Color crossover - an illustration

[pentaxuser]

IMO elimination of all crossover merely produces a technically perfect result like billions of others that could have been taken with any half decent phone.
In my view it is preferable that pictures taken on film should not look like digital.

Alan this is not meant as a "smarty pants" question but does this mean that you will not be currently pursuing your Dignan process to eliminate crossover

I had hoped that your endeavour might have eventually resulted in a home user colour developer process that was cheaper, simple to make and use but squared the circle of being the equivalent of the C41 process

If that's not your objective that's fine but if you could confirm this that would be helpful

Thanks

pentaxuser

 

[Alan Edward Klein]

Ektar is a tricky film - that is to say, any material that has high saturation will pretty ruthlessly punish any error, however small. A slight filtering mishap that might not be very conspicuous on say Portra 400 can become a glaring issue on Ektar. T
Truth be told, I never enjoyed color correcting C41 scans...I found it a frustrating business, overall!




I hope they're OK now

Because of my frustrations as well, I stopped shooting C41 color negatives and stick with E-6 chromes. They're just easier to scan and color-correct. Plus you can see if you're close to the original film's color. Most of the time I never actually compare. I just adjust until it looks good to my eyes.

 

[koraks]

I have the same experience @Alan Edward Klein and also preferred scanning E6. Fortunately, my family used to prefer E6 back in the days as well so most of our family archive happens to be E6. Technically, E6 should be more difficult to scan due to its higher density range, but I think you and I are not alone in finding it more intuitive to get it to look right on a computer monitor.
To remain on-topic BTW: E6 also crosses over, especially as it ages. Boy, do I have some spectacular examples of this on my harddrive, LOL! Not all of those slides from the 1970s stood up to the test of time all that well.

 

[Alan Johnson]

Sure; it's done!

OK thanks, I just thought my experiment not using C-41 might be rather misleading as C-41, ECN2, E6 are the processes you would wish to address.

@ Pentaxuser, I think that we established that Dignan process will not give results completely without crossover so it is not expected to match C-41.
I will continue with it using movie film for a bit.

 

[koraks]

OK thanks, I just thought my experiment not using C-41 might be rather misleading as C-41, ECN2, E6 are the processes you would wish to address.

Sure, I see what you mean. And even if you had any other reason, you wouldn't have to explain; it's your image.

I hope you keep us updates on the Dignan process and that you don't see my critique concerning crossover as a reason to not share your results or otherwise dismissive of what you're doing. I'm following your progress with great interest, in fact.

 

[pentaxuser]

@ Pentaxuser, I think that we established that Dignan process will not give results completely without crossover so it is not expected to match C-41.
I will continue with it using movie film for a bit.

Thanks Alan

pentaxuser

 

[DREW WILEY]

A lot of this is easier to sort out if it's taken a step at a time rather than getting secondarily complicated due to scanning issues all at once. I'm glad I optically print instead. But if I do need a scan for some reason, the quality of that scan can make a huge difference with respect to crossover itself, and not just detail capacity.
I won't go into it here.

And as far as taming Ektar goes, I've commented many times before. You need to appropriately color temp filter it AT THE TIME OF THE SHOT, and not assume you can efficiently post-correct it afterwards. Otherwise, a certain amount of dye curve crossover might be already baked in. That's true of other color neg films too, but less apparent due to their lower contrast and different expectations respecting hue saturation and purity.

 

[Adrian Bacon]

The issue of crossover pops up here from time to time, and it sometimes involves questions that boil down to "what are we looking at, exactly". Yesterday I did my best to put it into words and even pictures (scroll down from previous link to see it), but this is (1) buried rather deep inside a thread about something else and (2) I wasn't entirely happy with the visual example I worked out for it. So today I spent much of the day preparing a better explanation and example images of the crossover issue.

You will find my (amateur-perspective, hobbyist) explanation of color crossover and what it looks like on my blog, here: https://tinker.koraks.nl/photography/sickly-colors-the-crossover-issue/

In all honestly, most of the examples there are pure digital because that was the quickest and most flexible way to make a full mockup. I've also included a couple of real-world (scans from color prints) examples, but you may have seen these already on the forum because I've posted them on here before. Much of the article references to the practice of RA4 color printing and it was written with that practice in mind, but the insights expressed there are also relevant to color slides/positives and also hybrid/digital workflows.

Do with it as you please - learn from it, debate it, torched it to the ground; it's all fair game. Discussion on here is generally enriching and mostly enjoyable, so take it away, please!

(Mods, if you feel that as a result of the digital content included in the post it better fits the hybrid category, feel free to move it, but given its primary purpose of aiding people in working with film- and paper-based materials, I posted it here in the analog section.)

Nifty write up. I've been meaning to do something similar for a while, but have just been really busy.

 

[ludwighagelstein]

I really like the article, but wanted to add a small thing I noticed:


If you'd somehow manage to develop a color film in a process that would result in these curves, your main problem isn't going to be color balance.
From a practical POV, achieving D-Min that looks like this is next to impossible in a just slightly off process. A more realistic version of this plot would probably be something with ok D-Min, and then the blue record tanking between step 8 and 1, touching, or even crossing the green record.

 

[koraks]

From a practical POV, achieving D-Min that looks like this is next to impossible in a just slightly off process. A more realistic version of this plot would probably be something with ok D-Min, and then the blue record tanking between step 8 and 1, touching, or even crossing the green record.

Ah, thanks for your remark, but I'm afraid you may have misinterpreted this bit. Please let me clarify:

1: The image you quoted represents the neutral, "as intended" Portra 160 curves as published by Kodak in their datasheet, with one modification (see next comment). So this isn't something that represents a "slightly off" process - it represents a perfect process. As such, it's very realistic, except for one artistic liberty I introduced for the purpose of this illustration:
2: I shifted the red and the blue color curves so they overlay the green plot by means of a 'creative' manipulation to demonstrate how well (or, if you're critical, how badly) the color curves match for this film when processed properly. On a critical note, I perhaps should have removed the numbers on the vertical axis as they are no longer meaningful; they only apply to the green channel, but I moved the red and the blue ones to show the parallelism between the three different curves.

If we get a bit more technical:

If you'd somehow manage to develop a color film in a process that would result in these curves, your main problem isn't going to be color balance.

Well, yes, the problem would be exclusively a color balance one and not a crossover one because no crossover was introduced in that illustration beyond the crossover that's inherent to the film itself, according to Kodak's very own datasheet! Moreover, the color balance problem would be extremely severe since I totally obliterated the absolute difference beween the curves (e.g. removing the orange mask etc.)

 

[ludwighagelstein]

Ah, thanks for your remark, but I'm afraid you may have misinterpreted this bit. Please let me clarify:

1: The image you quoted represents the neutral, "as intended" Portra 160 curves as published by Kodak in their datasheet, with one modification (see next comment). So this isn't something that represents a "slightly off" process - it represents a perfect process. As such, it's very realistic, except for one artistic liberty I introduced for the purpose of this illustration:
2: I shifted the red and the blue color curves so they overlay the green plot by means of a 'creative' manipulation to demonstrate how well (or, if you're critical, how badly) the color curves match for this film when processed properly. On a critical note, I perhaps should have removed the numbers on the vertical axis as they are no longer meaningful; they only apply to the green channel, but I moved the red and the blue ones to show the parallelism between the three different curves.

If we get a bit more technical:


Well, yes, the problem would be exclusively a color balance one and not a crossover one because no crossover was introduced in that illustration beyond the crossover that's inherent to the film itself, according to Kodak's very own datasheet! Moreover, the color balance problem would be extremely severe since I totally obliterated the absolute difference beween the curves (e.g. removing the orange mask etc.)

No, I havent misinterpreted it. The first curve you show, shows the correct sensitometric curves. The second example I quoted shows yours. If the Red Dmin Density is above the Blue Record Dmin Density, something would be seriously wrong with your process.

 

[brbo]

No, I havent misinterpreted it. The first curve you show, shows the correct sensitometric curves. The second example I quoted shows yours. If the Red Dmin Density is above the Blue Record Dmin Density, something would be seriously wrong with your process.

Are you saying that Kodak's published characteristic curves are wrong? koraks did nothing more than shifted Kodak's published R and B curves to align all three curves at the flat part of the curve.

 

[koraks]

No, I havent misinterpreted it.

I'm afraid you might have, though please see my earlier response.

The first curve you show, shows the correct sensitometric curves. The second example I quoted shows yours.

The second example you quoted shows the correct sensitometric curves, deliberately shifted on the red and blue channels to show the extent to which the curves are parallel in a properly processed Porra 160 negative.
For a real-world negative to show the same red and blue shift, something indeed would have to be extremely out of whack and I think in fact that it will be neigh impossible to achieve this without resorting to dyeing the emulsion multiple times. The intent of the image was thus not to demonstrate a real-world example of crossover (those are shown as actual scans from prints at the end of the article), but to demonstrate the degree to which the color curves are parallel when everything is in order. The ensuing crossover examples are all exaggerated for illustrative purposes.

I hope this added explanation clears it up adequately. I still feel you misunderstood the intent and nature of the illustration, which I'm afraid got lost in translation somewhere, for which I apologize.

 

[ludwighagelstein]

I'm afraid you might have, though please see my earlier response.


The second example you quoted shows the correct sensitometric curves, deliberately shifted on the red and blue channels to show the extent to which the curves are parallel in a properly processed Porra 160 negative.
For a real-world negative to show the same red and blue shift, something indeed would have to be extremely out of whack and I think in fact that it will be neigh impossible to achieve this without resorting to dyeing the emulsion multiple times. The intent of the image was thus not to demonstrate a real-world example of crossover (those are shown as actual scans from prints at the end of the article), but to demonstrate the degree to which the color curves are parallel when everything is in order. The ensuing crossover examples are all exaggerated for illustrative purposes.

I hope this added explanation clears it up adequately. I still feel you misunderstood the intent and nature of the illustration, which I'm afraid got lost in translation somewhere, for which I apologize.

That's exactly what I meant.

"For a real-world negative to show the same red and blue shift, something indeed would have to be extremely out of whack and I think in fact that it will be neigh impossible to achieve this without resorting to dyeing the emulsion multiple times. The intent of the image was thus not to demonstrate a real-world example of crossover (those are shown as actual scans from prints at the end of the article), but to demonstrate the degree to which the color curves are parallel when everything is in order"

 

[koraks]

Alright, so...what's your suggestion then?

 

[Mr Bill]

Alright, so...what's your suggestion then?

Hi, what's happening here is that the densitometer readings are just not representative of how the film is seen by a an RA-4 (light-sensitive color paper) paper. The status M densitometer looks only at three narrow spectral zones. Whereas the color paper has much broader spectral sensitivities.

My considerable experience with Portra 160 film, shot with pro-grade electronic flash, is that optical prints (on the appropriate pro papers) have no problem producing neutral gray tones over a wide range of exposures. The implication being that this film, as "seen" by the color paper, has three neutral color "records" that remain parallel over a wide range. It's only when they are "seen" by something like a Status M densitometer that they are no longer parallel.

The Portra data sheets also show what they call "Spectral-Dye-Density Curves" for a "Minimum Density" and "Midscale Density." Looking at these one can see how "wiggly" the curves are, making the point that the one's "sampling results" can be very dependent on the specific spectral characteristics of the "device" used to "see" the film.

In Giorgianni and Madden's book, "Digital Color Management..." they show examples of some arbitrary color-neg film where they compare characteristic curves made with a densitometer vs what they call "printing density." The densitometer curves are similar to what you (koraks) used for your example of color crosses. Whereas the "printing density" curves are parallel. (As I recall Hunt's book, "The Reproduction of Colour" has similar examples.) These were both well-known color "experts" from Kodak, so one might presume that this is likely reliable information.

The "solution" to this? I guess one could accept that densitometer readings are not a good way to make precise judgments about color-crossover problems in optical printing. The ideal solution would seem to be to transform the densitometer readings into so-called "printing density," but this would not be simple.

 

[Mr Bill]

Fwiw I did read through your writeup on color crosses, and aside from this issue it looks fairly good in the case of neutral gray reproduction. But... there is a much bigger complication for non-neutral colors. Most of my experience has been with a large portrait chain where we put a lot of effort into high quality color. We did extensive testing in film and paper combinations before committing to anything new. Aside from a few limited tests we always stayed with the Kodak professional portrait film of the day (a limited test could go so far as to put a couple hundred full-time portrait studios onto a specific film, with an individual cine processor dedicated to that film; a replenished system varies slightly with the specific film byproducts).

Anyway, many years back we observed that it was possible to have a good neutral gray reproduction (no color cross) with "bad" (our judgment) color crosses in skin tones. And vice versa... good skin tone reproduction but significant color-crossover in gray scales. So we concluded that fairly extensive practical testing was also necessary for a thorough film/paper evaluation. Sensitometric testing, alone, was not enough. When Portra film was introduced (late 1990s?) it was very good in all these respects. But... I can't say what happens when shot outside of nominal 5500K color temp. There are multiple finely-tuned interactions between color layers during processing, so if one of the color layers is unbalanced with respect to the others I'd expect some disruptions in the color reproduction. How bad, I have no idea. From what I read on this forum I suspect that the great majority would not notice such things - it is something of a fine edge. Anyone who used to be satisfied with run-of-the-mill mini-lab prints would probably not be remotely bothered by any such issues.

One might wonder why a chain studio outfit would worry about such things. Well, it's the old rule of QC - why try to control things that you can barely measure? Tolerances tend to stack up, often tending to cancel each other out. But sometimes they mostly stack up in one direction or the other, the result being some out-of-spec product. So, especially with very large volumes, it can be cost-effective to control things to what most photographers would consider unreasonable levels.

Here's an older post describing part of our film testing routine...
Ps, the preview doesn't look quite right, but the link is...

What's the exposure latitude of Color Neg film?

Somehow, I don't think Drew photographs weddings. The initial answer to the question posed in the title is a question - "For which photographic usage?" If you are controlling the light, and aiming for extremely accurate and/or dependable colour rendition, Drew is completely right. If you are...

www.photrio.com

 

[koraks]

@Mr Bill thanks for your thoughts; they're very clear and to the point as always! The only thing I'd add is that your views are a (very welcome!) practical perspective that I feel very nicely complements the more theoretical and basic approach I took in the blog I wrote. My blog, indeed, only intends to answer the question "what is color crossover and why is it called that way", by means of a few mostly synthetic and exaggerated examples. Hence, your additions are very helpful in answering a few questions (e.g. the meaning of the divergence of the curves in the Kodak datasheet) as well as provide a useful warning that the issue of crossover is sometimes more complicated to figure out in reality than in my 'digital lab setting' experiment.

The status M densitometer looks only at three narrow spectral zones. Whereas the color paper has much broader spectral sensitivities.

Indeed, and partly overlapping ones, at that, whereas status M likely (haven't looked into it yet) considers a couple of wavelengths that are actually separated.

Whereas the "printing density" curves are parallel.

Yes, I can imagine that - and an interesting angle on this would be to combine photospectrometer readings of test strips with color densitometer readings and see how they can be correlated.

Anyway, many years back we observed that it was possible to have a good neutral gray reproduction (no color cross) with "bad" (our judgment) color crosses in skin tones.

This sounds like you might have run into metamerism issues, but it's just a guess based on likelihood more than thorough analysis.

 

[Mr Bill]

Indeed, and partly overlapping ones, at that, whereas status M likely (haven't looked into it yet) considers a couple of wavelengths that are actually separated.

Hi, yes, I don't think status M specs are readily found on the internet, so roughly they have peaks at about 450 nm, 540 nm, and 640 nm. Then the mostly useful width of each response is about 10-20 nm wide, depending on what one considers useful. Tabulated data is available in the IS&T Handbook of Photographic Science and Engineering.

Yes, I can imagine that - and an interesting angle on this [trying to get to printing density] would be to combine photospectrometer readings of test strips with color densitometer readings and see how they can be correlated.

Yeah, that would be sorta the holy grail. But probably more difficult than meets the eye. I've never seriously played with that sort of thing, though.

This sounds like you might have run into metamerism issues

Nah. The earlier color neg films, circa 1980, printed onto the matching pro papers, really did have color cross issues. I don't recall when these were (mostly) resolved, maybe it was on VPSIII?

A typical sort of issue would be the skin tones, as they approach specular highlights, would go slightly "cool," meaning a bluish or cyan shade. Acceptable for most people, but for someone with very pale skin their skin might get a "dead" look. If you corrected it (by removing cyan) other people in the shot would get reddish, even white clothes going that way. So it was a narrow path to walk.

Fwiw the RA-4 papers tend to be much more tolerant to light sources than ink jets, etc. The rule of thumb is, if it has a larger color gamut - the ability to make stronger colors - it is also more subject to metamerism (failure). The reasoning is: more pure colors are the result of dyes with narrower peaks, and this also makes them more sensitive to the light source.

 

[koraks]

Then the mostly useful width of each response is about 10-20 nm wide,

Yes, but this will depend a lot on the exact implementation of the device. Put differently: especially 'back in the day', getting 10-20nm buckets was pretty challenging especially in desk-top equipment that had to be quasi-affordable as well. I wonder how different systems compared in performance in particularly this respect.

I've never seriously played with that sort of thing, though.

Me neither. Lacking a densitometer and frankly, the inspiration, I don't think I'm going to go there. I might unleash the photospectrometer on RA4 stuff at some point, but currently have no concrete plans in mind...well, perhaps a bit, but they have to wait for now...

he earlier color neg films, circa 1980, printed onto the matching pro papers, really did have color cross issues.

Ah, no doubt!
I'm still turning around in my head that interesting issue of being able to print neutrally balanced grey scales while having trouble with skin tones. I'm trying to figure out how that would work sensitometrically.
As to the practical implications with cyan skin vs. red garments (simplifying what you said) - yes, that's a familiar problem I think to anyone who has worked with crossed-over color film. It's the exact kind of hole you find yourself in with no chance of ever getting out of it.

more pure colors are the result of dyes with narrower peaks, and this also makes them more sensitive to the light source.

Yes, that would certainly be part of it. There are other issues as well. For instance, in pigment inkjet, PB15:3 is a popular pigment, but it's prone to bronzing, creating a copper/reddish luster to dense areas of blue and cyan tones. Then there's the issue of how a microporous surface responds to light, as well as the surface material itself. There are, in fact, many factors that interact when it comes to metamerism and other/related forms of light-dependent (both spectrum and angle of incidence) color effects. Darn complex subject matter, that is.