Goldberg Carbon – Harnessing inkjet-based halftone for carbon transfer

Last summer, I wrote a sort of sit-rep on where we stand with hybrid alt. process printing (i.e. alt. process prints like carbon transfer from digital images) and what the in my view promising ways forward might be. There’s one thing I had in mind to test for some time, and yesterday, I finally bit the bullet and tried it on for size. It’s an attempt to overcome some problems with inkjet halftone negatives by means of an internegative stage.

First, the problem. Consider Exhibit A, an inkjet-printed sort-of-halftone negative:

I made this using QTR and a custom curve that only uses the matte black channel on my Epson 3880. The idea behind this is that an inkjet printer is essentially a halftone printer. However, used in the regular way, it’ll gang up several channels, resulting in overlapping dots of varying density. For normal printing, that’s great, because it creates (sort of) smooth tonality. But for a halftone image, it’d be nice to use the raw halftone/binary nature of a single ink channel. So that’s what this curve does.

At a distance, our Exhibit A halftone negative looks decent enough. But when I make a carbon transfer from it, things aren’t that rosy. Here’s what it comes out like:

Not too horrible, you might argue. Until you realize it should in fact look more like this, which is a positive inkjet print of the same image, again using the matte black channel only (so effectively a halftone image):

As you might notice, this version is a lot lighter. In other words, the carbon transfer comes out way too dark. And what you may not notice in the digital image above, but what is painfully present in the physical print, is a huge load of density in the patches and border area that should be pure white. What’s up with that?

I think the main problem manifests itself mostly in the dense areas of the negative. Let’s have a look at a small section of the inkjet negative, taken from the top edge and including a strip of the printed mask area around the patches:

Alright, typical inkjet striation pattern, actually not too bad given that it’s just a single channel and kudos for this 15-year old Epson holding up so well despite periodic abuse under my hands. But if we look a little closer still and bump the brightness a little, the problem becomes apparent:

This is the dark patch at the top, which is supposed to be a solid black. As you can see, in reality, it’s not so solid after all. And while it looks quite alright if this were the final print on a suitable paper, it is intended here to work well as a negative. The solid black patch actually is riddled with holes where the inkjet dots do not align perfectly. These pass through light, and that results in density on the carbon print. Here’s the corresponding area in the carbon print:

Note how the black patch (in the negative) does not result in a perfectly white patch on the carbon transfer. Initially I was fooled into believing this carbon tissue was fogged, until I examined it closer and saw the inkjet dot (actually, hole) pattern in the ‘white’ areas of the print.

This problem of imperfect density/coverage plays out in what are supposed to be the clear whites, but they also affect the other tones. If I take all this together, what happens is a bit like a dramatic case of dot gain. This is why the carbon transfer print looks far too dark. I actually made a transfer curve of it and you can see how it sags very badly:

The green line is the perfect/theoretical 1:1 transfer function. The thin grey line are the actual measurements, which as you can se show some variance. The orange line is an idealized trendline through the actual measurements. The excessive sag in this line would have to be compensated with a dramatic correction curve. This is just a really, really bad starting point for linearization. And this has been one of the key reasons why I never really pursued this particular route.

What could be done to resolve this issue? If only I could modify my inkjet printer so that it does the opposite of what it does now. It should really print transparent dots into an opaque sheet, instead of printing opaque dots onto a transparent sheet. Well, there’s a technology that has been around for some time in the printing industry that does exactly this: laser ablation printing. It’s essentially a high-powered form of laser printing that shoots holes in an opaque layer on purpose-made film (‘DITR’ film). Nice. But totally out of reach of a consumer like myself.

But wait…there’s a possibility. What if I let the inkjet printer do what it does best – which is to make a positive print. Then devise a way to turn that print into a high-quality halftone negative, and use that in turn to make the carbon print from? That’s actually pretty feasible, as I could just contact print a film-based inkjet positive onto a suitable silver gelatin film and then develop that to high contrast. Just the kind of thing I’ve been keeping some x-ray film at hand for!

The first contact print onto x-ray film wasn’t so great because it suffered a bit from halation due to overexposure. I made a second negative, cutting back exposure by 2 stops and increasing development. This got me a usable negative; see above. At least theoretically, exposure and development of the negative don’t matter much, since it’s supposed to be a halftone image anyway. As long as the dense areas fully block all exposure of the final print, it should work. Here’s a close-up of the negative, brightened so the contiguous nature of the black patch can be seen well:

The quality of the contact print looks quite good insofar as the flatbed scanner can resolve to this degree at 2400dpi (which really is a bit too much to ask of it):

This gives a good starting point to make a final print from. In my haste, I did get some stains and smudges onto the negative, but it still worked for a quick test:

This print isn’t too stellar as it does have some fog – I used a carbon tissue I was about to discard as it had some problems, and in the handling process it apparently collected a small bit of fog as well. Still, the print will work to get a feeling for how the curve response is. And at first glance, it’s already apparent that the distribution of tonal values is much more in line with the original than the previous print shown made directly from the inkjet negative:

I measured the whole thing, which involved some issues due to the defects and some issues with strip-scanning the patches. The print really wasn’t good enough for that, but I could still clean up the data and construct a reasonable response curve from it.

The blue line above is the ideal 1:1 transfer curve. The orange line is the cleaned-up, smoothed-out curve based on actual measurements. As said, this was a quick and (literally) dirty test; I could get a lot cleaner result with some more diligent work. Still, the general tendency is clear from even this imperfect print: the response is a whole lot better than the direct print from the inkjet negative shown earlier.

So, move forward then, eh? Well, I guess not. There are reasons why I had been putting this test off are because I just don’t see myself implementing such a workflow, even if it does seem to be promising. There are a few obvious reasons for this:

1: Carbon transfer printing is a complex process to begin with. Adding another step involving internegatives just isn’t what I’m looking for. Keep in mind that if I were to scale this up to color printing, the simplest approach using three-four negatives for a print (CMY or CMYK, no tonal separations) would involve three-four additional steps of exposing and developing internegatives.

2: I print small and will probably keep doing so, but in principle, there’s no very clear limit to the size I could potentially/hypothetically print with carbon transfer from digital/inkjet negatives. That is to say, my Epson 3880 will print up to A2, which is way beyond any size I ever see myself making carbon prints at. However, if I need to introduce a step involving silver halide film, the film size I can get will be the bottleneck/limiting factor. The practical implication is that my prints would be limited to a little smaller than 8×10″ (8×10″ film and some margins to fuss with). Not necessarily a show stopper, but added to #1 it’s also not an attractive proposition.

3: And of course, I’m a cheapskate which doesn’t combine well with using a film in this process that is increasingly difficult and therefore expensive to get hold of. The test I did here was with some leftover Carestream B/RA mammography film. This is single-sided x-ray film, in contrast to the more generally available (and generally significantly cheaper) double-sided film. The latter would be of no use in a halftone application because the fuzzy image created on the backside of the film would create problems. That would have to be stripped or removed, adding yet another processing step, … Moreover, ‘generally available’ is a bit of a generous term looking at the supply situation here in Europe. X-ray film is available quite easily in the US, but European sources have dried up to the best of my knowledge. Importing it from the US or Japan (FujiFilm seems to be the main remaining manufacturer) drives up costs dramatically.

I may at some point continue the experiments a little further, but for now, I’ve proven what I set out to do. I had known about the non-linearity problems with these inkjet negatives for a long time and had been mulling it over from time to time. In my head, I had already figured out the route involving internegatives, expecting that it should help. I’ve now been able to determine that is indeed the case. The question now becomes how to translate this into a practical solution. I’ve not figured out that one, yet, I’m afraid.

Oh, and coincidentally – you may have noticed the peculiar random patch pattern I’ve been testing with. Those of you familiar with digital calibration tools may recognize it. I had used these before as part of the xRite software suite I use with the i1Pro calibration device. However, this weekend I’ve started to explore Argyll CMS, which an open source suite for calibrating and profiling just about anything you can connect to a computer. And while there’s a learning curve involved, so far I think it’s absolutely fantastic! It’s super flexible and I can get things done with this that I wouldn’t know how to even begin with in the xRite domain. I’ll definitely be exploring that further!