Color carbon made feasible, part 1: halftone screens

In my previous blog post, I outlined a couple of tough challenges I met in the color carbon project. These challenges are partly inherent to the carbon transfer process, and partly stem from working with digital negatives, in particular continuous tone negatives. In this blog, I will outline the two main components to a solution to these challenges: halftone screen negatives and sensitizer-incorporated tissue.

To summarize the challenges from the earlier post:

1: Dot gain in both inkjet digital negatives and carbon prints resulting in poor linearization and necessitating drastic correction curves. And in my experience, the more drastic correction curves are, the more prone the process is to disturbances and variations.

2: Highlight breakdown or ‘tonal threshold’ behavior of the carbon printing process, which involves very delicate highlights not rendering reliably. This creates a coarse/abrupt transition from paper white to the lightest printable tone. The problem is inherent to the carbon printing process since it relates directly to the fact that a gelatin layer needs a minimum thickness to survive processing.

3: In sensitizing carbon tissue with dichromate, drying time is crucial and needs to be long enough for the tissue to be sufficiently dry, but not so long that the dark reaction kicks in. Moreover, there are challenges in achieving perfectly even sensitization of carbon tissue when using brushes or rollers. The result is various opportunities for variations in contrast and density creeping into the process, which makes for linearization a shooting match with a moving target.

A more minor issue is that of pigment load; while this needs to be controlled tightly, this doesn’t seem to be a major problem as it’s feasible to make or buy pigment dispersions that are easy to measure with sufficient precision. This leaves the other three challenges as roadblocks that need to be resolved for decent color reproduction to be achievable.

The solution to these challenges is twofold and requires intervention in the negative- and tissue-making stages. In this blog, I’ll go into first part: the nature of the negatives and possibilities for optimizing these. A future blog will go into the tissue-part of the solution.

Continuous tone negatives and their problems

Most publications on digital negatives for carbon transfer or other alt. printing processes aim at replicating silver negatives in a digital fashion: they produce continuous tone negatives. But continuous tone negatives are not the only option, and I think not the best either when it comes to carbon printing. I already mentioned the issues of inkjet dot gain and the problem of the tonal threshold making linearization difficult. But there’s one more issue.

In the Carbon Transfer Printing book by King et al. the authors provide a way for making digital negatives using QuadTone Rip. The essential steps are these:

  1. Determine which ink channel(s) will be used to make a negative. With the Epson 3880 and InkjetMall ConeColor Pro K3 inks I use, this is typically something like Photo Black, Light Black, Light Light Black and/or Yellow. The more inks used, the smoother the end result can be, since there will be less influence of dithering effects.
  2. Establish the exposure time to reach maximum density (dmax). I would say, the time required to hit the required density, since dmax is a bit of a moving target in carbon transfer, and for color work it’s also not a sensible target to begin with. Anyway, establish an exposure time.
  3. Determine how much inkjet ink density is required to just achieve paper-white. QTR makes it relatively easy to adjust the maximum ink density per channel.
  4. Print a step wedge using QTR and then print that using the intended carbon transfer process.
  5. Measure the step wedge densities and input those to QTR’s linearization tools, and build a correction curve. I used an xRite i1Pro photospectrometer and xRite’s ColorPort software to measure densities, but a regular densitometer or even a scanner should do fine, too. (If you happen to use an i1Pro device, I found this process guide quite useful, although I had to improvise a little with QTR’s command line tools.)
Ink pattern calibration page from QTR; this is used to determine which inks are good UV blockers, and it also gives insight into the relative opacity of the different inks. The print on the right is a carbon transfer (ammonium dichromate) showing the blocking power of the different inks.
To determine the required ink density, I made this aid by printing strips at different ink densities using the pre-determined ink channels and taping them together. This helps in approximating the correct density, but after this, iterative testing is still needed to determine the density exactly.

I’ve tried this process and in principle, it works. Sort of. The problem, apart from the ones already mentioned, is twofold. Firstly, the ink density to reach paper white (so the negative dmax) is pretty tricky to establish – a lot of finetuning is required to get it just right. And combined with the tonal threshold problem, this makes it very, very tricky to build a correction curve that accurately renders highlights.

In particular, the problem with the highlights is that a very pronouncedly upswept curve shape is required to get things to straightened out, something like this:

Note the upswept curve shape on the right-hand side.

Which brings me to the other issue: QTR’s linearization tool has trouble handling this kind of curve shape. In fact, the example above should have been even more extreme, but the curve creation tool failed to produce a curve for it, so I had to tame it down a bit – at the cost of linearity in the highlights and upper midtones.

Now, what happens if there’s a slight drift in the ink deposition by the inkjet printer? If InkjetMall decides to slightly reformulate their inks, or have minor batch-to-batch variations? Or if I switch to a different type of transparency film? That’s right – I’ll have to re-determine the correct ink density for the negatives, and potentially redo the linearization. Not to mention that the whole approach is sensitive to more obvious parameters such as minor differences in exposure, sensitization, etc.

Halftone screens and the continuing problems of inkjet

There’s a solution though, and that’s to not print a greyscale anyway. If you just print tiny dots that are all the same density, you can get the same impression of a continuous tone. It’s a matter of either varying the size of the dots (called ‘amplitude modulation‘), or the number of dots per surface area (‘frequency modulation‘) – and of making the dots small enough so that they’re not too visible. Such a dot pattern is called a ‘halftone screen’ and indeed, this is the approach that several color carbon printers appear to follow, such as Calvin Grier.

The benefit of this halftone screen approach is that there’s no more worry about setting the density of the negative. After all, the negative only has to allow the printing of dots with one density, and the area on the negative that’s not exposed/printed at all (i.e. the ‘white dots’ in the final print) just needs to block all UV radiation.

Halftone screen negatives also remove the problem of the tonal threshold, since a delicate highlight will still print as high-density dots – just very small ones, or very few of them spread out at a wide distance from each other.

In theory, this is great. And in practice, it works, too. I tried it out with the hardware I had anyway, the Epson 3880 printer I also used for the continuous tone inkjet negatives.

An inkjet printer actually uses a halftone screen approach already – it sprays tiny dots in a quasi-random pattern, spacing them wider apart for low density tones and putting them closer to one another for high densities. I.e. an inkjet printer primarily (perhaps only; I’ve not looked into this) uses frequency modulation. The way it rasterizes the image is generally controlled by the printer driver; in the printing industry, dedicated software is used for this end, called a ‘raster image processor’ or ‘RIP’.

A relatively low-end and affordable RIP package for desktop use is AccuRIP Emerald. It is intended for screen-printing applications using inkjet (or laser) transparencies.

AccuRip-generated halftone negative printed with an Epson 3880 onto screen printing transparency film.
Close up of the halftone negative shown above. Note the regular dot pattern, but varying dot size – this is an amplitude modulated screen. The square patches

Here’s an example carbon print step wedge I mage with the above shown halftone negative:

Carbon step wedge printed with an AccuRIP Emerald generated digital negative. Ignore the poor scanner color calibration; it’s out of whack, big-time.

When I printed the strip above, it was a revelation. Seriously. I had been messing around with regular (continuous-tone) inkjet negatives for weeks, trying to linearize them, and the rather drastic compensations needed to get them to sort-of work always bugged me a lot. The print above has some issues that I’ll get to in a minute, but is so much more well-behaved than anything I had done before that I immediately knew that I should stop mucking about already and get serious, or get out.

Now for the issues. Look at this gem; it’s the print above, zoomed in a little. Well, a lot – the squares are about 13mm in reality.

A closer view on the step wedge above. Note the raster pattern.

Or even closer:

1200 dpi scan of the rasterized step wedge. The individual dots are clearly visible here. The softness is mostly due to the flatbed scanner as the dots are quite well-defined in the real print.

I tried making some digital negatives with AccuRIP and it seemed that the resolution limit is around 88lpi with the Epson 3880 inkjet; beyond this limit, the lightest tones don’t print anymore due to the individual dots becoming smaller than what the printer driver still recognizes as a printable detail. But 88lpi is pretty darn coarse to my taste. Maybe it would be borderline acceptable for the yellow layer in a carbon transfer, but for any other layer, it’ll just produce newsprint and not a decent photograph in my book. So that’s out, but the experiment was very worthwhile indeed. Dots are the way to go – but I need smaller ones!

AccuRIP makes its own dots, basically, and then uses amplitude modulation to constitute greyscales. Note in the image above that the distance between the dots (or gaps, in the higher densities) remains constant, but the dots get bigger and merge together as the tone gets darker. To print these dots with varying diameters, a substantially higher output resolution is necessary, which inherently limits the dot size I can make with AccuRIP and an inkjet printer.

So why not use the inkjet dots themselves, directly? Well, we can force an inkjet printer into using only one channel – for instance using QTR, but it’s also possible from the Epson driver directly by selecting ‘black only’ (I need to select ‘plain paper’ as a medium for this option to be available).

Inkjet halftone dot screen using QTR. This used only the Photo black channel. This particular one was printed at 720dpi.

I printed a number of step wedges at various resolutions using QTR. See above for an example, which looks quite OK at some distance. However, from a close distance, things are a little less rosy:

Close up of the same step wedge shown above. Note the artefacts (see text below).

Note two things: the dark patch all the way to the left doesn’t show any distinct clear spaces between the individual dots anymore, but it should have, since it’s not the darkest patch. It’s an extreme example of dot gain, and to correct for it, I’d have to apply an adjustment curve with a very steep part. It would in fact be the same kind of adjustment curve to overcome the highlight problems with continuous tone negatives, so that doesn’t help me much.

Secondly, there’s an ugly moiré pattern in the right-side patch, with blank wiggly lines appearing in what should be a more or less even grey patch. In other test negatives, I also got banding problems, which tend to stand out much more in single-channel prints than if several ink channels are used in combination (as Epson intended the printer to be used!) Indeed, Epson’s ‘black only’ setting is supposed to work only with plain paper because of this kind of paper exhibiting severe ink bleeding, which results in any banding or moiré patterns not being visible anymore.

Yes, I did try to print some of those inkjet halftone screen negatives on carbon transfer, but moiré and banding problems of course show up as you’d expect, and linearization is still very poor due to the poor dot control.

My conclusion is that if any kind of decent resolution is required, an inkjet printer just isn’t capable of producing usable halftone screens. It’s OK for very coarse screens, so great for screen printing t-shirts and whathaveyou, but unfit for the kind of prints I’d like to make using color carbon.

A proper solution – or is it?

There is a way to remove the problematic inkjet printer from the equation. Actually, I can think of two. First up is using a laserprinter. They’re not exactly known for their strengths in reproducing photographs, but in this case, we just need good dots, and potentially, a laser printer should be a capable of this. I’ve looked into it a bit, and there are some concerns.

One is that laser printers, like inkjets, generally control their dot screen behavior through a combination of a printer driver and the printer firmware (and hardware) itself. For most desktop laser printers, there’s no possibility to influence things like dot geometry or screen angle. Higher-end printers offer PostScript level 3 compatibility, which does offer this kind of functionality at least from PostScript-capable software (Adobe Photoshop, Illustrator – not GIMP as far as I know). I’ve seen some Xerox printers that probably start at around €700 or so.

Another issue with laser printers is that their resolution limit is generally 1200dpi. That sounds plenty high enough, but keep in mind that when manipulating a halftone screen, we need to create dots of varying sizes (at least when using amplitude modulation). If you want to image for instance 256 discrete tonal values, you’d have to be able to make 256 sizes of dots. Imagine each dot is made up of individual pixels in an 8×8 matrix, this means that a 1200dpi device can only print 1200/8 = 150 compound dots per inch. In halftone screen parlance, this is actually called ‘lines per inch’. So a 1200dpi printer outputting 8-bit tonal resolution will have a theoretical screen resolution limit for amplitude modulated screens of around 150lpi. That’s not too bad – but not all that magnificent either. A 150lpi screen is visible with the naked eye when viewed close, and that idea doesn’t appeal too much to me for my carbon prints.

Perhaps it’s possible to exploit a laser printer’s halftone screening capability just like I tried with inkjet: let the printer driver and hardware handle things autonomously, which will probably result in frequency-modulated artwork being printed. I’ve tried it with my very (very…VERY) old LaserJet 6P, and the results were…horrible. Density was very poor indeed, low resolution…totally unusable. A good laser printer on good transparency material perhaps aided with a density-enhanding spray will probably perform better, and it’s something I intend to test at some point.

Then there’s the proper way of doin it, and I’ve mentioned it in previous blogs a couple of times: an imagesetter. This is a device somewhat similar to a laser printer in that it uses a laser to expose tiny dots. But instead of doing so on an electrostatic drum as happens in a laser printer, an imagesetter exposes a high-contrast silver-halide film that is consequently chemically processed to produce a silver gelatin negative. Imagesetters generally have (1) very high resolution, making them capable of producing very fine halftone screens and (2) offer control over parameters such as dot shape, dot pitch and screen angle. The latter is done through RIP software, like Harlequin, that translates a regular image into a dot pattern that the imagesetter can print.

The advantage of an imagesetter is that the dots it produces are tightly controlled and sharply defined, unlike the inherently fuzzy nature of inkjet dots. They are also of high density, and problems with dot gain are far reduced or perhaps even absent, at least in the film negative itself.

Sounds like the way to go, eh? Well, I haven’t had the chance to try this route yet. does offer an imagesetter service using their local partner, but I’m hesitant to farm out my negative production. I don’t really feel like having to sit around while my negatives are being shipped halfway across a continent, I find the marginal cost quite high (even though Grier’s argumentation on the pricing is transparent and reasonable overall) and, well, I really prefer to do things here, at home, and by myself.

Purchasing and operating an imagesetter would be a theoretical option, but it’s a costly endeavor starting around €7500 or so for the cheapest imagesetter and an old computer with RIP installed, not to mention the cost of film and the hassle of processing it. There’s also the issue of future material availability, as imagesetters are making way for direct-to-plate technology in the printing industry and smaller shops moving to inkjet or other technologies due to exploding maintenance costs and impossible repairs. For all intents and purposes, this appears to be a dead-end technology, and overall not very attractive for an extremely low-volume situation like my in-house darkroom environment.

So there’s a potentially good solution, but it comes with severe hurdles. It’s something I intend to look into, but I’m not sure if it’s going to be worthwhile. Part of it is also that I somehow don’t really like the idea of a halftone screen. It’s irrational and subjective, and as such, I can’t really explain it, but there’s something very fundamental to the approach that repels me.

Concluding this entry, negatives remain a concern. In principle, there are promising avenues, but from where I stand now, I can see no guarantees whatsoever that they will actually work well in practice, and make (economic/hobby) sense. I’ll have to do some more asking around and perhaps some testing here and there. But it’s a tricky proposition altogether.

One thing is clear to me: continuous tone digital negatives are not something I’m going to spend much more time on for now. I’ve spent literally weeks working my way through endless linearization exercises, and my conclusion for now is that it’s a dead-end street riddled with potholes to boot, at least for color carbon. It may work OK for monochrome, although I found the results I got with it so far lackluster and disappointing in comparison with in-camera silver gelatin negatives.

Leave a Reply

Your email address will not be published. Required fields are marked *