Green Shadows firmware fix?

[indergaard]

Advertisement (gone after registration)

There is always a A/D conversion in-camera. This is required for generating a digital raw file based on the analog information from the sensor. In this A/D conversion stage there is always processing applied. The A/D conversion stage is a processing stage in itself.

 

Without proper knowledge about the M240, I would maybe make a guess that the green shadow problem is further significantly amplified above 1600 ISO probably because that's when the M240 starts using digital amplification/push, and up to 1600 ISO the amplification/push is done analog? Just a guess/assumption, since I don't really know when the M240 shifts from analog gain to digital gain. Many other cameras usually shift from analog to digital around 1600 ISO, and when that shift occurs, it is usually a benefit to shoot at the last analog stage and rather push the files in post when you have more control over what details are pushed, instead of increasing the ISO value in-camera and thus getting a linear flat push to the entire image by the in-camera processing.

[jaapv]

Nice read;

"The M240 exhibits nonlinear response at ISOs 200, 400, and 800, artificially depressing darker tones. The darker the tone, the greater the depression. In my previous noise floor tests, I saw the noise floor to be artificially depressed at these ISOs."

and this points to the artificial blacks in the out-of-camera JPGs: shadows are deepened excessively up to the point that the output is not appreciated any more by me. For instance, at 2500-3200 asa at 3200 K walls under eyes turn up darker. 

Was this done to remedy a color cast? Comparing with the DNG's the latter turn out excessively flat sometimes under similar conditions.

 

When I compare the output in adverse situations (more than 800 asa) then my wife's X100 files look clean/linear. 

It was probably done to reduce shadow noise.  It is quite possible that you like the files of the X100 better at high ISO. It all depends whether you like Fuji's in-camera processing as compared to your own efforts from the M240 DNGs (or Leica's JPGs)

[adan]

1. I've only used the M240 a couple of times. I did not notice any green shadows, particularly.

 

2. However, most raw processing software (e.g. ACR) has, in the calibration panel, a slider specifically for shadow tint, which slides left for more green, and right for less green (more magenta). In my M9 personal profiles, as it happens, I've set that about 5 points on the green side. but anyone who wants something different can set it to the magenta side. Which leads to....

 

3. I've never used the canned profiles provided by Adobe, Leica, Sony, Canon or any other manufacturer. The very first thing I do (shots 001, 002, 003 etc.) with any new digital camera is shoot a Macbeth color card and build my own complete profiles (I do separate ones for sunlight and tungsten). They are always substantially different from the defaults.

 

Sony R-1: "Colors way too saturated, skies too purple" - they weren't once I did a correct profile

M9: "Ohh, the skin tones are much too pink." - they weren't once I did a correct profile.

 

My general impression - from the very first raw shots I ever made - is that software engineers know about as much about good color as the average artist knows about writing elegant computer code. I.E., not much. But fortunately we do have the capability to do our own profiles, save them so they are applied to every shot automatically, and thus correct the engineer's goofs seamlessly.

 

It's just one of those "if you want it done right, do it yourself" things.....

[jaapv]

Couldn't agree more. But goofs - maybe. Taste in colour rendering is such a personal (and cultural!) thing that it is hard to discuss it objectively. I would say that the differences between digital cameras as smaller than those between colour slide films were - and are.

[Guest JonathanP]

My only issue with using profiles to correct this is that because the problem exists before the demosaicing stage, the difference between the RGB levels is light level dependant once you hit the curves shown above, i.e. the ratio of R to G to B varies as the light level varies over the lowest 4-5 stops of the cameras dynamic range. So you'd need lots of profiles to cover all the possible light levels (and/or a more difficult to produce 3D profile).

 

If the roll-off was deliberately introduced to reduce shadow noise, then the big goof was to do it before demosaic - if it had been done after demosaic for the JPEG output (and the raw data left alone) then the JPEGs could have been "improved" without this side effect.

[jaapv]

According to Leica, they want to reduce noise that cannot be reduced as effectively in postprocessing. FWIIW.

[Guest JonathanP]

Advertisement (gone after registration)

Thats interesting, thanks Jaap. It's what I suspected, although I don't agree with the implementation 

[Lenshacker]

Most cameras (i.e. Canon, Nikon) don't touch the data beyond averaging out faulty pixels.

Most Nikon cameras store the .NEF file using a "Visually lossless" (ie lossy) compression algorithm using a fixed Huffman coefficient table. The high-end Nikon cameras allow lossless .NEF, but it is still processed data. That's what all those on-chip processing engines do. Nikon uses Motorola format big-endian data.  Noise reduction is done on chip. I suspect Canon is the same, never owned one. Most manufacturers seem to use TIFF format and have private tags and data for the proprietary information. DNG is TIFF with standardized extended Tags. Nikon started out using ".TIF" files for raw files for their first-generation DSLRs. The D1 switched to NEF, and on that camera performed white balance on the raw data before storing it in the file.

 

If you get the calibration pixels in the file, it's as close to Raw as you will find. The M8 in service mode does this. The M9 and M Monochrom in service mode still store DNG files, but larger images and bad pixel correction is not performed. Otherwise, just rename the ".RAW" to ".DNG" and open them in Lightroom.

[david strachan]

Maybe it is because I shoot only Raw and use a profile for ACR but I don't see this issue. I am not saying it should not be corrected by Leica at some point (simply because it should not be that way for either RAW or JPG) just that I don't see it. I am attaching a file shot at iso 1600 -- not for its artistic merit but because it was fairly easy to get rid of any green in ACR and CC. (And I am not saying this is perfect WB either! - just no green) So if your question is solely related to out of the camera files, one answer is in the workflow. Again, I am not saying Leica shouldn't correct any greens in the shadows but for your work there are ways to deal with it.

Hi Alan

A very radioactive face, or extremely jaundiced.

I'd much rather the green shadows.

 

Honestly, that orange face just looks bizarre, no matter how you cut it.

 

cheers  Dave S

[AlanJW]

Hi Alan

A very radioactive face, or extremely jaundiced.

I'd much rather the green shadows.

 

Honestly, that orange face just looks bizarre, no matter how you cut it.

 

cheers  Dave S

Yes, David.  I never said I had nailed WB there.  I looked for high iso files I had because I just didn't remember ever seeing green shadows. It was just godawful light and 1600 iso.   Maybe it was because I didn't care about recovering the shadows, as blocked shadows there were not an issue.  If the shot were good enough to spend more time on I would go back and fix the flesh tones.  But this post and my computer were the only places it has ever been.

[Guest JonathanP]

Just quick heads-up that I've released an updated version of the Lightroom plugin - better at higher ISOs - details in the post processing section: http://www.l-camera-forum.com/topic/229863-correcting-the-green-shadows/?p=2864431

 

Jonathan