Colour Negative Orange Cast

[carllooper]

I've started this new thread, because Roger's thread was being somewhat hijacked by this subject, and I'm interested in this subject.

So researching the orange bias in colour negative I've come across a number of articles. This one is from the National Film and Sound Archive in Australia, which is in charge of preserving the nations film heritage, so I'm hoping their awesome responsibilities don't include publishing misinformation on the subject. However the reading is quite light so not much to go off.

http://www.nfsa.gov.au/preservation/han ... enic-dyes/

This is their brief explanation on the orange:

In practice the dyes do not precisely absorb all the specific wavelengths and may have secondary absorption characteristics. To improve the performance of the dyes 'masking layers’ are incorporated in colour negative materials. The masking layers use dyes that are also formed during processing. This is noticeable as an orange appearance to colour negatives.

I'm assuming there should be a comma following the word 'dyes' in the sentence: "To improve the performance of the dyes [comma] 'masking layers' are incorporated ..."

I'm not exactly sure what is meant by 'masking layer'. This statement is not rhetorical. I literally don't know. Looking at diagrams of the structure of film, there is a yellow filter layer and a red filter layer (although many diagrams I've found only show the yellow). Jumping too readily to conclusions (because I'm otherwise naive on the subject) I'm erroneously imagining (as I later discover) that these filter layers is what is meant by the 'masking layers", based on a correlation between the orange appearance of film and the yellow/red filters. But if masking layers "use dyes that are also formed during processing" it must mean something else. Furthermore, by 'performance of dyes', it can't be referring to what happens during exposure because the dyes are not yet there (not yet performing). The dyes perform during projection. So a 'masking layer' can't be meaning the filter layers.

With respect to the filter layers, (and no doubt related to masking layers as well) one thing I came across is that silver halide is naturally more sensitive to blue. So in order for the silver halide of the lower layers to respond to red and green and not be swamped with a blue signal you would want to block the blue signal, the top layer having already encoded a good response to the blue signal (because sliver halide is naturally biased that way). The next layers down, no longer exposed to blue, can then only respond to red/green light. Since green light is closer to blue light it makes sense that the second layer would be responsible for encoding green and finally the third layer, responsible for encoding red, would need a red filter in order not to be swamped with green light.

Thinking out loud ...

So really the top layer actually encodes the entire spectrum (red green and blue) but gives much greater weight to blue because silver halide is biased towards the blue end of the spectrum. The next layer encodes red and green only (because no blue light reaches it) but because of the silver halide sensitivity towards the blue end it encodes a stronger response to green. The last layer encodes red only.

Now the encoding is done by using dyes that are the reciprical of the incoming light - staining the top layer yellow, the middle layer magenta, and the last layer cyan ...

So the top yellow layer while principally encoding blue, is partially 'polluted' with an encoding of red and green. Likewise the middle layer, principally encoding green, is partially 'polluted' by an encoding of red. Only the bottom layer really has a clean colour signal (incoming light). But is it correct to read the layers as 'polluted'?

Ok. Back to masks. The following seems like a really good explanation and clears up my conflation of filters and masks:

http://armycommunications.tpub.com/SS05 ... 140040.htm

Negative masking to correct dye error.
Color masks, which provide a neutralizing effect for unwanted dye absorption, are readily seen in the unexposed areas of a color negative. There are two such masks: a yellow mask in the green-sensitive (magenta dye-forming) layer, and a reddish magenta mask in the red (cyan dye-forming) layer. Together, they produce an orange tint in unexposed areas.
There is no need to mask the blue-sensitive (yellow dye-producing) layer since the yellow dyes usually have negligible unwanted dye absorption

So the bottom layer, while it has a clean incoming signal (red only light), is encoding it in terms of a cyan dye, but that dye is not as good as it otherwise might be. It is in error with respect to the light it otherwise represents. Likewise with the magenta layer. Ironically, the layer which is responding to the least cleanest signal (in terms of colour separation) - the top layer - is responding with the best dye (in terms of absorption).

Hmmm.

I think I've got enough information now to write this up as a rudimentary algorithm and test assumptions. I'd start with an emperical curve representing the sensitivity of silver halide to the wavelengths of light. This would be a constant. Then I'd need three terms to represent the filtration going on. The first is actually the UV filter layer. This would transform the reference curve to a new one which didn't include a UV response. A second transform would use the term for the yellow filter, to transform the curve again, and finally the term for the red filter to obtain the last curve.

The first curve, following UV filtration would very closely match the yellow dye forming layer. The second curve, however, which should match the magenta dye layer, won't do so because of inefficiencies in the magenta dye. Likewise the third curve. Furthermore, due to the masks introduced to correct such inefficiencies, there will be new terms required for this correction, if one is to obtain a good representation of the overall absorptions.

From there one should be able to adjust the strength of the filter terms and masking terms (in combo with a similar representation of print film) to guesstimate optimum values for the masking terms, or otherwise produce a rough model of the process anyway.

One thing I don't know (yet) but assume is the case (keeping in mind that assumption is the mother of all f**kups), is that the filter layers are dissolved (or transformed in some way) during film processing.

Carl

[richard p. t.]

Hi Carl,
I'll put in a few cents worth here:

First, regarding the colour sensitivity of Silver halide, Silver halide crystals in their basic state are ONLY sensitive to light with at least the frequency of blue light. It is hard to make Silver halide sensitive to other frequencies as well. The reason it is hard, as I understand it, is because the lower frequencies don't have the energy for their photons to dislodge the relevant electron in the Silver halide molecules. As I say, this is how I understand it. To make Silver halide 'sensitive' (in the way we mean that term in photography) to other frequencies requires the addition of sensitizers. These sensitizers are actually themselves dyes. I believe the way these dyes work is that the incoming light is able to affect these dyes, and then an electron from these dyes is then able to nock the relevant electron in the Silver halide molecule off its perch. In its native state, Silver halide is a creamy colour. Black and white print stock is pretty much the colour of native Silver halide. Black and white film for cameras, which is almost always panchromatic, isn't this creamy colour. One of the reasons its not this colour is because of the sensitizing dyes that make the emulsion panchromatic. These sensitizing dyes come out of the emulsion during the reduction of the Silver halide to metalic silver and end up in the developer (and the fixer in the case of the non-reduced Silver halide that remains after developing). Regarding colour film, the top layer, the blue sensitive layer, can be understoon as effectively un spectrally sensitized. The subsequent green and red layers are sensitized such that they are blue and green and blue and red sensitive respectively. The red layer usually isn't sensitive to green at all.

Regarding the orange masking; in the chromagenic colour process (as is used with all modern colour films except Kodachrome type films) it is commonly said that the colour dyes that will make up the colour image at the end of the process are 'already in the film' as opposed to incorporated later in the process or some other way as in other colour photographic systems. In fact, the colour dyes are only partly there in the emulsion; they can be understood as incomplete colour dyes or part of the components of colour dyes. For the dyes to actually form, they require the presence of a) a chromagenic colour developer which contains more than just a Silver halide reducing agent (ie additional components of the colour dyes) and b) the presence of oxydizing agents that are produced during the reduction of Silver halide to metalic Silver. All three things are necessary to form the finished dye - the dye components in the emulsion, the dye comonents in the developer, and the reaction products of development. Its clever and complicated stuff. Now, it is possible to make chromagenic dyes this way that are pretty neutral and produce the colour we see in colour reversal film - ie, true colours more or less - so why introduce a colour mask in negative film? The answer I believe comes down to the fact that the roll of negative film is for duplication. This requires the chromagenic process to work twice (at least twice, in the modern motion picture industry this reproduction happens through a number of generations, typically 4 - the camera neg, an inter pos, a dupeneg and a print). Any error in the dyes will be magnified by this repetition of the process. But it proved difficult to make chromagenic dyes with all the required properties that were also as accurate as multiple generation work would require. It was easier to find a yellow dye that was the correct colour and fit these other requirements than it was for either cyan or magenta. The nearest colour dyes that were able to fit the requirements were a little 'off' the colour they ideally would be. Thus, the idea of biasing the entire negative in some way - both the areas of the negative that had formed these slightly off colour dyes and the areas that had no dye - was conceived. Any particular area of the negative will either have this slightly off colour dye, or else have a slightly off colour mask. The 'off-ness' would be the same 'off-ness' thus would be correctable. The component of the chromagenic dye that is in the film emulsion is called the colour coupler. So what they did is decide to make these colour couplers themselves have a colour. They become coloured colour couplers. The yellow coupler is not coloured (I believe), but the cyan and magenta ones are. When the complete dye is formed in the chromagenic process, the colour of these coloured colour couplers leaves the film and enters the developer. Areas that have no colour image to form, however, remain this coloured coupler colour, even after fixing. So, the ENTIRE negatige is colour masked or 'biased' in this way and to the same extent.
That's how the colour mask of colour neg film works in my understanding.
hope that is of some interest,
richard

[carllooper]

That's how the colour mask of colour neg film works in my understanding. hope that is of some interest,
richard

you bet. Thanks heaps Richard. That has cleared up a considerable number of uncertainties I had.
cheers
Carl

[S8 Booster]

possibly not exactly what you look for but it works well - tried it:

http://www.photosolve.com/main/resource ... index.html

shoot/8

[carllooper]

possibly not exactly what you look for but it works well - tried it:

http://www.photosolve.com/main/resource ... index.html

shoot/8

Thanks for that. It's a good article and provides for how you might otherwise digitally correct for the orange. An even better idea, as others have said, is to introduce a physical blue-cyan filter (eg. an 80A, or CC50B + CC50C) during transfer, or a light mixer to do similar. The reason being that without such you would be sacrificing a significant amount of bandwidth for the green blue channels of the transfer (up to two stops, or in digital parlance: 2 bits) - there being so much orange, the digital sensor squashes the green-blue signal (passed by the neg), or after inversion what is otherwise the orange signal, it will have squashed. So by physically neutralising the orange during transfer it means you have a much stronger digital signal across all three colour channels of such, and therefore much more room to entertain any additional creative tweaks with the result.

But I was actually more interested in the reasons for the orange in the first place, there being a big hole in my knowledge on such matters. Richards exposition is great. I remain in awe at how photo-alchemists have ever invented photography in the first place.

Carl

Demonstrates use of CC filters:
http://www.youtube.com/watch?v=-fNH3yX58Hk

[carllooper]

Following Richard's excellent leads I was able to pursue more details and close a number of conundrums in my understanding of the colour separation process.

The main insight provided by Richard (if only to me) is that any curve representing the sensitivity of silver halide, with respect to the colours of light, would show that it cuts out for any wavelengths longer than that of blue light, ie. the silver halide, on it's own, is blind to red and green. So in terms of the design of film, silver halide would need to be combined with something else if it's to be active in sensing more than just blue light.

But this also means the top most layer of the film, which will encode blue light (in a yellow-dye forming layer), avoids being polluted by any green-red signal by simply being made up of silver halide in it's native capacity as a blue only sensitive agent. The yellow image that forms in that top most layer becomes an accurate analogue of the blue-only light, and without much effort. The universe is obliging here.

But with silver halide blind in relation to green-red light (yellow light) one has the problem of how to make the film sensitive to such. So the second biggest insight (for me) is what Richard covered on "sensitizers". A search on "silver halide colour sensitizers" pulled up a good deal of useful information.

The biggest insight for me here is that the dyes are not just there for creating an image (as I had always assumed) but play the crucial role of sensitizing the silver halide to green-red light. In other words they are also sensitizers. A text I unearthed on photo-chemistry puts it the other way (of course):

"Dyes are used not only for the sensitization of photographic materials, but also for producing images" (1)

So the performing aspect of dyes is not just during projection, (projecting a colour image) but during exposure+development as well: the sensing of colour.

In their sensing role the appropriate dye absorbs red or green photons and transfers the energy of such (the effect of such) to the silver halide, acting as a kind of colour convertor for the silver halide. One can think of the dye-silver halide combo as transferring the effect of red or green light (on the dye) into what would otherwise be the effect of blue light (on silver halide).

Between the photo and the graphy of photo-graphy is such alchemy.

Carl

(1) http://books.google.com.au/books?id=0Yn ... er&f=false

[Tscan]

I've never owned one, but a good thing to look at would be a still film scanner. They were real common before digital cameras took over for scanning slides or negatives. Did they use a filter or software to correct the image? Or both? The result would have to be an acceptable inverted image, but the process was always something that was mostly taken for granted. What was the process? Can it be applied to scanning MP negative film? Andres mentioned a cyan filter on his machine... I wonder if a cyan gell could be placed in the light path on the Retro8 scanner? I wanted to experiment with blue filters on my WP setup, but didn't have enough head room on available light. The 80A cut 2 stops and I was already using a 1/250sec shutter speed to reduce rolling shutter effect. The Retro8 is using a bright LED and shouldn't have any rolling shutter issues, so a color correcting filter shouldn't be a big problem for light loss.

[carllooper]

I've never owned one, but a good thing to look at would be a still film scanner. They were real common before digital cameras took over for scanning slides or negatives. Did they use a filter or software to correct the image? Or both? The result would have to be an acceptable inverted image, but the process was always something that was mostly taken for granted. What was the process? Can it be applied to scanning MP negative film? Andres mentioned a cyan filter on his machine... I wonder if a cyan gell could be placed in the light path on the Retro8 scanner? I wanted to experiment with blue filters on my WP setup, but didn't have enough head room on available light. The 80A cut 2 stops and I was already using a 1/250sec shutter speed to reduce rolling shutter effect. The Retro8 is using a bright LED and shouldn't have any rolling shutter issues, so a color correcting filter shouldn't be a big problem for light loss.

One thing I recollect Roger saying is that he was going to adjust the Retro 8 to accomodate underexposed film, which implies he previously had the exposure time locked off (a one light print as one might say). I don't know if that means he's going to make available some level control for such (or otherwise automate such) but either way it does suggest that an 80A in the light path might not be too incompatible with whatever adjustments he is making.

Now my understanding of white LEDs (which I've used myself in some tests) has a peculiar spectral response. Here's a curve I found for white LEDs - not sure if all are like this but I imagine they might be:



There's a sharp peak on the left there (blue) and a broader (better) peak centred around 550nm (between green and yellow) but a sharp dip in there around 500nm, between cyan and green.

Apart from the sharp dip this curve suggests the white LED would be doing part of the job for you already. From what I can tell that peak on the left is in more or less the same location you'd want your filter pass to be.

The dip between the blue-cyan and the green-yellow, with respect to the right peak, is not too deep so it won't be too bad, but it should have a bit of an affect on the green channel and a touch on the blue. Basically any cyan-green colour in the neg will be darker than it should, on the sensor, or reciprically, any corresponding red-magenta in the inverted result will be brighter than it should be (would stick out a bit). One could use a digital filter calibrated to tweak just that particular part of the spectrum.

The drop off to the left of the left peak is disconcerting. Magentas in the neg will lose bandwidth, or greens in the inverted become noisy and flat. The drop off to the right means as any colour approaches cyan in the inverted it too will tend to become noisier and flat.

It's not as ideal as it might be, the white LED. One should really have some sort of LED lightbox mixer like Franks idea. One can certainly write up a digital filter to counteract the white LED. There's no problem there colour wise. But the drop offs say there will be an increase in noise and a flat look, in the inverted result, for bright greens and cyans, and to a lesser extent for any bright red-magentas.

Carl

[Tscan]

I think he was talking about actually having a variable light source... aside from the camera exposure. I think your right, The LED may have a already blue bias to it and should do some of the work. I know he mentioned future software that could handle neg, but don't know if it's for sure. Personally, I don't think the format will survive much longer with only negatives available. The added costs of getting a decent scan is just too high to justify shooting it. Ironically, aside from costs, modern negatives in 50, 200, and 500 speeds makes the format more viable than ever before. The question of a DIY method of working with these stocks without pro studio costs will make or break the format. Roger says most of his customers are scanning old reversal footage... but I think the Retro8 scanner could be what replaces the projector for many of the people still shooting S8.

[carllooper]

What would be cool is a return of contact printing.

I had a contact print made in the early eighties of a Kodachrome film I shot in '79. The film was called "Futility". Its still one of the best films I've ever made - only because I was teenager at the time - it's full of angst and surrealist terror, strictly tempered by an Ozu-esque or Kubrickian photographic gaze. I still have the original film and the guy I made the film with still has the contact print (including mag sound track). The print was exceptional - just the faintest sense of any increase in contrast and grain.

But the thought of putting this film past an LED onto pillar boxed HD fills me with no excitement whatsoever. But that's just the filmmaker in me. I'm not Roger's market. I'd rather project the film on a projector to a small audience, so one can actually feel it, hear it running the projector. Experience it. Or spend some time blowing it up to 16mm, or doing so via a well made digital intermediate. Or spend some time enhancing it in the digital domain.

The time one puts into film just seems all so wasted if what one sees at the end is just some average copy. Id rather see experimental transfers done, even if they were just weird one's off a wall,from a projector, with phase issues, and then over-manipulated in after effects, or vegas, or avisynth, edited in unfamiliar ways - the experimental part of it being the far more interesting thing.

Anything would be better than just an average copy.

That's why I reckon Roger should make available a kit version - one that can be assembled oneself and work as is, or adapted and modified for experimental ideas - even if it's just to put a correction filter in. Would be great. He wouldn't have to build it! But it's probably not as simple as that. Or would require more work to produce something like that because one might want to ensure it could be assembled without too much skill or special tools - the result otherwise being a flood of questions when someone gets stuck. Still - it'd be great. I'd buy it. I may still buy it - it's just the transport system I'm interested in. I'd pull everything else out, attach my own LED lightbox, camera and write the software. Just wish there was some more technical information on it so I could weigh it up better.

In the meantime am just sticking with the UltraPan8/16mm transfer system I've been working on - with a 70's NAC Analysis projector. It has a remote control for stepping through frames one by one, but the digital electronics are so ancient, the docs on such buried in a Japanese vault for which nobody knows the location or has the key, that what's left to do on my side is make a robot hand (some solenoids) to operate the remote control and then it's done - and then next stop shooting some UltraPan8 film to run through it.

Carl

[Nicholas Kovats]

Carl,

Great stuff and discussion.

You want meat? By that I mean great gobs of technical meat regarding photochemical process and manufacturing? Apparently this is the new biblical reference for all wannabe nitty gritty Kodak photochemical engineers. Apparently self published by a retired Kodak engineer by the name of Robert Shanebrook. Only a few copies left, i.e.

http://www.makingkodakfilm.com/

"The technology required to make photographic film has been a secret held by a few companies. This book explains, for the first time, at this level of detail, how Eastman Kodak Company makes film. Photographic film is one of the most technically sophisticated chemical products that is used in everyday life. Over 200 complex chemical components are coated on to film base in up to 18 unique, precision layers which in total are half the thickness of a human hair.

This insider’s view explains in simple terms how the operation works. It is a picture book with over 25 diagrams and over 130 photographs of Kodak’s production materials and equipment. Sixty percent of the printed surface area is illustrations. The book is unique; nearly all the illustrations were made specifically for this book."

According to the index there are sections on Super 8 and 16mm, i.e. http://www.makingkodakfilm.com/Content.html

[Tscan]

possibly not exactly what you look for but it works well - tried it:

http://www.photosolve.com/main/resource ... index.html

shoot/8

If Kodak could make the same plug in for NLE and their MP stocks that would be perfect. But it would definately have to be independant from your standard color tools otherwise you would have the same bankdwith issues on any further tweeking. Looking at the Retor8, it appears that the camera portion on top can be covered with a gel, maybe tape it down around the sides? The right cyan gel that could slide into place would be an ideal place to start.

[carllooper]

Apparently self published by a retired Kodak engineer by the name of Robert Shanebrook. Only a few copies left, i.e. http://www.makingkodakfilm.com/

Thanks Nicholas. Great find. Am going to order it. I have a reasonably decent library on physics, optics, cinematography, film history/theory, computer programming, cgi, machine vision, philosophy ... but not one single book on the specifics of photo-chemistry.

Carl

[peaceman]

Hey Carl,

I recently wrote a longer post about the colorneg mask, its origin and pitfalls -- unfortunately in German only, but maybe it makes sense with Google Translate?
Here you go:
http://www.filmvorfuehrer.de/topic/1596 ... und-feind/

I also played with a massively complex photoshop procedure to "reverse" the mask exactly as it originated. Bending curves isn't good enough. Results are very good, but this certainly needs dedicated software to scale. Let me know if you are interested in the steps so that I can write them down.

Original Color Neg with Mask:


Just Inverted:


Auto-Levels and Auto-Color in PS:


This all isn't really very good. Even though the Auto-Algorithms in PS are pretty sophistictaed and go beyond Channel-Histogram-Normalization.

This is what I got with my algorithm, without any auto-level-magic: