Kodacolor Process (1930s) style camera

[holmburgers]

I'm glad you're hooked!

Here's what I'm thinking; the red is likely a result of the inaccuracies in registration. Since the red is present in the projecting filter, it would only take a very minute misregistration to render a red. I bet that if you assessed the presence of red very objectively, you would find that it doesn't correlate to red in the original subject.

The UV-light idea is interesting and historically pertinent, but I think that in Maxwell's case, it was a result of the incredibly long exposure necessary to achieve any kind of image through the red filter onto a color blind emulsion. Since your exposure is basically instantaneous, the red filter will allow very little usable light to an orthochromatic emulsion.

I would like to try using an ortho film behind my screen plates someday, as it should produce some interesting 2-color results.

Hope that helps..!

 

[yulia_s_rey]

thanks holms it does, all my mosaic screen trials are taken thru a yellow filter when I use ortho and I noticed the longer the exposure there was a slightly better color rendition. the registration errors make perfect sense w/ the Kodacolor system as well. today I was able to try my pervious idea out and scanned in a contact print taken thru a 300 dpi r,g,b mosaic in photoshop I made a perfect, or should I say digital, alignment of the mosaic and replaced the tones with red-orange to make a digital-intermediate attempt at a full color photograph, it worked quite well maybe i'll write a freeware code up for other tradigital photographers to play around and call it Orthocolor my goal now is to get a fine enough lenticular screen to complete the Kodacolor camera.

 

[yulia_s_rey]

after a flurry of tests I actually realized that the mesh method I spoke about earlier is not an mosaic color reproduction method, but in-fact a lenticular system! I stumbled upon this when I replaced the original red mesh with a fine steel mesh and put it about 1/16 from the paper/film and then used a half red-orange and half cyan filter over the lens and it worked better than any arrangement I made before. I think the mesh is acting like an array of "pinholes." this would kinda makes my search for a suitable lenticle screen much easier since there are fine metal meshes which a readily available.

 

[holmburgers]

That indeed sounds interesting.

It could also be working on the same principle as parallax-parallelograms, or in other words, the mesh is acting as a barrier-strip, only allowing certain angles and thus certain colors from the filter to hit certain parts of film; which is basically what a lenticular screen does in effect I guess.

I'd love to see some more examples if you've got them handy. So far I'm not certain if I've seen anything that is actually reproducing the colors of nature, but rather the suggestion of color and thus an effect akin to the Rorschach test. That's not to say that they aren't aesthetically pleasing and colorful though.

 

[yulia_s_rey]

the barrier strip is what I'm most familiar with as some 3D tvs work on this principle and was what lead me to starting this side project.

as for the color present in my images as being suggestive rather then having color information I would have to disagree because I get some values when I do color filter separations in my darkroom & in Photoshop (for comparison w/ projected or scanned thru negs).
If I need near true to life color reproduction I'll use Ektachrome or another color stock that had all the years of r&d behind it to reproduce the "colors of nature" as best possible in the photographic medium. Early forms of color screens were made before a truly pan emulsion was available, yet they exhibited a limited color palette, some colors were more accurately reproduced then others, but none-the-less there. Technicolor had its limitations even though it was a 3-pack shot w/ in-camera filters; dp s used to suggest the best colors to the production designer, today in modern day studios there are certain colors and garments to stay away from. I think no additive mosaic screen or lenticular system can really beat the color reproduction of multi-layered emulsions, last around was Polavision/Chrome but it didn't last long.
These additive screens are like a hobby and mainly to capture my family memories in an inexpensive way on film and in color using re-cans of 16mm Tr-X I have leftover from work. IMHO even having the best additive color screen made and emulsion to suit it, even by industrial means, can't beat a chromogenic film or print in color reproduction.

 

[holmburgers]

I'm not talking about needing to get Ektachrome colors, and there's no doubt that modern color films are technological marvels. I completely agree. Using these old color processes is not aimed at obtaining modern results, but rather to see the unique beauty inherent in the older processes.

The term "colors of nature" is just an old term to describe colors that are in some way representative of the original scene, as opposed to hand tinted or whatever.

As history has shown us, remarkable results can be obtained by these methods however. The key is understanding how the color reproduction works and to be as true to theory as possible. No easy feat when you consider the brilliant minds that have tackled this stuff before us.

Alternatively, on the other side of the coin, there have been many misguided inventors who tried to develop color schemes based off the wrong principles, like using the wrong primaries for instance, or not separating out a color from a certain layer of their tri-pack, resulting in degraded hues... this kind of stuff. There are many examples in that Friedman book.

I guess I'm particularly referring to your micro meshes, which to my eye, have yielded no color reproduction and have no backing principle as to why they should, AFAIK. So, I don't mean to sound discouraging, but I just want to have an honest discussion about the results.

Now, in post #18, what are the colors of the original flower? That one looks the best to me, and if I had to guess I'd say the left flower is either yellow or white, and the right flower is red or blue.

 

[yulia_s_rey]

I agree there is no better why to learn about historical processes than a "hands-on approach" about it. luckily, those brilliant minds did the thinking for us and their methods are published and readily available: the way I look at it and believe (seeming the most logical) is the mesh method is akin to shadow masks on a color crts, but again I could be off. my reasoning behind this is that it works best when the mesh was lifted just above the film emulsion after much trail and error (i'm still waiting for the friedman book to arrive as I am a much a newbie to this, most of my experience is how to photograph and light specific stocks or systems and not making them Post #18 is the Psuedo-Kodacolor using Tri-X Pan thru the finest lent. screen I could find. The flower on screen left is are white narcissuses and to the right are red silk carnations.

 

[holmburgers]

Knowing that about the flowers, I'd say you're definitely getting "the colors of nature".

That's awesome you're getting the Friedman book.. That means that at least two people have purchased that book in the last year... maybe in the whole world. Hahah

The thing that throws me about the mesh, is that if it's not very transparent, it would just block light. But since it's nylon, perhaps there is some trasparency, but how much compared to the clear areas? That difference in exposure between the two color separations would result in skewed balance, in a best case scenario. Also, since the green filter is over the whole image, that's going to block a lot of the orange that the mesh might be able to pass, and furthermore cause there to be no discrimination between the colors. Lastly, the ortho emulsion will not record any orange from the mesh.

Now, if the mesh was clear enough and dyed orange, and the spaces between the mesh were green, that might result in a kind of 2-color screen plate that would have the ability to reproduce some color in the original scene, if panchromatic. If using an ortho emulsion, using green & blue instead of orange might also get some semblance of reality.

Those are my thoughts on it.. and I could certainly be missing something; it wouldn't be the first time!

Your job sounds really cool by the way, from what little I've heard about it. I'm green with envy. . .

 

[Struan Gray]

I am, sort of, a third Friedman reader. It's an excellent book. For a while, the chapters which dealt with lenticular colour and screen-and-coloured-aperture photographs were viewable in preview on Google Books. Perhaps for those of you in the US, they still are. I've been searching for a physical copy at a reasonable price, but no luck so far.

Friedman does cover it all. Google searches get confused by the fact that 'screen based color photography' turns up all the methods with a coloured screen directly in front of the film, which includes a lot of fascinating, but irrelevant information. As it happens though, the screen-based technique is exactly the same optics as is used in making halftones with a ruled crossline screen. There is a wealth of literature about that out there, including complete calculations of the intensity distributions at the film plane which take into account diffraction off the lens and the screen - Fresnel diffraction at that.

A screen full of holes or a lenticular array placed in front of the film both place a small image of the aperture behind each hole or lenslet. If your reproduction chain can resolve the individual parts of the aperture's image you can extract them for each point in the scene. Autostereo and lenticular stereo images use the width of the aperture to provide stereo seperation. This kind of screen-based colour, and lenticular colour (including lenticular Kodacolor), use a multicoloured filter at the aperture so that the image of the aperture recorded on film includes miniature colour seperations.

Screens are easier to use: they don't need the spacing to be so exact, or constant, and it's easier to source materials which will work without needing contact between the screen/lens array and the film. Lenticular arrays are more efficient, so exposure times are shorter. Both sorts require the aperture of the lens to be matched to the screen/array and its spacing to the film for optimal results, but for adequate imaging, there's a fair bit of slop.

The colour you get can never be as good as a tri-pack, multilayer film because you can't make a set of good colour separation filters with identical filter factors. You are also, by definition, using the same B+W emulsion for all three colours, rather than individual emulsions optimised for particular spectral ranges and with matched curves. Still, with a long-straight line film like T-max, and a suitable amount of overexposure, you can get reasonable results.

The problem is analogue reconstruction of the colour. You have to align the mask, and insert a filter in the enlarging/projecting lens which complements the taking filter in colour response, and which matches it spatially. My solution, like Yuri's, is to add the colour information digitally to a scan of the B+W film, but that's not going to fly on APUG.

My experiments use a ruled crossline screen made by Polaroid, originally intended for small print shops to make halftones by re-photographing images onto Polaroid material. I have 65, 85 and 120 lpi screens. The screen fits into the space ahead of a 4x5 film holder and can be used with any international back and film holder type. My filter is a Hoya gimmick filter ("Tri-color") with three coloured sectors. The colours are not as pure as proper RGB separation filters, but they are adjusted in strength to have a constant filter factor. I 'mount' the filter by using the iris inside a Verito portrait lens to clamp it in place - the aperture is then about f8.

The method works. I only work on this project occasionally, but my current efforts are in the image processing software to automate the color conversion, and in trying to find a colour filter which will be more convenient to use. Edmund Optics sell filters especially for this purpose (see their Dead Link Removed ), but they're too rich for my tinkering budget. Instead, I'm looking at assembling something similar from gels and mounting them on a frame I can use with my various graphics arts lenses and vintage brass optics which include slots for Waterhouse stops.

For me, the digital step is essential to maintain sanity, which is why I've not been reporting results here on APUG. It's great to see that others are doing similar things though, and I look forward to seeing more results as your work progresses.

 

[holmburgers]

Struan, thanks for joining the conversation,

The digital conversion you're talking about is very interesting. I would like to do a similar thing with my TV-screen-plate photographs, which have a very regular grid of RGB. It seems that by having a scan of the b&w image, a program could very easily assign the right colors once it spotted the pattern. Have you posted anything on DPUG about this? I'm programming inept unfortunately.

 

[Struan Gray]

I wasn't really aware of DPUG - it was still Hybridphoto last time I looked. It seems like a more natural place to talk about the digital steps. I hope the Gods of Moderation will indulge me a little here.

I've written a fair bit of image processing software over the years, so it feels natural to automate things like alignment and registration of the screen array. One thing I need to measure rather than speculate on is how much distortion there is in a typical scan of a typical piece of sheet film. There are techniques to find the actual positions of the projected images of the aperture without having to assume an idealised grid, and they have the advantage that they will also work with stochastic screens - thus avoiding moire - so I'm putting them in now. The only downside I can see is a potential to posterise shadows to pure black, rather than a very dark tint, but that's not exactly a new issue in photography.

One thing that I can say for now is that the technique nicely preserves bokeh. My ultimate interest is ULF colour, and it's encouraging that the method retains the signature look of imaging with a large sheet of film. The only ugliness is with bright fully-out-of-focus highlights, which are full-size images of the aperture - i.e. look like giant tri-color filters superimposed on the scene. The technique doesn't work well for wide angles either, but I think that's partly the bodging I've had to do to try and get the coloured filter on the lens.

I want to do a bit more tinkering before writing the method up as a proper article. Mostly I want to get a feel for how it renders the world, and whether I do actually like that rendering as much as I think I am going to. The ideas are 150 years old, so it's not like I have to rush to patent


PS: I see that there are many, many affordable copies of Friedman on Amazon and Abebooks. When I first looked, admittedly a while ago, the prices were daft.

 

[holmburgers]

DPUG is unfortunately less populated and the depth & breadth of discussions is much less. I don't think you'll be met with any resistance in mentioning these things on APUG, at least not in this thread.

Well, I have learned something about this mesh/grid idea and it makes sense now, given a "barrier strip" and the multi-color filter at the aperture. A given point on the film only sees one of the filter colors, depending where it is in relation to the crossline screen. Is that about right?

I'd love to see an example, but if it's too early, I just hope that you'll work on it and share it with us in the future.

 

[Struan Gray]

I'm very happy to share ideas now, but I'd prefer to discuss measurements and results rather than speculations, and I'll have more of the former once my lastest batch of film comes back from processing (and I find time to scan it). From what I've seen on the ground glass, there's no question that the method works, it's more an issue of how well.

The key fact is that the pinholes in the screen or the lenslets in the lens-array screen have focal lengths which are much, much shorter than the main camera lens - the pinholes by placement, the lenslets by design. They therefore make a little image of the aperture on the film, whose average brightness and colour correspond to the small portion of the scene which would in normal circumstances be recorded on the film at that point. If you adjust the size of the little images of the aperture (through spacing and the main lens' f-number) so that they don't overlap, reconstruction is conceptually easy.

Lenticular images are slightly more complicated because they use only cylindrical lenses. Contrast along the lenslet axis is provided by the normal focussing effect of the main lens. Only across the lenslet is the aperture imaged onto the film - which is why in the Kodacolor system the striped filter has to be aligned with the embossed lenslets on the film.

In my setup the crossline screen creates a 2D array of pinholes, which project little pie chart images of the filter hanging at the aperture stop onto the film. Even with the 120 lpi screen, the pie charts are clearly seen with a normal focussing loupe (and so, should be well-defined in a simple flatbed scan of the film).

 

[holmburgers]

That makes a lot of sense; your description of how it forms an image and makes the separations.

It is indeed very similar to autostereoscopic or integral imaging, in which tiny lenslets are directionally discriminating. In fact, the first test of Lippmann's original proposal was by a Russian (IIRC) who utilized pinholes and imaged a bright lightbulb, since the technology to create the necessary "fly's eye" lenslets was yet to be pioneered.

 

[Struan Gray]

The stereo-colour link was investigated by both Ives and Lippmann. There is a lot of current research into 'plenoptic' or 'lightfield' imaging which uses the same ideas too. Some have added an aperture which varies during the exposure so that you get time-resolution.

In the end you are really just chosing whether to record non-spatial information on separate whole frames, or encoded as a pattern below the spatial resolution limit. With large format, or with high-res digital, you have more film area than you really need for the spatial information, so it becomes possible to bury the other signals in the image. I wouldn't want to try it with 35 mm or smaller - Yuri gets my respect there.

If you haven't seen it, there's a nice review of plenoptic imaging here:

Dead Link Removed

 

[yulia_s_rey]

sorry i've been mia (damned weather forecasts!) i've kinda been out of the loop. Would this be like comparing single chip digi cams vs three chip digi cams (in terms of both Chroma & Luma info) only applying to film?
i'd feel for it to best work for film wouldn't one have to use or make a specific emulsion that has a physically larger surface areas ei: silver chloride in order to counteract the tendency modern emulsions have to bury the color info. Digi Mosiac grids are usually locked pixel-to-pixel with the sensor behind it. explaining why I had some limited success with old Valca 6 Din film (its from the 1960s!) my only reasoning is that of all the stocks i have it re-chlorinates in a very dilute (blue) simple copper (II) chloride bleach which unlike other off-the shelf bleaches are way too harsh for such an old-school emulsion and tend to strip it, indicating that there may be a high presence of silver chloride salts which tend to be big and clumpy. (i could be very off but it seems to have a correlation in numerous trail-and-error runs) If so I would suggest that maybe to make the mesh system work best you need to find a suitable stock. This is a bummer for me as silver chl is very slow and would be impractical to use in motion film as you'd need full sunshine or high power arc lamps to get sync sound speed! If so I may abandon it for 16mm motion and stick to still as I had my best results using 120mm or 35mm. Also could explain why even enlarging paper would work only after tweaking the specific color intensities of the filters to complement the emulsion.

 

[yulia_s_rey]

by silver chl I also mean slver-bromo, etc; "old school" stox...

 

[Struan Gray]

Would this be like comparing single chip digi cams vs three chip digi cams (in terms of both Chroma & Luma info) only applying to film?

That's the basic idea. An exact analogue would equate a three-chip camera to a traditional three-frame B+W trichrome photo, and the single-chip with Bayer pattern to something like Dufay/Autochrome/Paget colour.

The point that has been forgotten in analogue photography is that there are other ways to encode the colour information than by using a coloured screen immediately in front of the film. Oddly enough, the techniques are alive and well in technical digital imaging - I lump them together as 'aperture encoding' - and are an active research topic, both for high-resolution imaging, and for supremely cost-conscious applications like cellphone cameras (where processing is cheaper than hardware).

I know - and respect - APUG's mission statement, but there are things to be learnt from the digimonkeys. Not least, that it can be useful to regard film as an information-recording medium, and not as some kind of mystic mimetic correlative with a deep physical connection to the thing observed.

i'd feel for it to best work for film wouldn't one have to use or make a specific emulsion....

Best would be a fast, fine-grain, panchromatic emulsion. Not easy, especially if you're going to homebrew. I'm using T-max 100 in 4x5, which at normal print sizes gives me a lot of unused space at the bottom end of the spatial scale. With motion film you don't have much real-estate to spare, which is why before tri-pack film took over there were so many attempts to minimise the information encoded (using bichrome instead of trichrome, using alternate frames rather than putting all the colour info on one frame, etc).

The combination of the filter and the crossline screen loses me about five stops of light. With modern 100-400 speed emulsions you end up with 1930s-era film speeds, which seems appropriate


PS: the most recent attempts were all hideously over-exposed. I suspect I had a brain fart with my Sinar shutter. It's been a while ...

 

[yulia_s_rey]

I share a similar outlook to the medium as well, at work we use both film and digital in unison, through a digi-intermediate system. Even sometimes generating 3D animation and then printing to film.
I feel that for film to be the continued medium of choice for professional applications we must be open to new ideas: Take for example "Storarovision," not only it allowed a more suitable format for telecine and digi conversion but also saves a lot of money. Right now the hype is with 3D imaging and that's where I came across this type of technology.

I get an average loss 3.5-4 stops over a range of films, although my tri-filter is homemade. I'm pretty much limited to MP re-cans from work, I found Tungsten balanced respond the best in general. I really wish I could find some more information regarding Kodak's research into emulsion design for Kodacolor Process 1. I really would like to get this onto 16 as I know there has to be some stock out there that would meet the above criteria you mentioned. I see what you mean now by "real-estate" as I thought it more having to do with the actual size of Au crystals and they being physically bigger size in the older days versus the modern smaller-sized crystal being implemented, in which case I was mistaken and there maybe no correlation what-so-over.