Weird Colour Loss - Sony A7RII
I was watching a video on YouTube by wolfcrow about using ETTR with the Sony and in one section he talked about why he doesn't use Capture One with the A7RII. It's very interesting watching.
This is a direct link tho the section where he shows the issue:
https://youtu.be/gNWWEMS_rEk?t=1081
Any ideas what might be causing this?
This is a direct link tho the section where he shows the issue:
https://youtu.be/gNWWEMS_rEk?t=1081
Any ideas what might be causing this?
0
-
It would be interesting to know if he has now run the same investigation using full RAW rather than S-RAW and if the results are the same.
Also whether others see the same results using similar shooting and editing methods.0 -
Could it be an issue with the preview only? Would be interesting to see if an outprocessed tiff has the same color desaturation.
Anyway, it would be a shame if C1 really would have lesser highlight recovery muscles than LR, especially for Sony, above all other issues this should be solved (if confirmed) as C1 image quality is THE argument for me to use C1.
Just my thoughts, for what it's good for...
Cheers.0 -
Hi Meanwhile
Thank you for your reference to wolfcrow's fascinating video on ETTR. At last, an ETTR post that is practical and easy to apply. Just set the in-camera jpeg style to "neutral" and "minimum contrast" and the in-camera histogram is shifted to the left to emulate the highlight end of the raw spectrum. It begs the question why manufacturers have failed to build this into a £2,000 camera?
This ETTR procedure seems to work quite well with my Canons with a couple of caveats. The end-point of the highlight tail of the in-camera histogram is rarely clear-cut. So personally I will continue to bracket. And the C1 histogram seems to clip the highlights prematurely compared with ACR and DxO. In my experience DxO makes the best use of the somewhat limited DR that my Canons (5D2 and 7D) possess. So I now use DxO as my primary raw converter (Standard preset) and transfer to C1 for fine-tuning the WB and Clarity, and the denoise/sharpen balance.
HTH
Peter0 -
[quote="meanwhile" wrote:
Any ideas what might be causing this?
the author (of the youtube clip) it seems simply does not know how to use raw converters (ditto for the topic starter and for the participant who alleged that author of the youtube clip knows how to use ETTR with A7R2 - no, he does not), I repeat my posting from dpreview where the same youtube video was floated a month ago :
"this screenshot below ( http://s8.postimg.org/62zx5vkyr/c1e.jpg ) is a crop from a raw that is 1/3 EV till clipping in raw channel (per rawdigger) in the white patch - do you see washed out greens ? pay attention to the profile and curve"
screen shot again is here = http://s8.postimg.org/62zx5vkyr/c1e.jpg
or with the profile appropriated from C1 CH edition for A7R2 purposes and linear scientific curve to avoid any attempt by C1 to compress highlights closer to clipping = http://s26.postimg.org/fcrkn8lt3/C1onemore.jpg0 -
Hi deejjjaaaa,
I am not really sure I'll get your point.
Are you saying that if exposed to the right, which you define that no single white patch or pixel is clipped, and if using a profile which is available in a 6000 USD raw converter (C1 CH), then you'll get the similar highlight recovery than with LR?
So the ordinary C1 Pro user which doesn't spend as much has no clue about converting raws correctly?
For C1 CH it is advertised
"Obtain high color accuracy with robust ICC profiles optimized for both color precision and three dimensional gradients at the same time. "
Isn't that exactaly what we all want even in a dumping price raw converter like C1 Pro?
I would have expected that (if not via the exposure tool then at least) via the HDR highlight slider C1 Pro with provided standard profiles could compete with LR highlight recovery (which the author of the video did not try).
Btw, how do you (or how would you) ETTR holding a camera in your hand?
Thanks
BeO0 -
Unless I misunderstood the video (which is quite likely) the conclusion was that you could expect to satisfactorily recover about 1/3 stop over exposure with full colour accuracy after recovery, right?
But even then with the sample files the author felt that noise was an issue. (I wondered if that was a compressed RAW factor - hence my question above.)
1/3 stop is probably typical exposure measurement latitude. I will admit I don't really do any colour critical work so I have not spent much time on investigation and understanding but in the main I would not be to worried about going a stop over measured exposure values for most of my stuff. More than that is viable with a Pro body but utility bodies seem to be quite a bit more finicky. As one would expect.
Nevertheless I would still bracket shots if results were likely to be critical . So I guess the main area of "need to know" would be Wedding shoots an similar?
I know nothing useful about Sony's system though.
Grant0 -
As the topic starter, whatever.
I put it forward as a question as to what was happening.
Secondly, I already suggested that he use Linear Response.
Thirdly, where does a mere mortal get their hands on that specific profile?
Fourthly, I didn't suggest that it was the correct way to ETTR with the A7RII.0 -
[quote="BeO" wrote:
Are you saying that if exposed to the right, which you define that no single white patch or pixel is clipped, and if using a profile which is available in a 6000 USD raw converter (C1 CH), then you'll get the similar highlight recovery than with LR?
not exactly - in my examples I used a simple home made matrix+trc profile or profile that is available (simply check the folder with camera profiles for backs) in a regular C1 distribution (even if it is intended for CH edition)... the point is that OEM profiles (from P1 or Adobe) do things behind the scenes, including boosting brightness, because an average joe is expected not to do ETTR... so if you ETTR you need to either use different profiles or modify profiles _AND_ use certain raw conversion settings, in C1 it is naturally better to use linear curves or use "auto" only when you know that camera profile itself has a custom tag inside instructing C1 to select linear curve (P1 does include such tag in some profiles for their backs when profile is intended for repro work)... but safer bet is to specifically use linear curve... and then yes, if you have a target and it is exposed (the most bright patch) ETTR-like but w/o clipping in raw channels then darker (than the most bright white patch) color patches (like greens) naturally are not going to suffer any ill effects... the author of the youtube clip publicly demonstrated that he has no clue what he is doing...[quote="BeO" wrote:
For C1 CH it is advertised
"Obtain high color accuracy with robust ICC profiles optimized for both color precision and three dimensional gradients at the same time. "
sure, but that does not mean that you shall believe every idiot out there who posts something on youtube just because it is youtube ?[quote="BeO" wrote:
I would have expected that (if not via the exposure tool then at least) via the HDR highlight slider C1 Pro with provided standard profiles could compete with LR highlight recovery (which the author of the video did not try).
why'd you need to use HDR sliders in C1 at all in this particular situation - nothing is clipped in raw and color patches are way further from clipping in raw than the most exposed patch (the brightest white patch) ?[quote="BeO" wrote:
Btw, how do you (or how would you) ETTR holding a camera in your hand?
if you are talking about Sony A7R2 (similar shall work for any A7 series/mark cameras - but Creative Styles might be different) then this is what I do to see in EVF or during the post shot review how close I am to the actual clipping in raw channels :
WB = UniWB
DRO/Auto HDR = D-Range Opt./DRO Lv5
Creative Style = Neutral (contrast -3, saturation -3, sharpness +3)
Color Space = sRGB
Zebra = 100+
or the minimal set :
WB = UniWB
Creative Style = Neutral (contrast -3, saturation -3, ...)
Zebra = 100+
with these settings I have both real time zebra and blinkies in post shot review indicating me clipping in raw channels with ~1/3 EV precisions... the price I pay is green tint from UniWB (I do not mind and actually prefer that), and some other minor things about face detection under the daylight, but when I am doing portraits I use the fact that A7RII (and A7 - I did not test with other A7* cameras, but I believe shall work in a similar fashion) with face detection and matrix metering will expose a detected/AF'd face (white, caucasian) with green channel ~2.7 EV below clipping in raw, so I switch settings effects off and simple add the proper exposure correction putting the facial skin where I need it do be (knowing that 0EV means skins will be around 2.7 stops below clipping) - the camera does the rest.
or you add a Picture Profile on top of the above to have greyscale image in your EVF - I developed the following set usable for me:
Gamma = Cine3 (reason : allows ISO below 800 + curve itself)
Knee = Auto (may be manual for Cine3 = Point :105%, Slope +5)
Saturation = -32 (and removes green tint too with Setting Effect = ON)
Color Mode = S-Gamut
Color Phase = 0
Color Depth = 0s
Black Gamma = Wide/+7 (reason : darker areas = brighter)
Black Level = +15 (reason : darker areas = brighter)
so you manipulate parameters for OOC JPG to make it so that you can get a quite precise indication when the actual clipping in raw data starts... yes, thibngs like Magic Lantern firmware for some Canon cameras is a good idea, but we do not have anything lie that for Sony... so have to mutate the OOC JPG instead0 -
[quote="meanwhile" wrote:
Thirdly, where does a mere mortal get their hands on that specific profile?
even a "regular" profile shall do with linear curve... now if you actually clip raw data for color patches then no profile will help much you as a raw converter will be forced to invent the data for 1 or 2 channels from unclipped one(s), but this is not the situation we are dealing with...0 -
deejjjaaaa, I use very close to your method myself for ETTR, I just can't stand the green in the WB, so it's not 100% accurate. Thank you for detailing it on the DPR site. 0 -
[quote="deejjjaaaa" wrote:
[quote="meanwhile" wrote:
Thirdly, where does a mere mortal get their hands on that specific profile?
even a "regular" profile shall do with linear curve... now if you actually clip raw data for color patches then no profile will help much you as a raw converter will be forced to invent the data for 1 or 2 channels from unclipped one(s), but this is not the situation we are dealing with...
So you think in this case simply changing to Linear Response rather than Auto will bring those colours back? It's actually what I suggested to him to try first.0 -
[quote="meanwhile" wrote:
your method
I want to note that it is __not my__ method... using UniWB or near-UniWB is known for years and UniWB is the key thing to get OOC JPG based zebra/blinkies/histogram as close to raw data as possible, I personally read myself about UniWB from Iliah Borg around ~2007 and I bet this thing was known to people well before 2007 ... I just listed my particular settings that I started to use when I got A7 in Feb 2014 and modified (Creative Style is different) for A7R2 when I switched to it from A7... before I was using the same for my Olympus E-M1...
yes, a lot of people do not like the green tint, hence again using Picture Profile allows to have greyscale image (histogram will be with R=G=B though, as it is till based on OOC JPG).0 -
[quote="meanwhile" wrote:
So you think in this case simply changing to Linear Response rather than Auto will bring those colours back? It's actually what I suggested to him to try first.
if he was exposing his color target in a manner where the most bright white (the the best reflectance) patch is not clipped in raw then he shall have no issues with colors at all...
even with overexposure -> I found an old raw on my computer where I have a passport (x-rite) target overexposed with clipping in raw for 3 most bright greyscale patches (mean around ~1.5EV overexposure /past clipping/ in raw in regards to the most bright white patch) and yet with a regular generic or standard profile for that camera, A7, I have no issues with getting colors w/o the effects shown in youtube clip for color patches by pulling the exposure in C1 by, for example, -2EV with linear curve or even w/ non linear curve like "film" (yes ~3 color patches are also clipped in green channel, but still they do not turn grey or whatever happened in the clip)... so the author of the clip did something very wrong or if you want to give him the benefit of doubt then we have to assume that C1 Experess (v8) was/is severely crippled in functionality.0 -
Thanks deejjjaaaa,
What does OOC stand for, !out of clippling?
I don't have an A7.. (yet) and try to remember then.
Regarding what you've written to profiles, what I read is that applying a profile in the base characteristics is destructiv? For example if a profile boost brightness and bleeds green/magenta, and you cannot compensate that with exposure, brightness or highlight slider, this is the only conclusion to draw...
I don't have special settings, I rely on the histogram and the visual judgement in the EVF, and on the generic camera profiles... reading thru your post I realized that I am much more the average joe than I thought... 😊
chees
BeO0 -
OOC JPEG = Out of Camera JPEG 0 -
I have requested the raw-file from wolfcrow. We cannot reproduce something similar inhouse, nor with other bodies, so we are very interested to investigate what is going on. 0 -
wow, that's great Christian! 0 -
[quote="BeO" wrote:
Regarding what you've written to profiles, what I read is that applying a profile in the base characteristics is destructiv? For example if a profile boost brightness and bleeds green/magenta, and you cannot compensate that with exposure, brightness or highlight slider, this is the only conclusion to draw...
you can create a profile that will certainly make things ugly ... you can increase or decrease brightness or you can introduce various color shifts (like hue shifts or saturation increase/decrease, etc, etc)... of course the actions with exposure/brightness/etc sliders are combined with profile effects... camera profile is applied after those are done on raw rgb data internally... I was trying to make a point that when you do ETTR (which generally /but certainly not always/ means expose as much as you can and then pull in raw conversion) you want to use profile that avoids boosting the brightness with 1D/3D luts inside it and you might want to use linear curve... now you can try to compensate a different profile by pulling more than you 'd otherwise in raw converter, but then I think the less math done with the data the better.[quote="BeO" wrote:
I don't have special settings, I rely on the histogram and the visual judgement in the EVF, and on the generic camera profiles... reading thru your post I realized that I am much more the average joe than I thought... 😊
you can still make beautiful photos without all of these... no amount of technical tricks can replace some artistic vision/talent/you name it 😄
you can perfectly use profiles supplied by P1 if you ok with how they render colors...0 -
9.0.3 released
including "Highlight recovery for A7RII fixed"
Maybe that's it, then Bravo Phase One!
Cheers
BeO0 -
Can anyone confirm 9.0.3 solved the issue?
cheers0 -
Great stuff if yes, thanks Phase One! 0
投稿コメントは受け付けていません。
コメント
21件のコメント