Kind of blue – A test with UV LEDs for DAS carbon transfer

So the 400nm LEDs I had been using since October or so worked really nice for dichromate carbon transfer, but I ran into trouble with DAS carbon with them. Some further testing with my bank of trusted Philips BL tubes and a little theoretical exercise suggested that at least some 365nm exposure is needed to get DAS carbon to harden reliably especially in the highlights. So I’ve been testing a bit with 365nm LED options – two in particular that seemed attractive. This is a small report on these tests.

First lets get the obvious option out of the way. I used cheap 400nm LED floodlights previously, so why not do the same for 365nm? I simply didn’t find any product to my liking. Whatever I found was either underpowered or overpriced, basically. I want something in the range of 100W for a unit that exposes 4×5″, so around 300-400W for an 8×10″ unit. I know this is the kind of power level that’s convenient in terms of exposure times with carbon transfer. And evidently, I wanted this in particular to be cheap, since I’m running a test initially to see how the 365 LEDs work in the first place. Yes, the UV tubes worked OK, but the 400nm LEDs didn’t work so well with DAS carbon, so I wasn’t feeling like wasting a lot of money on a LED option that may or may not work. So I was left with no other option but to DIY something.

I make no secret of my preference for cheap electronics components from AliExpress. So that’s where I headed in search of some 365nm LEDs. The one thing not so nice about shopping there for components is that the search function is kind of abysmal – you either get nothing, or you’re swamped with options. And most of the time it’s a bit of both: some relevant ‘hits’ among a massive number of totally different stuff that I’m not sure how they manage to match it to my search terms. Ah well.

In terms of 365nm LEDs, there’s a few options to choose from – but after whittling away the chaff, they boil down to a couple of options (but generally available from a multitude of suppliers). Cost is of course always a factor, so I decided to figure out how I could get the most bang (defined as power in Watts) for my buck (Euro, actually). After some copy-pasting, I could make a plot like this:

Some AliExpress 365nm UV LED options. Vertical axis: € per Watt. Horizontal axis: total Watts per quantity offered.

The vertical axis shows the effective price per Watt based on the nominal power as specified in the ads. The horizontal axis shows the total amount of Watts you purchase in one go within an ad. So for instance, if an ad offers 100 LEDs of 3W each and the total cost of this quantity would be around €170, the total power purchased would be 100 x 3W = 300W at an effective price of around €0.59 per Watt.

Furthermore, there are some colors in there, denoting the categories of LED elements I came across. I excluded large assemblies like floodlights and LED strips – mostly because they are barely offered, and in particular strips don’t generally offer that much bang for your buck at this wavelength, currently. They’re also somewhat inefficient, but that’s a personal gripe I have. I did include a couple of strip-like assemblies, but didn’t end up choosing them since they’re not very economical anyway.

It’s also pretty clear that the price per Watt is all over the place, but trends between roughly €0.20 and € 1.00 per Watt, with some exceptions in both directions.

COB LEDs. These are 100W RGB ones I bought for the color enlarger project, some years ago, but didn’t end up using them.

The grey dots denote high-power COB LEDs, which technically are quite interesting since they are close to a point source. Calvin Grier uses these in his fancy dual wavelength exposure units, together with lenses to focus the light so he can get a nicely collimated beam over a long distance. You’ll also note that the COB LEDs are on the expensive end of the spectrum, all of them being over €0.55 per Watt. COB LEDs come in various sizes and power ratings, ranging roughly from 5W up to 100W for UV units – and even beyond, for regular colors and white.

LED ‘beads’. These are 3W UV 365nm ones. We’ll talk about these later.

One step down in terms of power concentration are beads, which come in sizes of 1W through 5W these days. Currently, for pure 365nm wavelengths, these are an attractively priced option. I got a bag of 100pcs 3W ones and these particular ones came down to around €0.23 per Watt.

Then, there’s a whole slew of smaller SMD devices, ranging roughly from a 3535 form factor up to 6565 or so, with 3535 devices being the more common and affordable ones. Many of these turned out to be significantly more costly than the beads – except for one type that piqued my interest in particular.

3535 SMD LEDs. Interesting little ones, these are, since they are in fact dual wavelength LEDs.

Turns out the cheapest option were 3535SMD LEDs that are dual-wavelength. There are actually two LED emitters within each device: one emitting at 400nm, and another one emitting at 365nm. Apparently, this combination is used for hardening UV nail polish and these LEDs see wide usage in nail studios as a result. Well, I don’t care much about what they’re intended for, but they sounded like an interesting option.

When I was doing initial tests with DAS carbon, one of the things I observed was that I could get decent transfers if I exposed a tissue using two light sources, dividing the exposure time between both of them. That way I determined that e.g. a 10 minute exposure, I could split up into 8 minutes under the cheap & cheerful 400nm LED floodlights and 2 minutes under the bank of UV tubes. This seemed to be the sweet spot in terms of getting decent contrast (using the 400nm LEDs) and sufficient hardening in delicate highlights (with the tubes). Is it possible that I stumbled upon a type of LED that essentially packages this in a ready-made solution? Interesting! Oh yeah, these little 3535SMD dual wavelength thingies worked out at a little less than €0.10 per Watt. Very interesting, indeed.

So I ended up buying some 3W pure 365nm UV beads as well as a bag of ridiculously cheap 1W dual-wavelength 3535 SMD LEDs. Well, very nice, but now this needed to be turned into an actual light source, so I also got myself a 24V power supply and some inductors, resistors and whathaveyou so I could get going. I still had a few pieces of aluminum-core PCB material lying around, so I went ahead and made a PCB for both types of LEDs.

Here’s some snaps of the 3535 version that I took as I was building it:

Close-up of 3535 ‘dual wavelength’ PCB, showing the connector side. This is before mounting the components and while test-fitting the aluminum heatsinks (see black screws).
Top side of the PCB, with two aluminum heat sinks. Design aim was for 100W RMS power on a 16x10cm PCB, so that requires some provision for heat dissipation.
PCB on the hot plate for soldering. It’s just a bunch of LEDs and two connectors.
Finished PCB with components fitted and heat sinks installed.
Test firing the LED channels, one by one, using a bench power supply. The intense UV light does some really funny things with the camera on my phone!

Then I had to do something similar for the beads. I decided to make them pin-compatible and fit for use with the exact same driver, so I only needed to make a single driver unit for A/B testing.

Both finished test light source PCB’s. Dual 365nm + 400nm 3535SMD 1W LEDs on the left, 3W single wavelength 365nm beads on the right.

Both units dissipate a little over 100W. The 3535 board uses 96 LEDs to achieve this, while the beads are more powerful, so I only needed 48 of these and could run them significantly below their maximum rating for the same output power. On the 3535 board, I use strings of 6 LEDs each with two strings (so 12 LEDs total) being put in parallel for a single channel, for a total of 8 channels. The beads are not paralleled and are arranged in 8 channels of 6 LEDs each. Each channel is thus driven with around 500mA of current, or 12W at close to 24V.

Testing the 3535 version at full strength with all LEDs on. A little over 100W – and these are real Watts, not the fairytale Watts sometimes seen in ads.

Now for the driver. I actually wrote about making that PCB, because it was subject of a previous blog. Now you know what that was all about! The driver board is based on a couple of MP24894 LED drivers – the same ones I use in my color enlarger (I had a few leftovers anyway).

100W 8-string LED driver board. At only 4x10cm it’s pretty efficient, and it also wastes very little power at all.

The driver board takes a 24V input from a switching power supply. The MP24894 are buck topology drivers, that use inductors (mounted to the bottom of the PCB) to step down the 24V to whatever the LEDs consume at the current set by a series sense resistor. They’re efficient and when implemented correctly, they are quite stable and reliable. I ended up with these in my enlarger project because they are one of the few types that combine a low price with good documentation and a voltage rating up to 60V, making them fit for fairly high-powered designs.

Finally, I had to somehow mechanically bolt all this together in a usable form. I’m more of an electronics guy than a mechanics guy, so I went for something I could actually wrap my peanut brain around and used some scrap wood, screws, tie wraps and other unfortunate objects lying about the place. Here’s the contraption I built:

Underside of the test setup for the two LED types. I could easily swap out the LED boards with newly made ones if the need arises to test some other LED types.
And the top side. The metal brackets allow this to be mounted to a shelf over my worktop. The 24V power supply to the right; it’s actually a 480W unit, so it has room to spare for future expansions. The little driver board is mounted vertically just to the left of it; the inductors are shown here. The LEDs are underneath the heat sinks and the fans.
Here’s the unit bolted in place and connected to a UV exposure timer (the black and white box on the right). The latter is a bit of an experimental contraption that’s not entirely to my liking just yet, but it works in a way, for now.
Making some light. The little black box below the light source is a UV sensor, since the exposure timer on the right is a test to see if I can get UV integration to work, but there are some minor issues that need to be ironed out for this to work as intended. Currently, I simply use it as a timer without UV integration.

And now, the million dollar question…does it work?

Yes, yes it does. And that’s not a surprise, because from experience with the previous two exposure systems, I had a pretty good idea of the kind of power range to aim for. But there was a surprise in there nonetheless.

My expectation was that the 365nm beads would offer superior performance, especially in terms of exposure speed. After all, DAS carbon is significantly more sensitive to 365nm wavelength light than to 400nm wavelength:

Spectral absorption curve of DAS

Well, my expectation proved to be wrong. It turns out that the pure 365nm wavelength light source is about 1.5 stops slower than the 400nm + 365nm combined light source. That’s right: the LEDs that cost less than half of the 365nm beads offer more than twice the gelatin hardening action in DAS carbon!

How about the whole highlight problem that I set out to fix in the first place? Well, I’ve made 20 or so prints so far, and it seems that the highlight problem is certainly no worse in the cheap dual wavelength LEDs than with the pure 365nm beads. I’d say it’s more or less the same, or perhaps it’s even a little better with the dual wavelength light source, which implies that the amount of 365nm wavelength light in the cheap dual-wavelength option is sufficient to circumvent the issue.

How about some examples? I made a couple of prints with both light sources, aiming at comparable overall densities, so compensating for the difference in speed/efficiency. I used a tissue with the following composition:

% w/v in glop% of dry gelatin weight
Gelatin6.0%100 %
Sugar2.0%33 %
Glycerin 86%0.25%4.17 %
DAS sensitizer0.25%4.17 %
Kremer Pbk70.25%4.17 %
Kremer Pb15:30.005%0.08 %
Kremer Pr1220.16%2.67 %
Kremer Py1540.06%1.0 %
Formulation of the glop of the following two test prints

The glop is a very high contrast one in terms of pigment load – see in particular the >4% pigment to dry gelatin ratio. Note that the pigments used here are dry pigments, hence the low quantities compared to using e.g. India ink as a ‘pigment’ (which is actually mostly water). For instance, a typical and still fairly high-contrast tissue that I use with dichromate will have a dry pigment to gelatin loading of 2% to 3%, with the latter being a decidedly ‘punchy’ tissue that will print negatives in the range of normal grade 1 negatives or thereabouts. We’ll come back to this issue at the end, but first let’s look at how this pans out with the two different light sources.

First up is the 365nm LED bead light source.

Test print using 365nm only light source. Note the flaking of highlights along the edges of the clouds.

Next is a nearly identical print made with the 400nm + 365nm dual wavelength LEDs. Exposure time was less than half that of the print above to achieve the same overall density.

Test print made with 400nm + 365nm dual wavelength LEDs. Still one ‘flaky’ highlight, but less so than in the other print, which is probably due to this print being overall a tad denser.

I haven’t printed any step wedges (yet). For now, I’m mostly concerned with how the light sources ‘feel’ when printing in-camera negatives, which is what I enjoy doing most at this point. Going by this informal test (and a couple dozen more along similar lines not shown here), I note that there’s no big difference in contrast between both light sources, nor does the highlight flaking issue differ all that much between them.

This leaves me with the preliminary conclusion that the dual wavelength LEDs work at least just as well as the pure 365nm ones, but they’re much cheaper and much more efficient / faster. For this reason, I’ve been mostly using the dual wavelength light source over the past week or so in further testing, and it seems to work very well indeed.

Now for that highlight flaking issue, because the prints above aren’t exactly perfect, yet. My suspicion so far is that DAS just has a much higher ‘tonal threshold’ as Calvin Grier calls it. For a brief explanation, see e.g. this pdf on Gum Printing, page 19. The same principle applies to carbon transfer, and my experience so far is that dichromate-hardened tissues exhibit much less fragile highlights, and this pushes down the tonal threshold.

Remember that the tissues used for the tests above are very high-contrast ones: they have a high pigment load. The main way to push down the tonal threshold is to simply ensure that there’s a thicker gelatin layer remaining in the delicate highlights, making them a little less delicate. To obtain the same low optical density, this means the tissue needs to have a lower pigment load. Well, easy enough to test this!

I mixed up some more glop and reduced pigment load, but also increased the DAS load a bit. I figured that a little more DAS wouldn’t hurt in making the exposed gelatin tissue a bit firmer. I also made the tissue a bit thicker – since it has lower contrast, there will have to be a bit more gelatin to render the shadow areas, and I wanted to prevent exposing through the entire tissue, which leads to transfer problems. Here’s the formulation of that lower contrast glop:

% w/v in glop% of dry gelatin weight
Gelatin8.0%100 %
Sugar2.5%31.25 %
Glycerin 86%0.25%3.13 %
DAS sensitizer0.4%5.0 %
Kremer Pbk70.175%2.19 %
A reduced contrast glop with less pigment and a little more DAS

In the following examples, excuse the lack of proper calibration in the scanning process. The important thing here is how the highlights render. Also, you might find the negatives a bit odd in their lack of shadow detail. These are some of the negatives that were victim to my recent light meter failure. I ended up unleashing chromium intensifier onto them, followed by a healthy dose of pyrocat. After multiple rounds of poisoning them, they actually turned out to be…interesting. And they’re a lot nicer to print with than boring step wedges.

A print with the lower contrast glop – but a punchier negative, so it evens out in a way.
A closer look at the lower density areas.

In the high-contrast glop, the lighter patches in the leaves would have bitten out in several places. In fact, I also printed this negative with the higher contrast tissue and that’s exactly what happened. In this print, however, the transitions remain continuous. No, I mean, really – look:

1200dpi scan of the foliage area

In the zoomed-in examples above, the curves are not adjusted to make paper white align with pure monitor/digital white in order to make the highlight gradations more easily visible. As seen, they retain density, which was not possible with the higher contrast tissue.

Another example from the ‘accidentally underexposed but still kinda interesting’ series.
Closeup in 1200dpi scan

The same pattern in another print above; note once again the retained detail and density in the leaves, which in reality are extremely close to paper white.

As usual, scans do very little justice to the prints. While the negatives are evidently compromised, the compositions are somewhat gratuitous and the test prints are a bit of the quick & dirty persuasion, these little prints actually have something (in real life) to them that I quite like. Again, for me, this kind of testing beats printing step wedges, although the latter obviously allows for a more objective analysis and proper quantification. Perhaps I’ll put myself to that again one day.

For now, I intend to have some more fun with this new light source. I’ll mix up some more glop and make some more negatives to print with and to further my grip on DAS carbon. It feels like I’m getting closer to what I was able to do previously with dichromate – but with the improved ease of use of DAS-incorporated tissue. The only ‘sacrifice’ I’ll have to make is to really tailor the negatives a bit more to the process. This was a little less necessary with dichromate, but with DAS, there’s just no way around it. DAS just likes those really punchy, ‘classic’ carbon transfer negatives that you can use to armor a tank with.

Oh, and I’m already thinking about scaling up the nail-polish hardener LED light source to 8×10 size. Maybe I should order some more components and get cracking on that.

2 thoughts on “Kind of blue – A test with UV LEDs for DAS carbon transfer”

  1. Great find on the 3535SMD LEDs! They do sound perfect for DAS. I have a single 365nm 100W COB LED and frankly it’s been disappointing compared to my BLB tube unit, so glad there is an affordable alternative. Although I have no way to do SMD soldering and I haven’t had a PCB printed in years.

    A thought on highlights- I recently was struggling with some low contrast tissue I had that would retain highlights well, but shadows were not dark enough for my taste. So I ended up just printing twice onto a temporary support and then transferring both temporary supports to paper using visual registration as an experiment. It turned out great to my eye, so this is a long winded way of saying multi-layers prints might be another solution and it doesn’t have to be that complicated. This is similar to how Calvin printed an analog negative here I think

    1. Eric, great to hear from you again 🙂 So far, I’m very pleased with the ‘nail polish hardener’ LEDs! I haven’t looked specifically for ready-made solutions, but I wouldn’t be surprised if devices are turning up now that use these LEDs and that come pre-assembled. They will certainly be present in some nail polish hardener lamps, but these are generally small and…well, hand/finger-shaped! Not so suitable for printmaking. But the same wavelengths apparently are also popular in other industrial hardening applications, making me optimistic that these LEDs will pop up here and there.
      Another option is to put a second COB LED next to the one you’ve already got, but this one at 400nm. It’ll probably print faster (based on my experience so far), and combined with the existing 365nm unit you also should get good highlight retention. This is also how Calvin does it, btw! The main differences are that he (1) exposes the tissue in separate iterations, for shadows and highlights separately (with separate negatives; i.e. separations) and (2) he prints carbon generally with halftone screen negatives, which means that the whole tonal threshold issue is circumvented in his workflow. In fact, this is one of the main reasons why he apparently uses halftone negatives in the first place (the other being better linearization).
      So yes, I agree with your observation! I’ve also seen that print you linked to; it’s from a couple of years ago and seems to fit in the process he was going through at the time. I have a feeling things are more crystallized out at his end by now, but the main principles still apply.

      Btw, I’ve done the separation ‘trick’ as well, with large format in-camera negatives. Make two tissues, one a high contrast one and the other a low contrast one, and expose them through he same negative. Then transfer to the same polymer sheet just like you said, and in the end transfer to final support (paper sized with gelatin). I’ve in fact made colorful ‘split-tone’ separations this way. I still need to do a blog on this; thanks for reminding me! It’s an avenue I’ll be sure to explore further. It’s a very powerful technique indeed.

Leave a Reply

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