Digital Negative - Curve creation using spectrophotometer and not scanner

Hi Photo.NET Members,

I wonder if anyone here could help me out here to devise a method of linearising digital negative curves without the use of a scanner? I plan to use this to produce silver gelatine prints and who knows - maybe carbon/platinum prints later on.

I feel that the use of a scanner to assist in curve creation inserts a number of variables which may lead to problems further down the road - and I also don't have a decent scanner with the dynamic range to even produce a 21 step wedge which I could use....

What I do have is a SpydePrint spectrophotometer which allows me to read the values of the printed step wedge providing both L*A*B and density values. What I cannot work out is how I go about plotting this information against the curve. My concern here is that the maximum black/white point readings from the contact printed step wedge will not be 100% black not 0% white - thus throwing off all the other values in the process. All the documents I have been able to find discuss using the levels tool to set the black & white points before and readings take place - so that the scanned (analogue) step wedge dMax/dMin values correctly match that of the original (digital) step wedge. This makes good sense in my mind as the dMax/DMin of the actual contact printed step wedge will likely never be 100% black or 0% white.

I am completely new to this digital negative concept and have yet to print anything (my darkroom is still be put together). In the meantime I am trying to prepare and standardise my workflow. Reading the contact printed step wedge using the SpyderPrint seems like a far more accurate way of doing this instead of using a scanner. What do other people think? Ideally I would like to be able to read the values straight off the contact printed step wedge with the SpyderPrint and proceed with the curve creation without having to scan this.

Also, I cannot use any RIP based software as I plan to use a canon for this process.

Please could someone help a man out here? I think it would be great if this could be figured out and documented so others could benefit from a system which negates the need for a scanner and the variables that it may introduce.

Thanks very much to everyone in advance.

Karl

What I do have is a SpydePrint spectrophotometer which allows me to read the values of the printed step wedge providing both L*A*B and density values. What I cannot work out is how I go about plotting this information against the curve.

Hi, this is the gist of the problem. I've worked quite a bit, over the years, with measuring printer responses and tailoring tonal responses. But I can't really find a question in your post that can be answered directly - it's too vague.

If you wanted to characterize your paper printing, you would ideally measure the FILM density, then you could essentially plot film density vs paper print density (technically you would want to use the "brightness" of light coming through the the negative, essentially pick some arbitrary starting value, then "subtract" the film density from that). The result would be essentially the same thing as published "characteristic curves" for paper. But I'm guessing that you don't have a way to measure film density?

I don't know how you plan to make the digital negatives, but they will probably be subject to variations also. (If you use something like a film recorder to expose film, then the film stock and development both affect the result.) Or if it's something like an inkjet printer, setup options, etc., will affect it.

I'll be glad to make suggestions, but you need to be more specific about what you actually want, or at least how you think you're gonna do it. For example, do you anticipate taking a digital camera to a,scenic location, then photographing it? Then that is the basis to print from? Even something this seemingly simple is subject to interpretation - read up on Ansel Adams' zone system, for example. It was loosely a system to measure an original scene, then getting the important (to you) tones to be mapped onto a much more limited paper.

Hi Bill,

Thanks very much for the response. Sorry if I am coming across a tad vague - this is a new area to me and I still have much to learn here. :) That is something of an understatement...

I suppose I am trying to characterise the whole process - paper, negative and ink via the linearisation of the curve. All the processes I have read about talk about printing a step wedge to the transparency film, establishing the baseline exposure to produce dMax on a given paper and then printing/developing the step wedge onto the chose paper (silver, pt/pd etc..). They all then explain how to scan this developed step wedge and before taking any measurements defining the white point as wedge step 1 and black point as wedge step 21 (in order to effectively measure the differences between analogue vs digital wedges this step seems absolutely vital). It is this scanning part that I would like to try and figure out an alternative.

I would like to be able to read the values of the each step as L* (via the lab colour space) and utilise these readings to plot the correction/linearisation curve for the given paper, ink, transparency combination. What I keep getting confused about is how do I handle the black & white points. If I read the wedge black point with a spectrophotometer I will probably get a reading of say L*=5 and a white point of L*=96 (for example). In the digital domain, the first step would be to fine these values as L*=0 and L*=100 respectively. The question hinges on how do I go about normalising the curve with this taken into account?

One thing I have thought "may" work would be to plot the curve using the L* reading taken with the spectrophotometer (0%, 5%, 10%, 20%, 30% - 95%, 100% etc...) including the black & white points of L*=5 and L*=96 (respectively). This would produce a low contrast curve with no dMax or dMin being achieved. Once this is done using the levels tool I could then define the black and white points as L*=0 and L*=100...? Would this work in anyones opinion? The problem here is every print will require the low contrast curve to be applied, then corrected with the levels tool for dMax and dMin to be achieved.

I hope this makes sense... :) My head hurts, the more I read on this the more confused I seem to become. I hope I have explained better and that it does not come across like the ramblings of lunatic :)

Much appreciated,

Karl

I hope I have explained better and that it does not come across like the ramblings of lunatic :)

Hi, well as soon as you start doing it is when all of the things you overlooked start to stick out like a sore thumb.

Let me think about this a little bit.

In the meantime, as a comment, I don't know why you would want to redefine actual paper CIELAB readings from the actual values to range from L* = 0 to 100; why not just keep using the true values for the paper?

I haven't read any of the "instructions" on how to do this, so maybe there are legitimate reasons, I dunno.

What I think might be a better way is for you to define a tonal response that you want, ignoring the actual media that is finally used. Your defined response would be in the form of a complete curve for better control, coming off an "ideal white," then at the last minute corrected for the "real world paper white." Or perhaps no offset is actually needed as you can change print exposure at that time.

A couple of questions: when you talk about "linearisation of the curve," are you actually trying to make it linear in some fashion, or is this just a general term you're using?

And what sort of software do you plan to use? Photoshop? And how would you make the corrections, by putting precalculated numbers into Photoshop "curves?" Or is this sort of thing undetermined yet?

Hi Bill,

The plan is to print the 21 tone step wedge onto transparency which goes from 0% to 100% black in 5% increments and establish my baseline exposure which would make 100% as black as the paper allows with the least exposure possible. From there I would like to begin taking readings and plotting a curve to correct the non-linear way in which this will inevitably print (perhaps silver paper is more linear the pt/pd, salt etc...?). From readings, 50% gray on the step wedge will not print as 50% on the contact sheet - so the curve comes into play here to remap these tones according (say my contact she read 50% gray as 66% - my curve point would have an input=66 and output=50 to correct this). Unfortunately my spectrophotometer only produces reading in LAB and density - so in photoshop I plan to convert to the LAB color space and use L* reading instead of grayscale percentages.

The article below explains the process I am going to go through here - it's just the scanner part of this I would like to try and rehash. Using a scanner to perform these sorts of measurements seems fraught with potential issues (especially since a calibrated scanner does not appear to be required in the first place).

This intrigues me as a valid reason to resurrect my darkroom. I too cannot execute the photographic vision I have today in a totally wet, traditional environment--for the reasons cited above. My 'pipe dream' is conveying an image shot digitally to a Pt/Pd print. Getting a digital image to the enlarger has been the concern. Now the "general" question.

Are there specific printers and ink formulations that are more 'true' in creating transparency density than others? I see the methodology of normalizing the curve for approximation of range on paper. But printers will always believe at the HAL/engine level that they are printing on opaque media. Does this normalization method actually create a transparency using pigment inks that is sufficient for the genre we associate with 'fine art' Pt/Pd wet printing?

From what I understand, pt/pd and some other alternative processes require high contrast/density negatives. This can pose a challenge to inkjet printers but I believe it can be accomplished using colours other than black alone which are more effective at blocking UV. The method above (http://glsmyth.com/articles/creating-the-digital-negative.pdf) describes printing a colour chart to transparency and checking which colour blocks UV most effectively for your given process (Part 3 in the PDF). The negatives are then colorised in photoshop using the determined colour value. This seems like a good way to do it - as if black ink alone were to be used, I "suspect" that the density's required would cause problems in itself (this is based from reading alone). I have seen some great pt/pd prints online which have been derived from digital negatives so I have a lot of hope for this hybrid process.

For my purpose, I plan to use VC/MG paper - so will steer clear of colorised negatives due to the effect they may produce with contrast on these papers. If I later go with graded papers (providing there is any left by that time....) I may well colorise my negatives. For silver prints I understand the contrast and density requirements are nothing like those required for pt/pd etc...

I have read of people using both pigment and dye based inks with equally good results - but most seem to favour dye based inks. Not sure if someone has an opinion on this - I am open to any ideas on this process. Virtually every article I can find on this "method" go about linearising the curve in the same manner - a scanner - and seem to get good results. It just seems going forward on this idea that a scanner really is not the best tool for the job.

If I scan a stouffer step wedge (with no colour correction) and use the levels tool to set the black & white points to values taken from a spectrophotometer (0% and 100%) - you would think that the LAB values between 0% and 100% might resemble something close to their actually LAB values read with a spectrophotometer - but they don't - they are always out - and I have tried five different scanners. The scanner(s) response does not seem to be linear. With this said, I cannot understand why we would then go about basing a correction curve on readings which in no way relate back to the contact printed step wedge that was scanned. Am I overthinking this?

The machine I used to know as a spectrophotometer measured absorption spectra of liquids.

I do remember putting non-liquids in one, and getting results out.

But the idea is to measure absorption as a function of wavelength, which doesn't sound like what you are doing.

Densitometers measure either transmission density or reflection density (or maybe both), possibly

with built-in filters, such as red, green, and blue. I believe that they usually have a wider dynamic range

than spectrophotometers.

Measuring the absorption or transmission spectra of dyes and pigments would be interesting, though.

Hi, I finally dredged through the G L Smyth article and looked at his stepwedge tiff.

I don't fully understand what he's doing, so I must be missing something. As a note for anyone who's not read the article, he apparently is loading any digital image, either direct or from a scanned negative, into Photoshop. He then manipulates the image to his liking, digitally prints a negative image of it to a transparent printer media, then uses that to contact-print onto light-sensitive photo paper. One key to being successfully is to get the Photoshop view to match the resulting contact-print appearance. Smyth is doing this by adjusting a Photoshop "Curves" layer, until values of the scanned print match (roughly) the labeled values on his stepwedge tiff file.

Part of what I don't understand is how they can be made to match - changing the "Curves" should change both the screen view AND the final print. So my conclusion, although I don't see it in his article, is that ONE of these must be done with the "Curves" in force, and the other with the Curves inoperative. Smyth is calling the process of matching "linearisation" of the curve.

Karl's big question is, can he read CIELAB L* values on the print, and work with these rather than the scanned values as "%K?" My guess is that yes, this ought to work, EXCEPT that the labeled steps in the tiff file aren't labeled properly for LAB L* values, at least with an sRGB image. In the midscale they are off by about 5 L* units. (I can look up the right values if you want, but in Photoshop you should be about to just set one of the pixel-reading scales to LAB.)

As a note, Glenn is right that the SpyderPrint unit seems to be a colorimeter rather than a spectrophotometer. But there are spectrophotometers used in a similar manner - they just cost more (the i1 "pro" units are an example). Not that it matters much in this case as spectral data is not needed.

Here's a couple of comments on what I think I would do differently. First, it seems to be a lot of work doing multiple iterations of the test print, then readjusting the Curves in Photoshop each time.

What I would try is to just print a single image with a large number of steps. (One target I've used had steps like so: 0, 5, 10, 15...245, 250, 255; not publicly available, sorry.) Unless I've overlooked something here, which I well may be doing, this would let one make the complete, final Curve in one shot. Whatever values you need, just find a patch that's close, and read it. If need be, interpolate between two patches. Note that making the target image is a fair amount of work, but once you have it...

The second thing is with respect to the printing inks not controlling the photo paper well enough, ie, that a test is needed to find the most effective color. It seems to me that if ink looks dark black to us, but enough light comes through it to fog paper, there's a good likelihood it's light that we can't see (but the paper does), namely UV. (IR probably also comes through, but the papers are not normally sensitive to it. So I would probably try using a UV filter, like Wratten 2b on the light source. If no luck, I'd look for hard-cutting filters farther up the spectrum. Possibly a pair of filters, one above and one below the spectral range we think we're working in.

Anyway, these are just a couple of ideas that might help.