Evaluating White Balance and Tint for Leica SL and Leica M240 with DataColor SpyderCHECKR with 3 Lenses
This analysis should be of keen interest to any Leica SL shooter, or anyone considering the SL as a replacement for the M240. Or, simply to understand how the choice of camera profile drives very different settings for white balance and tint.
White balance and tint were giving me fits with field shots with the SL; the Leica SL delivered many greenish and bluish images in the mountains at its “Daylight” setting.. To process those images for good color rendering, it was necessary to establish a reference color temperature and tint for the Leica SL in sunlight. For comparisons, it was also necessary to do the same for the Leica M240, and to use several different lenses as cross checks.
Accordingly, the Datacolor SpyderCHECKR was used here in overhead mid-day solstice sunlight to to determine the appropriate white balance and tint settings for raw conversion in sunlight.
Three lenses on two cameras were used to establish whether the white balance and tint behavior is consistent. The Leica SL and Leica M240 were used as follows:
- The Leica 24-90mm f/2.8-4 Vario-Elmarit-SL ASPH, a native-mount lens with full electronic support on the Leica SL.
- The Leica 50mm f/2 APO-Summicron-M ASPH, Leica’s top-performing M lens, 6-bit coded and utilized with the Leica M-Adapter-T.
- The Zeiss ZM 35mm f/1.4 Distagon, an uncoded lens for which no lens corrections are applied by the camera, and a lens which I consider as good or better than any Leica M lens.
David W writes:
That was an interesting post regarding the white balance issues on the Leica SL. You may remember publishing an email I sent to you regarding the use of the A7RM2 in very cold conditions. At that time I mentioned difficulties with setting the white balance and I think that it was not seen as a problem. Well in the end after a lot of fiddling with exposure, temperature and tint I got to where I wanted to go but I still maintain that the “out-of-the-box” render had a blue cast that was difficult to correct. Following that exercise I have had a similar problem (as have many others) with a recent upgrade from CCD to CMOS on my Hasselblad. [DIGLLOYD: sometimes a channel blows out even though the histogram won’t show it; use RawDigger to check the actual raw data. I can’t rule out behavior in cold, but my cold weather experience has been fine].
I have come to the view that when a camera manufacturer is bringing a new model to market it arrives with an under-developed or thought-through firmware setup. More importantly the algorithm that presents the raw image to the raw converter seems to be a source of some of the problems. My understanding is that the sensor simply outputs data as a luminance measure per cell and that the embedded software interprets that according to the Bayer matrix and then the raw software converts to a visible image. However that takes no account of the way these sensor cells respond to light of different wavelengths until someone can take some measurements and establish a profile. Who decides what is good and what is bad? Is it colour fidelity or is it a match for the earlier version of sensor (CCD or otherwise). We then have the further interpretations of the raw converter but by that time we as photographers do have some control.
The chief complaint I hear from many people and one that I have experienced is simply that the output is different when compared with what they used to get and as a result new practices have to be learned to achieve the result envisaged - as you appear to have discovered with the SL.
What I do not understand is why the camera manufacturer cannot simply replicate the output qualities and colour values for the sensors predecessor as an optional profile? I may be completely naive here or have a fundamental misunderstanding of the technology but if so it is a position shared with many others who buy and then upgrade a camera because they like the way the images look in terms of colour, sharpness and so on.
BTW you have an opinion of merit on most issues and you are not bland or boring. It makes entertaining and informative reading.
DIGLLOYD: the choice of sensor definitely changes the look of the image. For starters, the CFA (color filter array) on the sensor may be different for each sensor, filtering light a bit differently for red, green, blue (and with overlap), and the sensor itself may respond a bit differently to the light it receives. There may be other spectral cutoffs as well suvh sd more or less IR blocking, more or less magenta/cyan overall tint; all of these things mean that the recorded data will vary by sensor. Add in changes to dynamic range and noise behavior and the output is just never going to match. A really superior profile could address some of this, but vendors never seem to bother.
In a similar vein, the difference in color between my workhorse NEC PA302W display and the NEC PA322UHD as for neutral gray is visibly different—even though calibration claims perfect neutrality for both (it has to do with the nature of the backlighting and non-continuous spectrum of that backlighting)—the PA302W rules the roost. Displays do not use continuous spectrum, and sensors use color filter arrays. No perfect match is going to happen. That is the source of the issue. Add in the raw converter, and all bets are off, except that companies like PhaseOne and others in the medium format area tend to try to minimize the differences between cameras.
Add in light with a strong color cast, and any filtration issues can be accentuated outside relatively narrow zone of “daylight”, which I would roughly label as 4600°K to 7000°K. Outside those bounds are warm sunsets and high mountain bluish light and any filtration subtleties start popping up like mushrooms in terms of sensor behavior. Not to forget the lens behavior either, which can attenuate some spectra and/or flare more. As a strong example for sensor non linearity, the Sigma DP Merrill has weirdly color reaction in bluish light, which is a challenge in processing raw.
As for the look of the SL sensor: based first and foremost on field shots but confirmed with the above color checker series, I rate its color and contrast as inferior to the M240 using equivalent processing settings (equivalent white balance and tint for neutrality as per above, but not adding contrast or saturation). The SL looks dull and flat by comparison to the M240, that is, with M lenses. Quite possibly the SL may be more accurate, and it definitely has better dynamic range and noise but I would say this: accurate is not the same as pleasing/appealing. The SL is less pleasing to my eyes than the M240 and the M240 in turn was less pleasing than the CCD-sensor M9.