Blue Cast on certain Color Checker Grey Scale with Portra 160

[AlexisMagni]

I believe we already talked about this somewhere in the past Christoph, but I am not sure. Anyway, now I have a proper setup equipment, using the best exposure practices, so I am bringing this once more into light.

I always shot a ColorChecker on the films I test. Using DC Mode on this Scanned Neg however generates quite a strong blue cast on one of the samples which shouldnt be there. I tried all the Portra 160 Film Types, and this doesnt help. This day I also shot a bigger color pattern I printed myself, which isn't reference grade as the ColorChecker, but you have more colors to check if you like that.

[C.Oldendorf]

I believe we already talked about this somewhere in the past Christoph, but I am not sure. Anyway, now I have a proper setup equipment, using the best exposure practices, so I am bringing this once more into light.

Yes, we did talk about it and there are two possible answers with the same solution.

a) We see development conditions that differ from what the manufacturer published in their datasheets.

b) The film data in ColorPerfect is about the straight section of the characteristic curve. That means if you have densities that fall into the shoulder region of the curve, the three colors (R G B or on the film C M Y) will behave and fail differently, which will introduce color casts in the values that come from those densest regions, i.e. the highlights in the positive.

The tool to fix this is FilmType / SubType / FilmGamma, but in case b) we do not have the same characterization leading to neutral grays throughout the image, and you will need to use selections or blend two images in editing. I lean towards this because we also discussed B&W shots that you exposed very, very generously.

[C.Oldendorf]

Portra 160 (2011).jpg (109.99 KiB) Viewed 617 times

[Full image link - opens in new tab]

Kodak typically does not plot the shoulder behaviour of the film, but in the data we do have, we see that at 0.0 log exposure red visibly builds up less density than blue and somewhat less than green. The blue line continues straight through. So if we see bluish brighter grays on the step tablet, that can be due to this. 0.3 log exposure is equal to 1 EV, so if you were to let your darkest tones fall at Zone III — that is, spot-meter them and expose the result by –2 EV — you should get recognition of substance and texture in those darkest parts, and about 7 EV afterwards should remain well behaved. If, however, you exposed your darkest image portions to be at, say, Zone VI, then you would have nothing less dense on the negative. You would still get a printable result, but you would be pushing into that shoulder. When I say Zone ... I mean it in the sense Ansel proposed for B&W.

[AlexisMagni]

Now I remember, after I saw that graph, which I believe you mentioned in the e-mail as well

About Exposure. This time, I shot the Portra 160 with my Swiss Knife of Analog Cameras, Nikon F100. This will do basically anything other than serve me coffee. Meaning, I used Spot Metering at the 18% Gray Sample, so Exposure should be as perfect as it gets. The 0 you see on my Smartphone is because I shot this at Zone V Exposure, and also a +1 afterward.

You are also lucky I have this same image scanned by my Nikon CoolScan, which should clear any concerns if the development of this film has led to issues.

[Full image link - opens in new tab]

Now for comparison, here is the DSLR Camera Scan but this time using SC instead of DC Mode.

[Full image link - opens in new tab]

To my untrained eyes, I would guess the DC Mode seems to be pushing the colors on the Blue Hues more aggressively than others, leading to this blue cast on the last Sample before Pure White?

[C.Oldendorf]

Let's look at the raw photo / scan then. Camera noise is another suspect but we'll see.

[AlexisMagni]

Sure Christoph! Just before I head out to the street, here is the scan!

[C.Oldendorf]

I don’t have a decent explanation yet — it could be about multiple things. My favorite theory right now is camera noise.
That can be proven or ruled out. When you’re set up to shoot repro the next time, would you please shoot the frame again?
Once with roughly the same exposure you have now, and one more with +3 EV or even +4 EV (a significant overexposure).
Just keep the camera and everything else as steady as you can between shots, since they’ll be blended for a test.
Using the darker regions from the second shot (in a brightness-matched form) lowers the noise level by the number of stops we overexposed.

If you can configure NEFs with and without pixel shift, that would be great as well. I don’t yet know if I trust that technology.

[AlexisMagni]

That makes me feel like I am doing my job as Beta Tester! We are here to find these corner cases to help you develop solutions. Happy to help, however long it takes.

Knowing I would probably be asked to scan more samples, this time I left my Kit mounted. Tomorrow I will have them here

[AlexisMagni]

Let's go

Correct Scan Exposure

Portra160_0_ScanExposure.nef

(31.15 MiB) Downloaded 3 times

+1 Scan Exposure

Portra160_+1_ScanExposure.nef

(32.17 MiB) Downloaded 1 time

+2 Scan Exposure

Portra160_+2_ScanExposure.nef

(28.36 MiB) Downloaded 1 time

And since I had it, I thought it would be useful to also Scan the +1 Exposure of the ColorChecker done at the moment of capture

Portra160_+1_CaptureExposure.nef

(28.64 MiB) Downloaded 1 time

They are all regular RAW, didnt use PixelShift for these.

[C.Oldendorf]

Alexis, thank you so much! I didn’t really understand what the Capture Exposure shot was about, but the rest was plenty to work with.
Attached is the merged negative data I prepared for the experiment. Maybe I can get away with not explaining how I put it together, but if not I can always write it down.

The red channel comes from the standard capture, the blue channel from the +2, and the green channel is partly from the +2 and partly not.

The effect persists. Since the behavior differs from the Nikon Coolscan example, it gives us something useful to think about.
Whether it ever amounts to anything, or if it’s simply part of the price we pay for pulling all this off with digital cameras rather than scanners, remains to be seen.

[AlexisMagni]

You dont need to thank me for that!

This is a reminder for me why I have always had trouble to engage with Color Negative Film, like I do with Black &White, and even Positive Film. It doesn't have to do with ColorPerfect, oh it has nothing to do with ColorPerfect. But the nature of Negative Film often leaves me more frustrated than in awe to the rendition I get out of it. Positive Film is much harder to achieve a good exposure, but the fact that we don't have to deal with the conversion makes it worthwhile.

I am also not very good with colors, I am considering going to Digital for Color Photography, and leaving Film to where it truly shines to my eyes, B&W. But of course I will keep doing my best to understand this and other behaviors and how to deal with it.

Another interesting observation is how the +1 Exposure during Capture doesn't show the Blue Shift, maybe it does a bit but on the darker side of the GrayScale, but the rest looks very good.

[C.Oldendorf]

What is the +1 Exposure during Capture NEF, please elaborate on how it is different.

[AlexisMagni]

What is the +1 Exposure during Capture NEF, please elaborate on how it is different.

I metered the Middle Gray for the first image, and overexposed +1 for the photo with the +1 on the screen. So it has a stop over the regular one.

[C.Oldendorf]

I'd like to understand the difference between "ScanExposure" and "CaptureExposure"

[AlexisMagni]

I'd like to understand the difference between "ScanExposure" and "CaptureExposure"

Sure

Scan Exposure - Exposure Compensation during Scanning. Doesnt have anything to do with exposure at Film Camera
0 - Optimum Scan Linear Histogram
+1 - 1 Stop over Optimum
+2 - 2 Stops over Optimum

Capture Exposure - Exposure used at the Film Camera in the moment the Color Checker was photographed

There are two different frames. One has the 0 on the Cellphone beside the Colorchecker. This one was metered for the Neutral Gray. The other Frame showing +1 on the cellphone was metered with +1 Exposure compensation.

The Scan of the +1 at Capture Negatice was done with optimum Linear Histogram Exposure, no clipping

[AlexisMagni]

I'm thinking, how else would we put DC mode to the limit? This is a big feature, and although it works very well most of the time, I feel like ColorPerfect is known for it's color purity, and DC Mode still seem to need some polish to keep the status earned by previous versions. There seems to be some non linearity issue, affecting the neutrality of the conversion. But other than shooting ColorChecker, I dont have many ideias. If anyone has any let me know!

[C.Oldendorf]

I don't concur. Scanners are scanners and digital cameras are not.

Albeit we are trying to overcome their limitations with math I never thought possible, it remains an unfair comparison. Let’s put it this way: in real-world scenes there are millions of surfaces with millions of spectral properties. All need to be reduced to sensible spectral averages to work in a three-color system. That complexity can be greatly reduced for film scans. There is only a very limited set of dyes and couplers that can chemically be used in film. That makes it very possible to engineer the light source and sensor characteristics in a way ideally suited for the task. Also, we have three discrete readings for every pixel, we can sample multiple times if we must, and so forth.

While it is possible that I have something to do with that blue hue, it is not a given. I guess all I’m saying is: don’t take what we have for granted, and don’t expect equivalence between scanners and cameras. Just the same as scanners will never be as fast as cameras, cameras will never be as accurate or as fit to the task as scanners. Both will be good enough within the constraints of time and image quality.

Ancient test images come out amazingly good (shot on Nikon D300s when it was new):

[Full image link - opens in new tab]

Way back when (shot on Nikon D200 and processed close to 15 years ago with what we had then):

[Full image link - opens in new tab]

Which, when color balanced, turned out to be due to camera noise in a way I won't go into here:

[Full image link - opens in new tab]

Today, with the same data and the same deficits therein and all, we get to:

[Full image link - opens in new tab]

The relation of film density to final image intensity was, and will stay, extreme. Overcoming the above needed some safeguards. It is possible that one of them goes into that exemplary bluish gray. If that is so, it could be fixable, or it could be of a nature that touching things steers us further away from a reached optimum. That is exactly the kind of work that I did, let rest for a few months, did again because it felt not quite there, and so forth. I think the shots date to 2011; I’m not even sure if PerfectRAW had seen the public light of day back then. I was younger, that’s a given and it was the very last opportunity to shoot Kodachrome and have it sent in to be developed by Kodak These were both shot on Fuji Superia Reala 100 - which I wish I had frozen more of.

[AlexisMagni]

Legendary Fuji Reala! I really wish I could have shot these Fuji films in their time...

I understand about how Digital Cameras are a Flawed Technology from the start, when we talk about Color Integrity. Unless we are back to using CCD Sensor, or other technology that doesnt require the chroma to be interpolated because we cant have each pixel to fully interpret all possible colors. But this still feels strange to me, because I have seen Digital cameras fail in their conversion, compared to my Nikon CoolScan, but usually when they fail, they fail for the whole image. It will have almost uncorrectable casts, weird color shifts, etc. What I have never seen happen in any other conversion or software is having the kind of non linear cast that happened in the Color Checker when we use the DC Function. That only happens on a specific area of density of the negative, the rest is fine, Pure White is perfect, which leaves me puzzled.

Negative Film Conversion isnt easy, as you said, you have a very limited range on the negative that has to be amplified to a whole spectrum of colors on the final image. DC Mode in its attempt to expand these colors even further than SC mode did, maybe is going to areas where an insignificant color cast is amplified to a very noticeable one?

I know this is not in any way scientific, probably full of flaws, but I have made a simple Manual conversion in Photoshop, inverting, Adjusting the Levels RGB Clipping points, and nothing more, and this if nothing, shows me the original image has perfect Neutral Gray Rendition, and very linear as well, as I see no issue happening, other than of course the overall image not up to what a professional software can do.

[Full image link - opens in new tab]

Why would we only see this issue in DC Mode? This makes me think it has to be something with DC, and not with the Camera Technology overall?

[C.Oldendorf]

In order to prevent the noise-induced brightness explosions mentioned above, some safeguards had to be put in place, and it is conceivable that one of those has this effect. But it is not a trivial thing to check — it could take weeks or longer to do. At the same time, flipping the histogram around does show that the densities are more or less continuous. But nothing more. It does not yield a single element of pure and convincing color, nor does it restore brightness relationships between individual items, because it is physically the wrong thing to do in the first place.