Whither hybridia – The challenge of digital output for alternative printing

I’ve never made a secret of my frustration with digital negatives. Yet, they also appeal to me, for the obvious reasons of consistency and as a crucial means to marry digital imagery with analog/alternative printmaking. For something like color carbon, I consider digital negatives the only feasible way to go – but man, are they fickle. In this blog, I’m going to try and list the options, although frankly, I have no solution yet that I feel comfortable with. Maybe getting it off my chest and structuring the whole thing will help some.

For this blog post I’m indebted to Andrew Burns, who has graciously provided lots of information and thoughts about the use of LCD’s and lasers especially for direct imaging. Thank you Andrew, also for proofreading earlier versions, suggesting relevant additions and offering some excellent illustrations for the direct laser exposure and LCD sections!

Contents

This blog is a pretty darn long one, so I’ve sectioned it up. Follow the links below to go directly to each section.

• Inkjet – using inkjet printers to make digital negatives; includes a note on halftone vs. continuous tone images
• Laser – using laser printers instead of inkjet printers
• Imagesetter – negatives on film exposed by pre-press imagesetter machines
• Direct laser exposure – skipping the physical negatives altogether and just blasting a laser onto the print/tissue directly
• LCD – using a liquid crystal display instead of a negative
• DLP – a bit like using an LCD, but using projector technology
• Film recorder – outputting digital images directly onto camera film, or using machines to digitally expose silver halide print materials

Inkjet

First up, there’s inkjet. The obvious route, used by many, and also the only one I’ve explored seriously myself. The advantages are clear: costs are manageable, resolution should be plenty good enough for real-world prints, consumables are easily available and operating a printer doesn’t require much in the way of specialist knowledge. Oh, and the inkjet printer you use for digital negatives will also churn out regular prints, so it’s not just a one-trick-pony space-eater.

But there are many disadvantages as well. Print quality is “on paper” (pun intended!) excellent, but on transparency film and in the real world it leaves a thing or two to be desired. I personally use an Epson 3880, which is a marvelous machine if you use it for what it’s intended to do (i.e. make prints). On transparency film, the notorious ‘pizza wheel’ marks are annoying – they could be overcome, I’m sure, but this comes at the cost of either modifying the printer (which will affect its ability to print on regular media) or accepting a wayward workflow involving taping film onto matte board and running that through the rigid media channel.

Apart from the pizza wheels, there’s the inherent nature of inkjet dots, which just doesn’t make for very appealing negatives in my experience. I’ve got this carbon-on-glass test print sitting on my desk that I feel illustrates the problem aptly:

The top part is an inkjet step wedge of my own design (it’s not linearized), the bottom is a Stouffer T2115. Even at this low magnification, the coarseness of some of the patches of the inkjet wedge is clearly visible – it just looks very ‘grainy’. Note also the darker blotch in the 4th square from the left on the top row – a typical inkjet problem, where some of the ink of the negative managed to release from the negative, leaving a lower density mark. Regular film (silver) negatives are far more robust.

Zoomed in, the contrast is all the more clear:

Note how coarse the inkjet negative patches are compared to the smoothness of the Stouffer. More so, there’s a regular pattern (visible as faint vertical lines) to the inkjet patches, which corresponds to the travel direction of the print head. The inkjet dot pattern isn’t a truly random dither – it’s a regular pattern, and this tends to show up especially clearly in highlights, where the inkjet dots on the negative are closely spaced.

Since tonal rendition of these inkjet negatives is particularly poor in the high density areas (i.e. highlights in the final print), I contemplated flipping things around. Print an inkjet negative, then contact print it onto regular film (e.g. ortho litho or x-ray film) and then use that to make the actual carbon transfer. I in fact did just this for making PCB’s in the past, and in principle the approach works well. However, the requirements for PCB making are very different than for photographic printing, so whether this approach really is much of an improvement for the latter, I don’t know. And it’s a very roundabout way of working, so I’d rather avoid it.

Diversion: halftone vs continuous tone

Then there’s an issue I need to touch upon, which is not necessarily inkjet-related, but more fundamental. It’s the choice between continuous tone and halftone negatives. I’ve written about this before, so I’ll not repeat all of it – moreover, don’t just take my word for it, and pay heed to what Calvin Grier has to say on the matter as well. The summary of a long story is that halftone negatives help to achieve better consistency and are easier to linearize, although this comes at the cost of potentially visible dot patterns.

Coming back to inkjet, that’s in fact part of the problem. Inkjet negatives are halftone negatives by definition, since inkjet cannot actually print continuous tone. It’s always distinct dots with identical optical density (within the same channel/ink). Intermediate tones between paper white and maximum density are rendered by either making the dots bigger or smaller (‘amplitude modulation’, or AM) or by placing identically sized dots closer or further away from each other (‘frequency modulation’, or FM). Inkjet relies mostly on the latter technique if you print in the regular way. The result with digital negatives created this way is along the lines of the example I’ve shown above.

It’s also possible to do amplitude modulation, i.e. dot size variation, with inkjet. I’ve played with this a bit for the earlier blog post I just linked to. Here’s an example image from that article:

As you can see, the tonality is particularly coarse. This is because with amplitude modulation applied to a machine that basically can only deposit equally sized dots, the penalty is resolution. In order to make bigger halftone dots, all we can do is print lots of the same inkjet dots close to each other.

Without going into the math, a relationship emerges between the tonal resolution (how many distinct grey levels can be printed, i.e. how many halftone dot sizes can be made) and spatial resolution (the size of the smallest image constituents, i.e. pixels or dots). In practice, this means that the print head that natively can image 720 dpi (i.e. 720 distinct dots per inch, per ink channel) print head turns into an approximately 45 dpi machine if I want to image 256-greyscale level halftone dots. That’s low-end 1980s newsprint resolution, so evidently not an option.

In theory (i.e., on actual paper…), the frequency modulation of an inkjet system is very attractive. But sadly, the side-effects in terms of linear features and granular noise (interference patterns, basically) make it perform less than stellar in practice.

To work around the undesirable pattern issues, I figured that it might be possible/useful to replace several (perhaps all) of the inks in the printer with matte black. (I might call it the “T-Ford hack” – any color as long as it’s black!) A tool like QTR could then be used to make a curve that uses the all-matte-black channels in tandem. This would achieve more nozzles being used, increasing effective dot density and potentially creating a smoother rendering especially at the higher and lower ends of the tonal scale.

What’s holding me back from trying this at this point is that I’d want to purchase a dedicated printer for this, as I’d like to keep the 3880 for general purpose use as a color printer.

Laser

I never really considered a laser printer as a viable means to make digital negatives. Two things changed my mind. Firstly, the work of Esmee van Zeeventer / Eamy, of which a particularly appealing example can be seen on Instagram. As I understand, she made the negatives for that installation using a generic office laser printer.

The other thing that changed my mind is the issue of halftone negatives and Calvin Grier’s compelling arguments for their usefulness. Like an inkjet printer, a laser printer is a dot machine: it can only image distinct dots of equal density. So it’s inherently a halftone printer, and perhaps this could be exploited for digital negatives.

However, similar problems can be expected as with inkjet. There will likely be artefacts stemming from the way the laser in a typical laserprinter scans across the drum. There’s also conceptual approach to a ‘laser’ printer that uses a linear bar of tiny LED’s (Kyocera’s Ecosys models use this AFAIK) which may or may not offer better performance in this regard.

I really don’t know how well a laser printer might work in practice, but it has two things going for it: firstly, many laser printers (esp. higher end printers) support PostScript, which potentially allows for a lot of control in making halftone dot patterns, including parameters like screen angle. Secondly, the dots made by a laser printer are potentially of better quality than inkjet dots. Inkjet dots aren’t really dots, but a bit like dye clouds on a color film or paper: they’re not very sharply delineated. Laser printer dots at least have the potential to be better defined – whether this is the case in practice, is for me an open question at this point.

On the other hand, one potential drawback with laser printer negatives is the generally limited density. Pigment inkjet easily achieves transmissive densities in excess of 2.1logD using a decent film and ink. I’m not sure whether laser printers manage this, since the thickness of the toner layer they can physically deposit is limited. It’s something I’d have to test, but it likely depends (a lot) on the specific printer and toner used. Sadly, the only laser printer I have at hand here is a positively geriatric HP LaserJet 6P, which surprisingly still works, but nowhere near well enough to even think about trying it for digital negatives.

As with an inkjet printer, with a laser printer both options of AM and FM could be explored. Frequency modulation may work well enough with the dithering algorithm of the printer and its driver, depending on the specific model used. For amplitude modulation, dedicated RIP software can be used, it might be feasible to use PostScript (exploit its capabilities to modify halftone parameters) or it may be feasible to use halftone generators like the one in GIMP (or Photoshop, for that matter) on a file at the printer’s native resolution – in the hopes that the printer driver won’t re-interpret the whole thing and overlay some random FM pattern on top of it.

All this depends highly on the specific printer, what capabilities it has and how the driver works. I’ve done quite a bit of rooting around in laser printer specifications, reviews etc., but information of even the most basic aspects of custom halftone screening is generally missing entirely. Since I don’t really feel like buying a random printer in the hopes of getting something useful, I’m sort of stuck at this one, for now.

Of course, there’s always the real-world resolution limit of 1200dpi for even the best laser printers. With AM halftoning, this limits resolution to around 75lpi (256 grey levels and 0 degree screen angle), which is still kind of crappy. So the printer’s native FM better be darn good…(in which case even 600 dpi would be plenty good enough for me!)

Imagesetter

The obvious next option is the imagesetter, which in a way is a glorified laser printer. In very general terms, an imagesetter uses a laser to expose a light-sensitive film, which is then developed, yielding a film-based image with high density. There are some variants to the theme, such as media that consist of a black coating on a film base, which is then ablated with a laser, so that the film does not need to be developed and is immediately ready for use.

The advantage of this approach is that imagesetters are generally more precise/accurate than office laser printers. In a regular laser printer, a light-sensitive drum is exposed by the laser, which then picks up toner particles and transfers them to a piece of paper. This indirect method has implications for dot shape and introduces all manner of potential issues (striping, striations, poor offset etc.) Also, the density of an imagesetter negative is very high and would (as far as I know) be difficult or even impossible to match with a toner-based laser printer system.

Calvin Grier uses imagesetter negatives with resolutions up to around 20um dot size (which roughly correlates with 1200lpi) and moreover uses multiple negatives per color channel (at least for black, magenta and cyan) so that any coarseness of the halftone pattern is smoothed out between the different layers. Katayoun Dowlatshahi and Tod Gangler also use imagesetter negatives, to the best of my knowledge, and Kees Brandenburg has occasionally used them as well, but generally prints from regular inkjet negatives to my understanding.

All of them, to the best of my knowledge, ‘farm out’ the creation of imagesetter negatives to providers of pre-press services. Calvin Grier used to offer a service where you could send him your files and he would forward them to his partner print shop, but as I understand, he discontinued it because virtually nobody made use of this service. I’ve personally not tried this route, mostly because of the turnaround time it introduces and the fact that I prefer to keep the entire printmaking process within my own control.

The main disadvantage of an imagesetter is that it’s basically out of reach of an amateur (like me) or even small-scale professional printer (like Calvin, Katayoun etc.) There’s a good reason all of them outsource their imagesetter work: these machines are expensive to purchase, it’s hard to justify the cost of consumables for a small-scale user, and operating these machines isn’t as simple as printing on a regular inkjet or laser printer. An imagesetter relies on a RIP (raster image processor) that’s used to convert a continuous tone photograph to a dot patter, and creating these dot patterns is an art in itself. I am aware of some of the basics, and that’s enough to understand that operating my own imagesetter would be the start of a very long learning curve. Not to mention that these machines are being decommissioned left and right, so getting them serviced is increasingly challenging.

Yet, the discrete nature of the dots on a good imagesetter negatives combined with my cursory testing with halftone step wedges makes this option tantalizing. Someone should invent something like an imagesetter, but usable in a home setting.

Direct laser exposure

Enter Andrew Burns, an Kiwi with a mechatronics mind and an interest in photography. He contacted me through this blog back in January, and unfolded his plans to build an x/y laser exposure machine. A laser is used to scan across a surface, exposing individual dots/pixels as it goes. It’s a bit like a regular laser printer, but in those, the laser is actually static and the light is reflected onto the drum through a rotating mirror, while the drum also rotates. In Andrew’s setup, the print is static and only the laser moves.

Important to note here is that Andrew basically just skips the negative. If you have a laser imaging pixels, why use a negative in the first place, after all? Why not just print directly onto the final medium? So he did just that.

Evidently, it’s difficult to get this right. It’s difficult to make a small enough laser dot that constitutes a sufficiently small pixel so that there’s potential for decent resolution. It’s difficult to space the dots evenly, and to get them to touch each other, but without overlapping too much – across the entire print surface, too. It’s difficult to get all this done at a reasonable speed. It’s difficult to rasterize the image so that the final halftone print looks appealing and is free of moiré patterns and other forms of interference, and with good rendition of shadows, highlights and midtones.

Keep in mind that solving all these challenges basically means re-inventing the imagesetter. Andrew suggested to me in an email that instead of doing this, it might be smarter and more feasible to ‘just’ take a second-hand imagesetter and modify it so it’ll directly expose prints/carbon tissue/etc. How feasible this is, I can’t say – it’s probably rather challenging than it looks at first glance (isn’t that always the case…)

If you’re not looking to DIY much and you have a truckload of cash to spare – there’s a ready-made solution that might work. The printing industry has been steadily moving towards CtS (Computer to Screen) technology, and as a result of this, a variety of CtS machines are now available for applications like screen printing. See e.g. Lüscher’s product offering. I did a quick round of Googling to see if I could find a list price for these machines, but it’s all “enquire within”, which generally means “bring a BIG wallet”. You’re probably looking at $50k at the very bottom end. I turned my pants pockets, but nope. No luck.

Anyway, getting this direct laser approach to work turned out to be difficult enough that Andrew so far has moved on to a different approach.

Liquid crystal display

Let’s follow Andrew a little more as he moved on to another idea for direct imaging: using an LCD. Yes, you’re looking at one, presently, but what most people don’t realize is that an LCD actually has things in common with a film-based image. An LCD is essentially a system that can selectively create optical density on a per-pixel basis. So in principle, an LCD could be used in place of a film-based negative.

Andrew has done just that and modified an exposure system for 3D resin prints, which is based on a high-resolution LCD. He has documented his progress in this Photrio thread – which starts with some example images and a few words about the laser exposure system I discussed above.

The LCD sounds ideal – essentially just a digital substitute for a regular film-based negative. But there’s a catch. On a film negative, the actual image layer is on the surface of the film. Therefore, it can be brought into close contact with the print medium (sensitized paper, carbon tissue etc.) for exposure. However, in an LCD, the actual image forming crystals are suspended between thin sheets of glass. These add thickness, and as a result, it’s impossible to ensure perfect contact between the negative and the print medium. The result is a fuzzy image with a lack of definition.

The solution to this is to collimate the light. In a typical light source, the light radiates outwards from the actual source(s). For instance, on the LED array I use for exposure, the light source essentially projects an array of overlapping light cones (think of street lights in the fog!), with the light hitting the print medium from various angles. A collimated light beam consists of light rays that travel perfectly parallel to each other. Andrew’s exposure unit indeed has collimator lenses for this purpose, but the lens array has the drawback of creating an uneven exposure across the print surface.

An obvious workaround is to simply increase distance between the light source and the print/negative sandwich. Think of sunlight – although the light from the sun isn’t necessarily collimated (it radiates outward in all directions from the massive fireball), due to the large distance between the sun and our planet, the light rays from the sun that reach the earth are virtually parallel to each other.

Again, Calvin Grier is one of the people who has run into this and solved the problem – albeit in a rather brute force fashion. His light source consists of an array of powerful COB LEDs, each with its collimator lens, and on top of this, the light source sits across the room from his exposure frame. (You can see the arrangement on his Insta here, with the light source in the foreground and the printing frame against the back wall, and here as well looking into the opposite direction.) The reason Calvin spent so much effort and arrived at a relatively unwieldy setup (lots of power, large distance) is that he works with halftone negatives with very small dot sizes, down to 20um. This is much more demanding on the light source than printing with continuous tone negatives.

The LCD approach effectively works as a continuous tone negative; each individual pixel can be modulated to a large number of densities (I’ve not checked, but I assume it’s 8 bit/256 density levels, but it might be even more). This means that I’m sure that simply increasing the distance between the light source and the LCD/print medium would clear things up. Andrew wanted to keep things more compact, so increasing distance was an unattractive option to him.

He ended up leveraging the flexibility of a digital imaging solution, employing a smart compensation based on a measurement of the unevenness of his light source and overlaying that on the image. It seems to work quite well!

An alternative route towards collimation Andrew has been contemplating is the use of a simple UV laser and use that to scan across the image area. I imagine this should work very well, since laser light is strongly collimated by definition. Especially if the laser is positioned perpendicular to the print, as otherwise, the angle the laser beam forms with the LCD pixels will result in pixel shape aberrations, and that in turn will likely result in a combination of dot gain and reduced exposure intensity. The latter can be compensated for, the former only to an extent. But simply keeping the laser at a large distance from the LCD, or having it move on an x/y stage would seem to be a better approach overall.

On a related note, some people have been experimenting with similar LCD’s in place of the negative in an enlarger. With the high resolution currently attainable with these LCD’s, it’s possible get good print quality even with some enlargement. Combine this with the advances made in recent years on the front of UV enlargers, enabled by the advent of high-power COB UV LEDs, and it might become possible soon (or perhaps already is…?) to print digital images directly from your computer as cyanotypes, Van Dyke Browns, Pt/Pds, carbon transfers etc. OK, there might be a challenge or two in the area of thermal management on the UV enlarger idea, but frankly, that sounds like a manageable challenge.

So, problem solved, right?

I’m not sure. One issue (thanks Andrew for pointing this out) is that the 8-bit greyscale resolution of the LCD may not be sufficient to work with whatever adjustment curves are necessary to linearize the output. Think of inkjet digital negatives: you generally need to apply fairly aggressive/extreme adjustment curves to the negative in order to get a linear greyscale to render faithfully on the final print. When using an 8-bit direct imaging system (such as an LCD), extreme posterization may occur. The very high optical density of the LCD ‘pixels’ does not necessarily help here (although it’s a good thing, because otherwise, long-scale processes like salted paper or DAS carbon would be non-starters!).

This can be worked around by exposing the print not once, but several times, each time intelligently modifying the pixel intensity so that the effective or ‘virtual’ bit depth resolution is multiplied. It’s very similar to what is commonly done with tonal separations in normal printing. Again referring to Calvin Grier: his series about calibration contains a chapter on this, which is worthwhile to read.

However, the main issue is that while the resolution of these LCD’s is definitely good enough for continuous tone with some enlargement, my search is really for something that work well as a halftone negative. The LCD Andrew uses has an effective resolution (on 1:1 enlargement) of 552dpi. If used as a halftone amplitude modulated device, it would become a 34.5lpi print device – kind of low. Dithering could be used, alternatively, so frequency modulation instead. I suppose this might be an option.

The print size would still be limited to the physical size of the LCD (setting aside the still hypothetical enlarger for a minute). In the case of Andrew’s LCD, this is 29.8×16.5cm or about 11.7 x 6.5″. That’s not too bad and roughly the same surface area as an 8×10″ print. However, for dithered/FM halftones, a higher resolution (Andrew suggests 1000ppi +) would be ideal, and the LCD’s that offer this kind of resolution tend to be much smaller.

So maybe. Maybe.

DLP / micro-mirrors

The approaches above are all based on things that have been tried and that I have come across/heard of so far, even if only in an experimental way. Andrew Burns pointed out that I overlooked an obvious (if you think about it) option, which is DLP (Digital Light Processing). This may sound arcane, until you realize that many of the digital projectors (‘beamers’) have worked this way during the last two decades. Moreover, if you go to the cinema, this is the technology that’s likely used to project the movie onto the big screen.

The core of DLP is a micro-mirror array, which is basically a chip that contains a large array of millions of tiny movable mirrors. Using MEMS technology, these little mirrors can be tilted, so that light that is shining onto the mirror can either be projected towards e.g. a lens and projection screen, or it can be deflected away. This effectively creates a pixel that can be turned on or off. By turning the pixel on and off at a very high rate, a seemingly continuous tone can be achieved (similar to how PWM is used to dim LEDs).

In principle, this sort of technology could be used more or less like an LCD for direct imaging of alternative process prints. Perhaps it’s even very suitable since the mirrors would theoretically have a good reflection in the UV band – but I never actually looked into DMD chip specifications for this. Of course, in a projected system, lenses would be involved in focusing the image onto the print, so raw light source power will be a concern just like with the LCD-enlarger concept discussed earlier. In fact, it’s all very similar – except that I can imagine that the DLP approach might deal better with the high UV power levels. An LCD will easily overheat because the dark pixels need to absorb the radiation, while in a DLP system, the unwanted light is simply reflected away onto a surface that can handle the heat. DLP systems are used in cinemas and this is likely one of the key reasons.

So far, I’m not aware of anyone who has modified a DLP projector for UV imaging purposes, although the use of digital projectors in photography in general is quite common. However, that generally relies on unmodified machines projecting regular color images at sensible power levels. Direct UV imaging for alternative prints using these DLP projectors would be a different ballgame.

Film recorders etc.

This one I’ll add mostly as a hypothetical footnote. There still are some film recorders (sometimes referred to as “LVT’s”, for Light Valve Technology) in use – some in commercial labs, some in amateur hands. These are conceptually similar to laser printers and imagesetters in the sense that they shine a tiny, collimated light beam (typically laser) onto regular camera film – color or B&W.

In the heading, I wrote ‘etc.’ because there’s another class of machines with conceptual similarity – the (mostly) laser imagers used to print on silver halide color and sometimes monochrome paper. The most common type is what you find in a typical digital minilab, which tends to expose rolls of color RA4 paper up to 12″/30.5cm wide. Larger systems are in use in specialized labs; think of renowned (but decidedly long in the teeth) systems like Océ LightJet and Durst Lambda.

This technology shares considerable similarity with imagesetters, although the latter are generally black & white machines with particularly high resolution and capable of printing onto fairly large media. Color RA4 imaging systems also print big (some of them, at least), but their resolution is generally more limited and tops out at around 600dpi (with one or two exceptions, if memory serves). Film recorders have higher resolution, but most that are still around are limited to 8×10″or even smaller 4×5″ film.

What all these machines have in common is that they’re not very feasible options for somehow trying to coax some digital negatives from them. They are either virtually absent from the second hand market (film recorders), too expensive and/or too unwieldy for a home or small lab setting, require too specialized knowledge and auxiliary equipment and software to operate – and there’s always the challenge of sourcing suitable film at a reasonable price.

For good quality halftone work, it seems to me that only actual film recorders would be a real option given their resolution (typically around 2000dpi). For continuous tone negatives, hacking a derelict digital minilab to print onto some kind of light-sensitive transparent film would be a mean feat. Someone up for a challenge? Don’t ask me where to start – I wouldn’t know! A few years ago, the obvious route would be to use a material like Fujitrans, which is a color silver halide transparent film. However, I understand that its production was recently discontinued, so that seems like a dead end (and I’m not sure if the optical density was actually high enough for some of the more demanding processes).