Color seperation filter used in Drum Scanners

[Ted Baker]

I am interested in construction of drum scanners and any information relating to colour filters used between the light source and PMT. i.e. the beam splitter or dichroic mirrors?

I found a few manuals online, but they don't any specs for the filters used. I am interested in there spectral characteristics specifically.

[Leigh]

Look on EdmundOptics.com

- Leigh

[Drew Wiley]

Well, don't just look at the Edmunds site, but order a textbook from them on this subject. It's fairly complex. I have a huge old OCL (Optical Coatings Lab) catalog that contained a tremendous amount of detailed information on dichroic filters; they also had superb technical assistance over the phone back then, when they were the biggest coating lab in the world. Now all split up, elsewhere. Spectral characteristics shift according to angle of incidence and temperature. It's an entire engineering field in its own right. There is an optical engineer who contributes to this forum as Nodda Duma.

[Ted Baker]

Look on EdmundOptics.com
- Leigh

Thanks I had a look, I tried a few searches but could not find any parts that might for any specific drum scanners.



Well, don't just look at the Edmunds site, but order a textbook from them on this subject. It's fairly complex. I have a huge old OCL (Optical Coatings Lab) catalog that contained a tremendous amount of detailed information on dichroic filters; they also had superb technical assistance over the phone back then, when they were the biggest coating lab in the world. Now all split up, elsewhere. Spectral characteristics shift according to angle of incidence and temperature. It's an entire engineering field in its own right. There is an optical engineer who contributes to this forum as Nodda Duma.

I understand the topic well enough. What I am looking for is some of the details of the ACTUAL filters or dichroic mirrors units used in some of these machines. Any actual specs for real parts are of interest.

[Bruce Watson]

Thanks I had a look, I tried a few searches but could not find any parts that might for any specific drum scanners.

That's understandable since most manufacturing of drum scanners ended in the mid-late 1990s. You are extremely unlikely to find any third party parts for a drum scanner.


I understand the topic well enough. What I am looking for is some of the details of the ACTUAL filters or dichroic mirrors units used in some of these machines. Any actual specs for real parts are of interest.

If that's what you need, you should talk to Aztek. They still make a few drum scanners a year (so rumor says), and have at least some inventory of parts they use to repair Aztek / Howtek drum scanners. It's possible that Dainippon Screen is still viable. Maybe ICG. But Aztek is more or less the "last man standing" in the drum scanner biz.

[Drew Wiley]

What makes you think scanners are all the same in that respect? Anyone seriously in the game would have ordered up a custom coating run to their specifications. You can't just go out and grab a filter. If you want "real parts" you'll probably have to cannibalize them from "real machines", just like everyone else who now has to do to keep these kinds of machines going, or else find someone in the business of providing parts via that same method.

[Ted Baker]

What makes you think scanners are all the same in that respect?

Not necessarily the same, but not that different either as they are all trying to solve the SAME PROBLEMS... There is always something to learn...

Plus you can contrast how they attempted to solve certain problems compared to CCD technologies.

These drum scanners in this thread appear to have very similar spectral responses: https://www.largeformatphotography.info/forum/showthread.php?150020-Scanner-Comparison-2019-Epson-Flatbed-Eversmart-Flatbed-Drum-Scanners

[Drew Wiley]

A former member with a lot of experience is Jim Browning, who still owns patents on big laser printers. He might still be listed. Otherwise, contact him at Digital Mask.

[Pere Casals]

Most drums use an hallogen lamp (of about 55w) and 3 dichroic mirrors to split the 3 beams.

This arrangement has an advantage, the 3 beams are mostly proportional so if the lamp changes a bit its power then the 3 beams will change by similar amount, not exactly the same because a shift in power may change a bit the effective kelvin.

The splits should be similar than those in dichroic color heads of enlargers or those in 3 CCD cameras.

Some manufacturers do state that they trim the band of each beam with additional interference filters, this is a dichroic mirror working as a filter. I guess that the cut-off nm are undisclosed, one may need an spectrometer to sample each channel. I guess that a blind region between 2 channels may help a consistent separation across several films.

IMHO, it is mostly irrelevant what specific spectral nature drums have. In theory, independently from scanner illumination, and for the same film, the output from two scanners can be perfectly matched with a 3D LUT. This is because scene spectral information has ben reduced to a 3 dimensional space.

Color layers of a film cannot have arbitrary shapes, for each silver level after first developer there is a resulting spectral shape in that layer(after color developer), this is what allows a 3D LUT to contain all color matching information between two systems.

[Drew Wiley]

Specific coatings can be employed to fine-tune such things, Pere. And then you've got the issue of patents. Workarounds in some manner or another either had to come into play between different manufacturers, or licensing was involved. And there was unquestionably an evolution within the prime time frame of scanners themselves. And there are various kinds of color film, some from obsolete processes quite different from what you describe, as well as other media, that get scanned as well. You're oversimplifying all this. That's why contacting an actual engineer involved in the specifics would be in order.

[Peter De Smidt]

Karl Hudson might be a good source.

[calebarchie]

Hi Ted,

Can you elaborate on the purpose or requirement / end goal you are chasing here? That way we can further assist you.

The general tonality differences and similarities you are seeing in that thread are inherent to PMT technology itself. Yes, most of the filters were custom made and possibly tuned to better fit the input of PMT (eg cut off bandwidth with broader spectrum PMT - just remember individual PMTs themselves are identical with adjusted gains). But mainly because they needed custom sized pieces of glass to fit in the entire optical assembly, since they were doing that they would of had to specify coating themselves regardless. The only ones I think of who may not have done this are Isomet, they used most off-the-shelf parts possible to keeps costs down (yet gained rep as best RGB scanners) unfortunately there is also no resources on them in general, much less technical and part numbers.

I guess what I am trying to say is don't think too much into it; most drum scanners if properly calibrated, tuned and maintained will give similar output regarding colour / tonality / resolution of ADC/type / PMT the rest is design, eg optics and mechanics. There are many, many CCD based scanners that used dichroics in similar way or circumstance just due to design of light path / module, a way of turning and filtering the light. Their output still does not match what you see in most PMT based scanners either so it cannot solely be attributed to a mirror or filter.

Perhaps you can find a cheaper, older ccd scanner to pull apart and study instead of potentially wrecking expensive drum scanner. Also, I have seen you interested in cneg conversions so if this is relating to that, just bear in mind at the time these drum scanners were mostly designed for slide and colour separation for prepress.

Bests

[Ted Baker]

Can you elaborate on the purpose or requirement / end goal you are chasing here? That way we can further assist you.


I am back working on my scanning software. What prompted the question was an earlier thread that had a number of drum scans of the same negative, inverted using colorperfect, that seemed to show remarkable similarities at least to me.

In contrast there is a lot of information on DSLR sensors, specifically matrices to convert RAW sensor output to XYZ, so you can work out the spectral characteristics of the sensor.

One of things I am working on involves building a custom light source, for a DSLR. Where the Spectral distribution correctly matches the colour negative materials. Or more correctly the spectral distribution matches the Positive printing materials i.e. (paper/interneg/interpositive).

This removes the need to calibrate the sensor, and at the same I would work out suitable matrices to calibrate other equipment (i.e. a Drum scanner, flatbed etc) to match the ideal measurements.

That is the idea anyway...

I learnt for example many drum scanners use off the shelf PMT's like a 931B Photomultiplier tube. So I just wondered if they used filter equivalent to the recommend Wratten combinations for making colour separations of positive artwork for dye transfer prints or similar. i.e. 47B,99 and 29.

The correct spectral distribution for color negative is different to that of human eye, most sensors in combination with a correcting matrix are design to match the human eye. However if you understand the actual spectral distribution you can better correct it to match the requirements for color negative.

Perhaps that makes some sense :)

[calebarchie]

Hi,

I cannot confirm that claim but due to nature of the filters needing to transmit and reflect they have still have band range (TR, RF) - look up thorlabs gives good idea. It is not complete monochrome cut off like in those filters, this can be seen by people using discrete RGB sources on drum scanners with improved results. In the end, PMTs cannot see colour only photons > electrons > signal (v) so it is up to dichroics to do so. I am not aware of any drum scanners with this kind of source, the problem back then LEDs were probably not bright or efficient enough despite PMT sensitivity. A monochromatic laser source makes the most sense, if an expert user can confirm - I imagine there would coherence complexities but PMTs still strive with those kind of problems.

I think for your project at hand, you would be better served by looking at minilab print systems and colour management regarding those. It is a complete circle system re calibration and mainly focused on c41 if you have any experience with those. I pulled apart a light module for an old konica machine circa 90s, it used an huge array of early individually controlled R , G & B diodes and one of the largest dichroic mirrors I have ever seen. Just goes to show how far LED tech has come. If you still want to try get a filter from Drum scanner and have a spectro on hand the heidelberg would be your best best - no expense spared, completely custom. Ask Karl!

Feel free to PM regarding your project specifics, although being so complex I am not sure it is practical to attempt with so many sensor types if you are aiming for universal system.

[Ted Baker]

I think for your project at hand, you would be better served by looking at minilab print systems and colour management regarding those

Indeed, along with arriscan, Scanity. (35 mm feature film, no expense spared)

At this stage it is just a line of enquiry. So far I have a sort of prototype that replicates the state of the art which would be the Scanity unit. But it is also worth taking a look back in time...

grateful for the help!

[Drew Wiley]

Ted, although I'm certainly not an authority on scanners, I do have good reason to think you're underestimating the complexity of the whole question, particularly with how color dyes interact in such a wide variety of manners, making any silver bullet unrealistic. And I have a fair amount of personal experience using color separations. That's not to say progress is impossible. But there are already people who have spent staggering amounts of time, money, and custom engineering to facilitate such ideas. And what you're describing is a moving target - films and papers frequently change, certainly enough to skew any kind of generic program or filter set. But I applaud your ambition.

[calebarchie]

Hi Ted,

I didn't want to clip your wings but Drew has been more or less direct to the point. I understand you may have found a way to measure or establish the characteristics of a particular sensor and its digital post processing (hence eliminate) and therefor take the next step in the process eg calibration/light. I believe when there are too many options and (potentially unknown) variables the correct word is 'difficult' rather than 'complex' (I have studied entire subjects between those two words).

But if you wish to follow your current line of enquiry especially focussing on the telecine or film-out part of the spectrum, look at early flying spot scanners which used PMTs and dichroic setup. These were just eventually just replaced with CCDs in similar dichroic set up which you might see in the arri (iirc arrilaser was the film recorder aspect). Again, I encourage you to look a simpler minilab systems first; it is a entire full circle system from scan to print output (especially regarding making profiles from reflective materials)

Also coming back to using lasers in drum scanners it just occurred to me that the gausian profile of beam may be a problem with the spot size of the light source. If you read about optical lab printers (and things like the lambda) you can see they shifted towards led technology because of this problem. This transition can seen in hi-end film-out LVT (Rhino/Kodak) vs laser based film recorders (oce lightjet 5000)

Caleb

[Drew Wiley]

Quite a bit can be learned by searching extant patents. An awful lot of hypothetical ways of doing things turn up, even though, for one reason or another, most of them never get to market. Therefore, an idea which one thinks is new might have actually been around the block several times. Of course, a patent prevents you from copying things for commercial purposes. Tweaking a concept strictly for personal use generally flies under the radar.

[calebarchie]

Quite a bit can be learned by searching extant patents. An awful lot of hypothetical ways of doing things turn up, even though, for one reason or another, most of them never get to market. Therefore, an idea which one thinks is new might have actually been around the block several times. Of course, a patent prevents you from copying things for commercial purposes. Tweaking a concept strictly for personal use generally flies under the radar.

I can attest to this, while back we were researching old ICE patents, had to use WebArchive to retrieve anything of any value but what we found was priceless. We were considering darkslide microscopy approaches as an alternate, the patent had clearly already investigated the potential of this and many others and of course we all know what ICE is now they went with IR approach. But this has directed our thinking to something that wasn't even on that long list of solutions just due to technological advances.

Research papers are very good so are old books, I recall suggesting to Ted to read a few a while ago now. I have read one recently which was pretty interesting let me see if I can find it again..
Found it: https://www.zora.uzh.ch/id/eprint/151114/1/FlueckigerEtAl_InvestigationFilmMaterialScannerInteraction_2018_V_1-1c.pdf?fbclid=IwAR1TLrB9_VnlzFV0Hjx47cp0x-SqsbuO4zd4uyZfHn7yPthsepMS0smOFBk

[Ted Baker]

Ted, although I'm certainly not an authority on scanners, I do have good reason to think you're underestimating the complexity of the whole question, particularly with how color dyes interact in such a wide variety of manners, making any silver bullet unrealistic. And I have a fair amount of personal experience using color separations. That's not to say progress is impossible. But there are already people who have spent staggering amounts of time, money, and custom engineering to facilitate such ideas. And what you're describing is a moving target - films and papers frequently change, certainly enough to skew any kind of generic program or filter set. But I applaud your ambition.

Not really IF you isolate the problem. The interaction between color negative and the color positive/inteneg/pos is remarkably consistent. i.e. nearly any color paper will work with color negative. Actually what I am interested is has been done and is being done. I am just interested in how it can be done using the equipment I or others could afford. The correct points to measure color negative are around 690nm, 540Nm and 450nm. There is blind spot around 590Nm which is why you can use a safelight with positive/interneg/positive materials. These wavelengths aren't the same as status-m but not too far either.

Modelling the rest of it, yes is very complex, but there is no need to re-invent it all, nor I am I attempting too.



Research papers are very good so are old books, I recall suggesting to Ted to read a few a while ago now. I have read one recently which was pretty interesting let me see if I can find it again..
Found it: https://www.zora.uzh.ch/id/eprint/151114/1/FlueckigerEtAl_InvestigationFilmMaterialScannerInteraction_2018_V_1-1c.pdf?fbclid=IwAR1TLrB9_VnlzFV0Hjx47cp0x-SqsbuO4zd4uyZfHn7yPthsepMS0smOFBk

Indeed! that is a good paper, which I have read BTW. However it is not an old document, it actually is a summary of the state of the art and very current 2018!
My original question is indeed just research into old technologies to consider how certain problems were solved or not solved. This forum has a lot of expertise with these older technologies, by shaking the tree sometimes usefull bits of information fall out... :)

[calebarchie]

Ted,

Never said it was older text, it is recent research paper which I had read not long ago now. I do not know if you recall, I suggested you to read some older staples on Photrio in a thread where you were experimenting with cineon conversion (although not quite there).

If you want reliable and trustworthy sources of information I would stick to patents, academic texts and older publications less so forums, though sometimes you might get lucky :p

Bests

[Ted Baker]

Ted,

Never said it was older text, it is recent research paper which I had read not long ago now. I do not know if you recall, I suggested you to read some older staples on Photrio in a thread where you were experimenting with cineon conversion (although not quite there).

If you want reliable and trustworthy sources of information I would stick to patents, academic texts and older publications less so forums, though sometimes you might get lucky :p


Sorry if my previous post appeared antagonistic, appreciated those comments you made earlier. I have collected numerous articles and several academic texts, as you suggested.

This BTW visually describes what I am interested in:


https://i.ibb.co/gSNbcQF/Screenshot-from-2019-03-29-10-10-34.png

Its a very good diagram it does not include the colour coupler dye. But if you understand it you see why the colour coupler dye is required.

[calebarchie]

Not at all, I have confusing way of writing also.

I see what you are trying to do, the bayer array just makes your life a lot harder. There is reason why the arri uses a tri mono ccd setup as opposed the cheaper options in that paper. Though it doesn't necessarily mean one is better than the other, just different this is evident in conclusion. We have developed system that is capable of both, however at roughly three times the price point with its own inherent limitations. The point being it is a system and highly integrated yet modular but still a somewhat self-contained system none the less.

This reminds me of that rather rudimentary Lightroom inversion plugin, I can't recall the name but it had for profiles each camera and its sensors particular characteristics - it would just not work that well otherwise and it wasn't doing anything particularly complex in the actual inversion. Can you keep up with every new camera that gets pushed out in the industry? Not so viable, if you have discovered way to streamline this process please do elaborate. PM me if you do not wish to discuss in thread.

[Drew Wiley]

Oh my. Babe in the woods. What is a hypothetical ideal set of spectral sensitivity curves (which even what you posted certainly isn't, though you do note how it's minus the coupler effect), won't help much with the manner different types of color neg films are actually engineered to have a degree of overlap in their dye curves to achieve a certain effect. And they differ, sometimes significantly. Even the exact hue of orange masks differs enough to mess up reproduction is you don't precisely subtract the exact type. If you actually applied a model like you posted, the primary hues would not only be harsh, but the greens would be way off. But sensitometry is important; and if you compare different dye curves between tech sheets for different products, it will help you to scratch your head and ask why they differ. For example, in the make-believe model you posted, acquired from gosh knows where, it looks the green separation is way to the left because some kind of old Vericolor film was in mind, which rendered a "poison green" distinctly contaminated with cyan. Look at the work of Stephen Shore, and how he capitalized on repro flaws in Vericolor L to obtain simultaneous contrast between poison green and pumpkin yellow almost every composition - a dangerous game just a step from the edge of the cliff of outright clash, which he managed masterfully. The application of the same spectral separations with a modern neg film like Ektar, and even more in relation to chromes, would be disastrous. Same goes for the red - it's pushed too far right in a hypothetical attempt to get rid of some of the excessive warm contamination in the yellow-orange category, which these older films had a horrible time resolving, because they were intended to lump all those kind of hues into "pleasing fleshtones" which would appear muddy in a different kind of subject, like landscape situations. But if you could move around both the peak spectral sensitivity of your three primaries, as well as the width of nm capture, that would give you something you could customize to the application. That was possible in dye transfer printing, not only through specific filter choices per film type, but by tweaking the printing dyes to match.

[calebarchie]

Hi Drew, picture is from paper and shouldn't be viewed out of context. Please give it a read, its just a theoretical example to explain a basic concept.

But you are correct, there are many many many many many many variables but the OP looks to be focusing on colour negative which limits it slightly. That is why I strongly suggest he look at minilab systems which had to deal with all this kind of variability with focus on C41, some did well others not so much it was incredibly complex which they had to control somehow. You can learn a lot just by looking at how they attempted to do it, where they excelled and where they fell short. But from my time working in a lab, I just dumbed it down to more or less; crap in = crap out :)

Bests

[Drew Wiley]

Well, it is an interesting paper, but obviously with an emphasis on restoration of older color motion picture films. And that should in itself provide an important clue. More of this kind of research is going on in Hollywood than anywhere else, because that's where the financial incentive is - at least as an industry, if you widen the definition of Hollywood to include a lot of related work being done by certain moving image labs here in the Bay Area. Not terribly helpful, however, since a lot of trade secrets are involved. And the sheer variety of dyes used in the Technicolor process - the majority of which are still literally locked up somewhere as secret - shows what a daunting task is potential involved if one starts from scratch. I'm just a kindergartner at this kind of thing myself, though I have gotten some precise separations by basically updating older all-darkroom sheet film methodology, and can do it precisely in the field too, provided nothing moves between exposures! The exposure range linearity is much better on pan film than digital capture.
Well, a step at a time; that's the best any of us can do. But it's also how long memorable journeys are taken. Nothing ventured, nothing gained.

[calebarchie]

The exposure range linearity is much better on pan film than digital capture.


Now thats a whole other can of worms, what is digital capture? CCD or CMOS? Consumer digital cameras are mostly linear for the sake of it I think.

http://hamamatsu.magnet.fsu.edu/articles/ccdlinearity.html
http://www.harvestimaging.com/pubdocs/213_EI2017.pdf

[Ted Baker]

Not at all, I have confusing way of writing also.

I see what you are trying to do, the bayer array just makes your life a lot harder. There is reason why the arri uses a tri mono ccd setup as opposed the cheaper options in that paper.

The Bayer array makes it hard yes, and the IR cut filter harder still. My prototype is functioning, but not yet a success. :-) I have not modified my DSLR, and hope to avoid that if possible.


Oh my. Babe in the woods. .

Well the purpose of the thread was, to learn about a little about Drum scanners, not to provide a dissertation. It important to understand that many of the variables you discuss are intentionally minimised at the negative/positive interface, which are not designed to viewed by the human eye. This is entirely different to reversal system where this interface does not exist.

On of the strength of the neg/positive system is its suitability to create many generations (copy a copy), in order to do this, the variations at the interface have to minimised.

This is spectral sensitivities of Kodak Endura Paper, complete with notch for the safelight.


https://i.ibb.co/hKtM9G2/Screenshot-from-2019-03-30-11-23-58.png

These are indeed very different to the spectral responses of the human eye, or any device or film for that matter which is either attempting to mimic a human response directly or via some correction (matrix).

[calebarchie]

I think all drew is trying to say is it a bit naive to think any colour negative and any ra-4 paper will work seamlessly, you still got to a lot of work to balance what wp with a dichroic CYM colour head (yes there were additive heads too). Its like if I took my phone and took a photo of a neg on a lightbox and did a basic invert in an app. Sure it actually works, you will get a picture - but will it be any good?

What you have going for you are there is only 2 major players in this space left Kodak and Fuji both do paper and film so work with that. Again, please look at minilabs!

[Drew Wiley]

Dissertation? Hardly. I'm not even remotely qualified to write one. I have built very effective RGB colorheads and use those, so do have at least an appreciation of the inherent complexity of the problem. I'm just pointing out that there are apt to be some icebergs out there - big icebergs, with only 10% obvious above the surface - even before the Titanic launches. What might seem like a new set of ideas, simply isn't. CMY is far easier to engineer when it comes to printing. But RGB laser printers obviously exist too; and it took some pretty serious engineering to get there. In some of this thread, where a particular model was published in relation to vintage movie films, you would have to factor in a philosophical and esthetic question - what are you trying to achieve? - an "ideal" form of reproduction according to some industry standard, or the replication of the actual look of those older materials? Same thing can be said for modern films and papers. What is the look you are after? You can't just select specific dyes like Technicolor did for specially furnished movie sets. Every dang tweak in your spectral sensitivity choices will have an effect. I got pretty good at predicting it with Ciba printing, and had all kinds of filter options. But I'm still on the learning curve with RA4. It has it's many idiosyncrasies too. Getting competent commercial quality RA4 prints is super easy. Getting it to really sing takes a lot of experience. Some of my best results are actually from Portra internegs from masked chrome films - bagging certain subtle hues I haven't seen cooperate in any photographic print medium so far. But Velvia dyes matched to Portra dyes and then RA4 - Nope. Showy colors, easy, but no real dye match. Any fool knows how to saturate color. But look at the neutrals or complex hues. Inkjet is horrible at it unless one endlessly slithers and dithers. Guess I'm frustrated. Why can any competent watercolorist achieve certain complex hues in a matter of seconds, but no photographic process ever invented can? And most color photographers don't even have a clue what I'm talking about; they think color is synonymous with turning the volume up.

[Ted Baker]

Some of my best results are actually from Portra internegs from masked chrome films - bagging certain subtle hues I haven't seen cooperate in any photographic print medium so far. But Velvia dyes matched to Portra dyes and then RA4 - Nope.

No doubt your very skilled, and like other contributors, that is a resource that prompts me to ask questions.

It might be worth considering, while portra is suitable for making internegs from chromes it not suitable for internegatives/positive from original camera negatives, even if you fix the problem with the gamma or at least it is not optimal.

Anyway, perhaps I should just get on and do some more work.:)

[onnect17]

I must be way drunk (or blind). I can’t remember anything in the c41/ra4 space close to a well developed kodachrome/e-6 when it comes to color accuracy.


Sent from my iPhone using Tapatalk

[interneg]

I must be way drunk (or blind). I can’t remember anything in the c41/ra4 space close to a well developed kodachrome/e-6 when it comes to color accuracy.


Sent from my iPhone using Tapatalk

Kodachrome has significant (but pleasing to some) colour inaccuracies - colour neg will always give more accurate colour & negative/ positive processes have a range of significant advantages over reversal ones.

[Bruce Watson]

...colour neg will always give more accurate colour & negative/ positive processes have a range of significant advantages over reversal ones.

Yes. This. Color accuracy is what that orange color correction mask is all about.

One of those other advantages is of greater dynamic range captures. Said another way, color negative films can record a scene with more stops between the black and white points than can a tranny film. Indeed, modern color negative films come close to the dynamic range capture of B&W negative films.

[onnect17]

Wider dynamic range I agree but I feel E6 registers blues cleaner than C-41, IMHO.


Sent from my iPhone using Tapatalk

[Drew Wiley]

More misleading stereotypes, as if color neg films were all the same, or all chrome films were. It depends on the specific hues you're trying to reproduce, and via what medium. For example, I have an old print hanging around the corner made from old Ektachrome 64, which was capable of wonderful blue nuances and blueish sage colors, but due to red contamination of the green-sensitive dye, was incapable of producing a bright saturated yellowish "spring green" unless you jumped through a lot of hoops like masking the separation neg for a dye transfer print with around a 25% exposure to remove that bias. On the other hand, fairly early on, Fujichrome 50 was noted for vivid clean greens, but had a hard time with greens which were in fact kinda muddy to begin with. I've printed old pre-E6 Agfachromes that captured certain earthtones like no other film I've encountered since, and even certain fluorescent lichen colors; but the green repro was mute, almost like it had been deliberately skewed for a Godfather movie. Kodachrome was no silver bullet either, but bagged certain kinds of colors and neutrals in its own wonderful way, which I miss. But that didn't help much, because it was already long extinct in sheet film when I began. I shot a lot of Kodachrome 25 as a kid, and made a few nice little Cibachromes - a wonderful marriage of media; but "little" pretty much summarizes it. But rather than whimpering about all these limitations, there is an aspect to limitation itself which can be cultivated in an intelligent way to produce stunning prints from all these various films and print options. That require appreciating those limitations and not pretending you've got a silver bullet. For quite a number of years I exploited the idiosyncrasies of Cibachrome - no, not loud reds! It was a very idiosyncratic but potential beautiful medium. The significant limitations of C41 films have never particularly appealed to me in my personal work, though I used them for at least some of my occasional portrait work. There are other practitioners of muddy color negs like Meyerowitz who have built entire careers of eloquent work based on the repro shortcomings of good ole Vericolor L. I find Ektar far more appropriate to my own needs; but it too has some significant quirks. I don't know when I'm going to get around to printing up a few sets of black-and-white separation negs onto Fuji RA4 paper; but the negs are already made, and I already have all the necessary pin registration system. It's like having too many flavors to choose from at the local ice cream shop, with all being potentially rewarding if you're patient.

[calebarchie]

Most drum scanners = CMYK separation workflow
Isomet = RGB 'photographic' workflow

Sorry not enough time to discuss more..

[Sasquatchian]

Could you expound?
Most drum scanners = CMYK separation workflow
Isomet = RGB 'photographic' workflow

Sorry not enough time to discuss more..

[Ted Baker]

Could you expound?

I didn't write that post but I am the OP. I took what Calebarchie was saying is that most drum scanners used a "densitometric" approach rather than "colormetric" approach. I.e. for the former measure the amount of CMY dyes in the camera original (transparency) so that you can then reproduce that with CMY dyes on a piece of paper. You could argue that the later colormetric approach is a folly since you are limited to the CMY dyes used in your final print. The spectral response of a drum scanner is much closer to these CMY values, it in no way matches the spectral response of the human eye. Think status-a measurements.

Contrast that with a modern digital camera, where the sensor still doesn't actually match the human eye response (to expensive). Where a matrix is used to give some sort of correct colormetric values, and then in "photoshop" various matrix's are used to vary the output to match your monitor and finally the print.

I am not sure where Isomet drum scanners fit in this timeline exactly though.

I have perhaps oversimplified things but this may convey the gist of it.

This what partly prompted my original post.

[Sasquatchian]

Well, I know that there are three color filters after the splitter and before the PMT's, so my guess is that the ultimate gamut of the scanner is defined by those filters. I had a long conversation with John Pannazo about that twenty years ago but to no logical conclusion. And then, if you use something like Hutcheson Velvia scanner target to profile your scanner, that effectively limits it to the gamut of that film, which is going to still be quite a bit larger than any printing material you can find today. The specific target and especially the software used for the profile seem to have a pretty big influence on how that is all interpreted. The bottom line is that if your scanner can record pretty much everything that is one your film, that's about is good as it's going to get.

While it's true that DSLR's have a much wider initial gamut seemingly than the PMT's of a drum scanner as defined by the sep filters, in the end, all that extra gamut may be useful in capturing saturated colors in the real world, but is largely wasted when using the same camera to make a single shot scan from color film. And you still have to decide how you want to get from wide open color gamut to something you can actually use. When I compare the gamut plots of my 5DSR to my Howtek 8000 in ColorThink 3.0, the Canon eclipses the Howtek by far but both provide exemplary results.

I haven't spoken with Evan for a while so maybe I'll give him a shout out or just make that an excuse to drive to Orange County and hang out with him and pick his brain on the subject. Maybe in the fall.

[Pere Casals]

Well, I know that there are three color filters after the splitter and before the PMT's, so my guess is that the ultimate gamut of the scanner is defined by those filters.

IMHO, not at all.

The ultimate gamut is defined by the spectral responses (r, g and b) from the particular film we use (or from the dyes on DSLR pixels). The image capture is the critical step.


You can match the "ultimate gamut" of two scanners by a well done calibration procedure, if you use a 3D LUT for that you can get a perfectly matching result, even in the case that the fiters in the scanner are substantially different, you only need a different calibration for each film to equal result from different scanners.

...but you won't be able to make a "general calibration" to match Velvia with Ektar or Portra, because it won't work when spectral nature of light or subject changes.


After image capture you have spectral color information of each spot reduced to 3 values, it can be 3 voltage levels in a sensor or 3 silver density levels after first developer. In the taking we define the ultimate gamut, while the scanning/edition is irrelevant for the ultimate gamut if we have the good software tools.


There is advanced math demonstrating that.


_________________


Scanners are IT8 calibrated, but different scanners may require a 3D LUT calibration for each film for a perfect match.

[Drew Wiley]

I think you're oversimplifying this, Pere. I don't want to get personally involved with discussing specific scanners; but I do know a thing or two about color mapping and filters, and why it's virtually impossible to correctly replicate every kind of hue regardless of the technology involved. "Gamut" in the jargon of computer skills is a bit different than true color gamut of things in nature etc, which cannot always be predictably quantified. It's hard enough just film to film. Film dyes are imperfect. Then there are also real limitations of output, at the printing or repro stage. A skilled craftsman learns how to work within these serious limitations to his advantage. Denying them is a mis-step.

[Pere Casals]

I think you're oversimplifying this, Pere.

Drew, I'm not oversimplifying, it is film or on pixels dyes what makes the simplification. In the taking spectral information disapears, and the spectrum is what had the richeness.

Let me make a reasoning. Say that that an scanner measures the rgb values in an spot in the color film. For that r-g-b combination a film always has the same transmission spectrum, there is a single possible one.

Now imagine that you build a 3D table, it would be a cube with a cell for each r-g-b value, and in each cell you store the transmission spectrum of film having that particular r-g-b scanner reading.

This would allow to map an actual spectrum in the film for each RGB reading. This is something that's not necessary in a workflow... but with a few logic steps it can be demonstrated that always there is a 3D LUT that converts the output from an scanner to the output from another one. This is something that it's well known by proficient colorists having an scientific education, and because of it they say that what is critical is the capture, the rest can be maped.

[Drew Wiley]

Here's just a small clue about something you're missing. The ideal narrowness or broadness of the bandwidth of RGB capture, individually, is not an nm fixed space, but potentially differs not only between films, but within a given film itself. For example, the red segment of Kodachrome ideally needs to be nm broader than that of green or blue. This is because of the specific dyes involved. Old Ektachrome 64 had a lot of red contamination to its greens, so if you want to reproduce that as is, you'd need different capture filtration than if you were hypothetically going to try to correct that kind of response or scan later Ektachromes without the same problem. Fujichromes differ among themselves too, with Velvia being a bit different than the others. In other words, one shoe size does NOT fit all. Scanner programs can be optimized for just so many given film categories, namely, certain ones extant and popular when the scanning software was developed. That's one of the reasons Kodak updated Portra films for sake of scanning parameter commonality; but then Ektar still has idiosyncrasies. Then you have sampling size issues which are related to dye curve shape. The better the scanner and bigger the sample, format-wise, the better the end result; but it's rarely ideal. People like to post colorful images of how good things are, but to a truly trained eye they often look contrived. I am a proficient colorist in my own right, Pere. I taught color matching. Any halfway decent watercolorist can mix in a matter of minutes complex hues which would utterly baffle the entire photographic and scanner industry, no matter how many tens of millions of dollars they've already spent on color R&D. Color mapping goes back to the 1920's. Of course, doing it on graph paper was extremely slow. But here's a question for you: why do inkjet printers use a 4-axis color mapping model, while film uses only a 3-axis model?

[Pere Casals]

why do inkjet printers use a 4-axis color mapping model, while film uses only a 3-axis model?

First, because commercial CMY inks are not enough. Second because it's cheaper to obtain a grey level or an unsaturation level by using black ink.



The ideal narrowness or broadness of the bandwidth of RGB capture, individually, is not an nm fixed space, but potentially differs not only between films

Drew, there is always a 3D LUT that matches results from different scanners for a particular color film, probably the color matching you were teaching did not include 3D LUTs.

But remember, each film requires a different 3D LUT for a perfect match.

[Drew Wiley]

If there's such a thing as a "perfect" match anywhere in this kind of technology, Pere, then go patent it. You won't find it, and clearly don't understand anywhere near all the major variables of the wider topic yet. Certain things always get lost in translation. One attempts to match the film to the scene, then the scanner to the film, then the scan to the printer output. Each of these speaks a different color dialect, and you're trying to juggle it all using yet another language - that of your software and monitor. It's amazing how far all this has come along in recent decades, but it's still far from perfect.

[Pere Casals]

and clearly don't understand anywhere near all the major variables of the wider topic yet.

Drew, do you know what a IT8.7/1 target is ?

Let's guess you know what it is... well, all colors in that table are exactly matched by all scanners with the same RGB values, in fact you can even download a calibration file for each individual target to account for the target manufacturing variations.

Colors that are between individual cells are interpolated, but from cell to cell there is little color distance, so the interpolation is a very, very good match.

If the taget was made with Ektachrome then that target would allow a perfect match between scanners for ektachrome scanning.

For Provia 100F that calibration won't be as perfect, but then you can use a Provia 100F targets, ISO 12641-1, and then you have again a perfect match between scanners for Provia.


If you take the Ektachome calibration and the Provia 100F calibration, then there is a 3D LUT that converts one to the other. You can also calculate a conversion 3D LUT for any color film.



Drew, look, you cannot compare ancient color matching with what a 3D LUT does. 3D LUTs map any source color to any destination color, when scene-light-film spectral natures are removed from the equation a 3D LUT accounts for all.

[Ted Baker]

. The bottom line is that if your scanner can record pretty much everything that is one your film, .

A monochrome sensor will capture it too :-) In the end it is about reducing metameric errors. None of these sensors old or new match the eye. They use one or more colour matching functions in their imaging pipeline. That IS the whole purpose of scanner target you mention. When you save a JPG or just open a RAW file on your PC a matrix or "3D LUT if you like" is used to transform the data to match the spectral output of your display. (modern imaging programs will do this in several steps). In the case of plain JPG it is hard coded to sRGB by your camera when you hit the shutter.

What seems to be the case is that drum scanners are set up to measure CMY dyes and transparency very well, to be printed with CMY dyes on a piece of paper... This should not be a huge surprise. (transparencies at least, not so much colour negative which is a bit different).

As an example the cones in the Human eye that measure the long wave radiation have there peak at about 566nm this is a bit different, from the spectral peak of the red you will find on a drum scanner and also on most if not all DSLRs. Its also different too the spectral peak of a cyan dye, or the red phosphor of a CRT.

Its doesn't seem to make sense trying to reduce metameric errors for colours that CANNOT be displayed using a CMY dye system, but instead make the most of inherit capabilities of CMY dyes. i.e. build a scanner for scanning CMY dyed film for printing CMY dyes on paper.

[Drew Wiley]

Interpolation. Sometimes it tries to pole-jump a ditch, sometimes something too wide to be unrealistic. I'm done with this thread. Too little knowledge of how dyes and inks and so forth actually work, which is never ideally.

[Pere Casals]

Interpolation. Sometimes it tries to pole-jump a ditch, sometimes something too wide to be unrealistic.

194101

This was an IT8 specific for Provia, Astia and Sensia. With it different scanners do deliver same result for those films. Colors in the array are exactly matched, other hues would not be far from calibrated points, so interpolation also gives an excellent match.

Then add what it can be done with 3D LUTs for matching other films.

Drew, IMHO it would be interesting you play with 3D LUTs, in the digital workflow it gives a clear knowledge about how dyes and inks actually work.

[Sasquatchian]

The problem with the idea of exact color matching, and why it always falls flat when dealing in non-scientific images, is that you never ever really want to really match the transparency. That's what we asked for thirty years ago when we were too fucking arrogant to think that either the film or our idea of color filtration could be off at all, when in fact it always was. Since I got into this little drum scanning experiment over twenty years ago what I've found is that isn't a single piece of film that I ever wanted to be scanned exactly like it looked on a light box. I can always make it better either through scanning software adjustments or through post production adjustments in Photoshop on its way to whatever output device it's intended for. And for those outputs - inkjet printers, laser and LED chemical printers, digital and offset lithography printing presses, custom ICC output profiles are plenty sufficient particularly when there are so many variables along the way. And, that IT8 has been drastically improved upon years ago by Don Hutchison with his hand made hand measured targets which are simply the best available. And yes, a single Velvia target works just fine for every E6 film out there and works pretty damned well for Kodachrome which only requires a tweak to take the blue out of the shadows. You can either do that manually at the time of the scan, use Gretag's Edit Module to tweak a copy of the profile or fix in in thirty seconds in Photoshop later on. Take you pick. And the cool thing, is that if you're able to think in the abstract, you can actually use that same transparency profile as a source starting point for color neg scans. Well, at least you can in Trident. You end up embedding that profile and converting to the profile of your choice in Ps.

[Pere Casals]

a single Velvia target works just fine for every E6 film out there

E-6 is easy because the real film is the real reference. Problem is Color Negative film because there is no standard inversion so it's "à chacun son goût".

[Ted Baker]

And, that IT8 has been drastically improved upon years ago by Don Hutchison with his hand made hand measured targets which are simply the best available. And yes, a single Velvia target works just fine for every E6 film out there and works pretty damned well for Kodachrome which only requires a tweak to take the blue out of the shadows.

When you deconstruct what your actually doing here. It is pretty much what happens when you open RAW file from a DSLR on your PC with the exception of the demoscacing. i.e. your calibration applies a tone curve or curves, and either a 3D LUT or a 3x3 matrix to your input data. Of course the actual values vary, because the sensor has different spectral characteristics. I don't use Photoshop so I am not sure how to turn off this initial processing in that tool, but you can use other tools if you want to see a "version" of the RAW photograph before any transformation are added.

If your happy to post your calibration file, which I am guessing in the form of an ICC or ICM file, we could use a tool like ICCtoXML to see what exactly it contains? From your description I can guess what it roughly looks like.

This is all relevant if your interested in building scanning hardware and software.

[Sasquatchian]

I've used at least half a dozen different apps for scanning color negs and the best and easiest I've found is still Trident, the scanning app for Howtek drum scanners. When you're analog printing color neg, it's an iterative process always involving an estimating starting point and then making a visual assessment of the result. Since the result is quite subjective you can have several different but equally correct results, depending on who is evaluating the prints. It takes a printing technician to interpret and decide if the print looks right. Scanning a color neg is pretty much the same except you get immediate feedback during the pre-scan, but what IS required no matter what, is a high end hardware calibrated reference monitor and software that does a proper inversion and lets you adjust end points - y'know - black point with detail, white point with detail values, as well as overall color balance, saturation and gray balance. YOU, the operator must know what you're looking for, as it is subjectively variable - just like making an analog print. And sometimes, you still need to make fine adjustments or local adjustments post scan in Photoshop. You are using your calibrated screen as the intermediate output device, adjusting to make your image look great there and then making all of your tonal and color adjustments, both global and local, with adjustment layers in Ps. From there it's a simple matter to convert a copy to any output device you want to using the appropriate custom icc profiles.

As far as input profiles, another reason the Hutch Target is preferred is that is has an opaque patch that you use instead of the IT8 target patch for 0,0,0 RGB, which is, by definition, the d-max of the film, but if you use that you will invariably end up clipping the deepest blacks in the film, something you only want to do if you're doing it on purpose. I've used at least four different software packages for making scanning input profiles and by far the best has always been the now discontinued ProfileMaker Professional from Gretag, which was superceded by the current X-Rite offerings, which make fabulous output profiles but still not quite as good input profiles. YMMV. Of course you need to keep a computer running Snow Leopard around to run ProfileMaker, but that's not a huge problem either.

[Ted Baker]

I've used at least half a dozen different apps for scanning color negs and the best and easiest I've found is still Trident, the scanning app for Howtek drum scanners. When you're analog printing color neg, it's an iterative process always involving an estimating starting point and then making a visual assessment of the result. Since the result is quite subjective you can have several different but equally correct results, depending on who is evaluating the prints. It takes a printing technician to interpret and decide if the print looks right. Scanning a color neg is pretty much the same except you get immediate feedback during the pre-scan, but what IS required no matter what, is a high end hardware calibrated reference monitor and software that does a proper inversion and lets you adjust end points - y'know - black point with detail, white point with detail values, as well as overall color balance, saturation and gray balance. YOU, the operator must know what you're looking for, as it is subjectively variable - just like making an analog print. And sometimes, you still need to make fine adjustments or local adjustments post scan in Photoshop. You are using your calibrated screen as the intermediate output device, adjusting to make your image look great there and then making all of your tonal and color adjustments, both global and local, with adjustment layers in Ps. From there it's a simple matter to convert a copy to any output device you want to using the appropriate custom icc profiles.

As far as input profiles, another reason the Hutch Target is preferred is that is has an opaque patch that you use instead of the IT8 target patch for 0,0,0 RGB, which is, by definition, the d-max of the film, but if you use that you will invariably end up clipping the deepest blacks in the film, something you only want to do if you're doing it on purpose. I've used at least four different software packages for making scanning input profiles and by far the best has always been the now discontinued ProfileMaker Professional from Gretag, which was superceded by the current X-Rite offerings, which make fabulous output profiles but still not quite as good input profiles. YMMV. Of course you need to keep a computer running Snow Leopard around to run ProfileMaker, but that's not a huge problem either.

I spent years printing RA-4 some of it making a living... I also understand the inner working of a lot of the software you discuss so I can describe in technical detail roughly what is happening in each step. I don't have the source code so it will always be a guess. Trident may be the best thing since sliced bread, but is it not a defunct product?

I AM interested in re-inventing the wheel... Hence the reason to look at old wheels... I tried earlier to explain roughly what your doing in terms how that works in a modern equivalent using a DSLR. I would love to access to all this OLD equipment, software etc. Failing that any samples, profiles, etc are all ways worth looking at.

Actually this thread describes some of what I was working on if your are interested:

https://www.largeformatphotography.info/forum/showthread.php?147672-scantools-a-few-tools-to-help-with-film-scanning/page1

[Sasquatchian]

Colorbyte stopped development of Trident twenty years ago. I spent a fair amount of time to convince them not to stop but there was no database of Howtek users to hit up for upgrade fees to move it to OSX and very little incentive with the impending demise of large scale film usage. I have no idea what they are doing behind the scenes in their color neg processing but what it is, it works well. You could try giving Panazzo a shout over in Florida and see if he still wants to talk about it. My guess is no. But he'll still sell you a dongle for a grand. ha. I'm pretty sure you can still find a copy of ProfileMaker out there. I'd try and pick up that along with a dongle and start playing with it. Chromix in Seattle would be a good place to start along with their excellent ColorThink Pro.