happen with DP Quattros (I just tried with one DP2Q raw file).

BTW, in x3f_io.h the camera ids are the same for DP2M and DP3M - this might be a bug, but the values do not seem to be used anywhere and it is probably not related to the issue, which occurs both for DP2M and DP3M:
#define X3F_CAMERAID_DP2M (uint32_t)78
#define X3F_CAMERAID_DP3M (uint32_t)78

[rolkar-kivra]

I have been looking at the commit comments for changes included in 0.57. Except for Quattro things, we also fixed a bug regarding dark point detection. There are four rectangles around the image where the sensor is covered with an opaque layer. There you can test for black level.

But, the Merrill sensors are so bad and have a huge spread in manufacturing tolerances, that sometimes those areas are damaged. We assume that, in order to get high yield, Sigma have done something clever in their software for compensate for broken black areas, instead of ditching those sensors.

Earlier we just took the mean value of the dark areas. But now we also try to be clever. Which seems to work badly for your camera.

We have to admit though that we have, more or less, given up on Merrill sensors. Always getting problems. So ... we might have made a mistake.

And .... unfortunately .... we have no time allocated for working with x3f :(

The main reason being that Sigma has made it unnecessary cumbersome to work with their messy files and we do not use Sigma cameras any more.

[geni1105]

The issue actually appears on both of my cameras, a DP2M and a DP3M - so it may be systematic and not related to a specific device. Anyway, thanks for the explanation and no problem, 0.56 works fine for Merrills.