Variable Contrast UV Exposure Unit for Pyro Negatives

Here's a short video about an exposure unit I just finished building.


In my college's alt process darkroom, I noticed that when I printed an x-ray negative developed in pyrocat in a traditional fluorescent black light bulb exposure unit, I got significantly lower contrast than I got when I printed it in our Cone Editions VerifiedUV LED exposure unit. This is because pyro negatives have increased density in UV light, and the density gets greater the deeper the UV light gets. I think the BLB box uses blue/black bulbs likely dominant around 395nm. The cone edtions box on the other hand uses 365nm LEDs, and with the 365nm light, I got about 2 stops more contrast. With that in mind, I designed this box to use both 365nm and 395nm LEDs, which can be dimmed by PWM separately. By dimming the 365m I get lower contrast, and comparatively by dimming the 395nm I get higher contrast. This will make it easier to get perfect contrast in alt process prints without needing toxic chemicals like dichromates, while also keeping the workflow fully analog (no digital negatives).

I'll be starting to use this to print malde/ware platinotypes over the coming month, and will update you guys on how it goes.

That's a cool idea.

I'd suggest that part of the contrast increase is possibly due to the more collimated light the LEDs put out compared to the very diffuse light from BLB tubes, not so much the difference in wavelength.

How did you determine the 2-stop difference? It sounds like a pretty large difference, really.

Interestingly I've been printing with a 400nm LED source over the past few weeks/months and in the years prior to it I used BLB tubes. The difference in print contrast as measured on a step wedge is quite small, although effects such as halation are decidedly less with the LED source due to its better collimation, making the LED prints seem more contrasty than they really are.

Ethan Brossard said:

I think the BLB box uses blue/black bulbs likely dominant around 395nm.

Click to expand...

You'll find the spectrum of BLB tubes in fact peaks around 370nm or so. Here's some charts from an Ushio datasheet:

And this is for Philips TL-D 36W BLB:

As you can see, even BL (not BLB) tubes peak around 360nm, but with a broader spectrum. Still, the amount of energy in the >375nm part of the spectrum is quite small quantitatively speaking. Below 350nm or so, attenuation in all sorts of materials (including the polymer the negative film is made out of) tends to increase rapidly, so much of the energy in the lower wavelengths is wasted for printmaking.

I don’t think collimation played much of a role in the increased contrast. The LED source I used is a Cone Editions VerifiedUV exposure box, which has 6 led diodes per square inch spread over a 22x30 inch area. In other words, it’s very diffused as well.

2 grades would be the better terminology rather then stops. The way I determined the difference was by exposing a stouffer step wedge onto a piece of film, and then developing with pyro to create a pyro stained test wedge. I then printed the pyro wedge onto some paper coated with van dyke brown chemistry in both exposure units. Then, by measuring the density of the wedge on the steps corresponding with white and black on the prints, I determined that the fluorescent bulbs required a dynamic range of around 1.8 while the LEDs required a dynamic range of around 1.5.

Ethan Brossard said:

6 led diodes per square inch spread over a 22x30 inch area. In other words, it’s very diffused as well.

Click to expand...

No, this is not accurate; it's still a quasi-collimated light source this way. It's not perfectly collimated, but much more so than a bank of tubes, especially at short distances.

Ethan Brossard said:

2 grades would be the better terminology rather then stops.

Click to expand...

Ok, that makes more sense, thanks!
Nice test you did there; kudos for taking the time to do this. When I was ironing earlier today it occurred to me I could/should make a UV transmission densitometer. That would greatly facilitate measurements like these. I never really needed one before, but lately I've been more into measuring odds and ends here and there.

koraks said:

I'd suggest that part of the contrast increase is possibly due to the more collimated light the LEDs put out compared to the very diffuse light from BLB tubes, not so much the difference in wavelength.

Click to expand...

No Callier effect in contact printing.
Enlarging with a point source: source is (if prperly set up) imaged to lens diaphragm. Diffused light will miss the aperture. No such mechanism in operation in contact printing, or if the souce is diffuse. So-called "condenser" enlarger with a frosted bulb is an intermediate case.

Thanks @bernard_L
So what's your take on the apparent contrast difference between BLB and 365nm LED? Given that they both peak around 365nm or so, with the BLB's peak evidently being more of a lopsided mount Fuji-kind of shape whereas the LED is a proper peak.

koraks said:

So what's your take on the apparent contrast difference between BLB and 365nm LED? Given that they both peak around 365nm or so, with the BLB's peak evidently being more of a lopsided mount Fuji-kind of shape whereas the LED is a proper peak.

Click to expand...

Don't know. Sure, you provide nice spectra from various types of near-UV sources. And there is an apparent contradiction between your observation and that of the OP, Ethan Brossard. To proceed, one would need the actual spectra of your "400nm" LED source, your prior BLB tubes, ditto for the two sources experienced by the OP. Plus, either the OP or you speak of "alt process" without more specific details (unless I missed it); the light sensitive ion could be Iron or Silver (or dichromate??); conceivably one of these might have a sensitivity to a long-wavelength (>400nm) tail in the spectral distribution of the source, that is not captured in the single specification of the dominant wavelength... Pure conjecture, agreed, but might indicate where to look for an explanation of the "contradiction".
--------------------------------------------------------------------------------------------
Cyanotype (Iron)

--------------------------------------------------------------------------------------------
Silver bromide


--------------------------------------------------------------------------------------------
Gum bichromate ??
--------------------------------------------------------------------------------------------

The process I did my initial tests with was van dyke brown, and thinking back, I believe they were actually actinic bulbs, rather than blacklight blue, so looking at the spectral emission, there are more spikes higher in the frequency range

bernard_L said:

one would need the actual spectra of your "400nm" LED source, your prior BLB tubes,

Click to expand...

For the BLB tubes, see earlier post above. I use a bank of the Philips ones I gave the spectrum for.
For the LEDs, it's a standard LED spectrum. I don't think it's relevant to figure out how narrow exactly the typical LED peak is as it'll boil down to pretty much the same regardless.

My testing was with Van Dyke as well in addition to dichromate. I might do some more measurements at some point, although frankly it's not very high up my list of priorities.

bernard_L said:

conceivably one of these might have a sensitivity to a long-wavelength (>400nm) tail in the spectral distribution

Click to expand...

Hmmm. That's a pretty wild one as guesses go Who knows, but I'm very skeptical based on how the electrochemistry would go for something like that. It's kind of unlikely that a sensitivity peak would accidentally occur for a wavelength that's far less energetic than the rest of it. That sort of thing is hard enough to engineer on purpose, let alone getting it as a free gift.

Here's a little proof of concept test I did today using a single negative, printed twice. Once using 365nm VerifiedUV V2 leds from Cone Editions, and the other time using 395nm UV LEDs from waveform lighting. The negative is a stouffer wedge I copied onto a piece of film, and processed with pyrocat HD to mimic the negatives I'll be printing with this box. As you can see, the 395nm LEDs require a longer dynamic range in comparison to the 365nm LEDs.

Ethan Brossard said:

Here's a little proof of concept test I did today using a single negative, printed twice. Once using 365nm VerifiedUV V2 leds from Cone Editions, and the other time using 395nm UV LEDs from waveform lighting. The negative is a stouffer wedge I copied onto a piece of film, and processed with pyrocat HD to mimic the negatives I'll be printing with this box. As you can see, the 395nm LEDs require a longer dynamic range in comparison to the 365nm LEDs. View attachment 326276

Click to expand...

Interesting study. This relationship would hold true even for regular negative, right?

While there seems to be some difference, I think you have to normalize both to their min exp for Dmax to see its real extent. What I can gather, 395nm is a couple of steps faster so it runs short at the highlight end. It also seems to me qualitatively (from the picture which may not represent the real thing) that the shadows are much shallower (perhaps longer toe?) in the 365nm case while for 395nm they seem to stop abruptly at 14. Dmax seems to be lower in 365nm too.

:Niranjan.

nmp said:

Interesting study. This relationship would hold true even for regular negative, right?

While there seems to be some difference, I think you have to normalize both to their min exp for Dmax to see its real extent. What I can gather, 395nm is a couple of steps faster so it runs short at the highlight end. It also seems to me qualitatively (from the picture which may not represent the real thing) that the shadows are much shallower (perhaps longer toe?) in the 365nm case while for 395nm they seem to stop abruptly at 14. Dmax seems to be lower in 365nm too.

:Niranjan.

Click to expand...

I printed these to reach dmax in the “shadow” area, and watched where the highlights fell. It doesn’t give a quantitative proof of how much contrastier the 365nm is, but it’s getting closer. The negative isn’t dense enough to give pure white in the 395nm test, so I don’t know exactly how much more contrast it has, I’ll have to make another test wedge for that.

Looking at the densities of both prints, the tones from 14 to 6 on the 365nm test cover the same range as 13 to 1 on the 395nm test. Measuring the densities of the negative at those points tells me the 365nm light required a dynamic range of 1.2 for that range, and the 395nm light required a dynamic range of 1.45 for that range. Not quantitative, but a qualitative proof of concept.

The theory behind this is that pyro stains make negatives denser to uv, and the shorter wavelength the uv the denser it gets, so with shorter wavelength uv a pyro negative will be higher contrast then with longer wavelength uv

So, I don’t think normal negatives would behave any differently with this system.

Sorry if I missed it, but is this Van Dyke or silver chloride?
Either way, good test and compelling result! I'd love to see it repeated for other processes - and no, I'm not urging you specifically to do so

The one thing that puzzles me is the density of the text on both prints - it's denser on the 395nm print than on the 365nm one. Is this because the 365nm print was exposed shorter and the text has rather high density compared to the 21 patch?

Ethan Brossard said:

the shorter wavelength the uv the denser it gets

Click to expand...

While your hypothesis makes sense and I'd like to believe it, your test does not yet prove it. You could repeat the test with the bare Stouffer (not the stained replica) and compare with this outcome. Perhaps you already did this?

Btw, there's a rather old/dated test by Sandy King of UV light sources; it doesn't involve LEDs as the technology wasn't available back then. Therefore, the correlation between contrast and wavelength is much more difficult to establish. How would you reflect on your findings compared to those of King?

Ethan Brossard said:

In my college's alt process darkroom, I noticed that when I printed an x-ray negative developed in pyrocat in a traditional fluorescent black light bulb exposure unit, I got significantly lower contrast than I got when I printed it in our Cone Editions VerifiedUV LED exposure unit. This is because pyro negatives have increased density in UV light, and the density gets greater the deeper the UV light gets. I think the BLB box uses blue/black bulbs likely dominant around 395nm. The cone edtions box on the other hand uses 365nm LEDs, and with the 365nm light, I got about 2 stops more contrast. With that in mind, I designed this box to use both 365nm and 395nm LEDs, which can be dimmed by PWM separately. By dimming the 365m I get lower contrast, and comparatively by dimming the 395nm I get higher contrast.

Click to expand...

Ethan Brossard said:

Here's a little proof of concept test I did today using a single negative, printed twice. Once using 365nm VerifiedUV V2 leds from Cone Editions, and the other time using 395nm UV LEDs from waveform lighting. The negative is a stouffer wedge I copied onto a piece of film, and processed with pyrocat HD to mimic the negatives I'll be printing with this box. As you can see, the 395nm LEDs require a longer dynamic range in comparison to the 365nm LEDs.

Click to expand...

Ethan Brossard said:

Not quantitative, but a qualitative proof of concept.

Click to expand...

I had myself a difficult time trying to persuade myself that one of the two step exposures had a higher contrast. So I spent a little time trying to make it quantitative.
No claim to accuracy, but quantitative and unbiased.

• Save image as tif
• Perform median smoothing on the exposed steps (because probe size in next step is not adjustabe)
• Load ("scan") into vuescan, use density probe
• Sample the steps, record in spreadsheet, graph

I repeat: the "densities" are not accurate, but are unbiased and objective. The plot below shows that one of the two LEDs provides more sensitivity (brighter? more suitable wavelength? who knows?), but IMO, the two curves (with their little imperfections) show essentially the same contrast. With better precision than "counting steps".

Being the skeptical type, I couldn't resist replicating @bernard_L's exercise:


I've plotted L* values, so the plot is the inverse of Bernard's.
Blue = 365nm
Orange = 395nm

Alright, I replicated it mostly because it's easier in this plot to see something that piqued my interest: the slope of the straight line portion is indeed very similar, but the toes and shoulders differ. This is difference is particularly present on the low-density side.

Does the penetrating ability of UV change with the wavelength? In carbon printing, I have a relatively thick emulsion, so I am curious if the difference between shorter or longer UV wavelengths is significant.

Using a thick emulsion (layer of glop) in carbon printing pushes some the effects of the types of light sources on the image. While far from a true point light source, my merc vapor lamps send the UV fairly well straight to the contact printing frame, with most of the UV going straight down into the emulsion on the quickest route to the bottom.

Some of the BL (or BLB) tubes' diffused light also heads straight down to the contact frame, but there is very high proportion of light entering the emlusion at shallow angles, whose light paths thru the emulsion are much longer, and will not penatrate (expose) as deep as the same strength light hitting perpendictularly.

In my experience, both sharpness and contrast was noticibly reduced using BLB tubes vs a Merc Vapor lamp...but not for processes with very thin or no emulsion (such as my platinum printing).

Both points made by Vaughn are valid. The second one

Vaughn said:

Some of the BL (or BLB) tubes' diffused light also heads straight down to the contact frame, but there is very high proportion of light entering the emlusion at shallow angles, whose light paths thru the emulsion are much longer, and will not penatrate (expose) as deep as the same strength light hitting perpendictularly.

Click to expand...

which is completely independent from the Callier effect (my post #5), might explain a relation between wavelength and contrast in certain processes.
If such a relation would be observed. The evidence so far is not compelling, to say the least.

bernard_L said:

Both points made by Vaughn are valid.

Click to expand...

Yes, but specific for carbon transfer or other processes with a thick emulsion.

Some of you raised some questions about my results I couldn't answer, so I went back to the darkroom to continue testing and see if I could come up with more concrete evidence proving either way whether different wavelengths of UV light can change the contrast of a negative developed with Pyro developers. I started where I left off with stouffer wedges copied to pyrocat HD negatives, but ran into confusion. I wasn't able to measure a consistent change in dynamic range between 395nm and 365nm light sources. Some tests seemed to differ by as much as 3/4 of a stop to a full stop, while others didn't differ much at all. Part of the problem was that the stain from pyrocat HD doesn't differ a large amount between 395nm and 365nm, but I also realized that as van dyke brown prints out, the shadows will self mask, creating the illusion of greater dynamic range when overexposed. With that blocking my way, I went back to the drawing board and did some more research. I found that Pyrocat MC is an all around imrpoved version of Pyrocat developer, and has a greater staining effect in the UV range, which I theorized would make it easier to see a difference in the dynamic range of two tests printed from a Pyrocat negative. I bought some from Photographers formulary, and it arrived today, allowing me to continue my testing.

I exposed a sheet of Fuji HR-U X-Ray film (what I've been working with most lately) in contact with a stouffer wedge under an enlarger. I then processed in Pyrocat MC 3:2:100 for 10 minutes. the resulting negative has a visible dmax of around 2.23, and a visible dmin of 0.32, giving a visible dynamic range of 1.91. I coated a piece of stonhenge paper with Van Dyke Brown chemistry mixed from scratch around 3 weeks ago, and cut two 1 inch strips from it. I then printed my Pyrocat MC test wedge on both these strips, one in a 365nm Cone Editions VerifiedUV exposure box, and the other in a box containing 395nm Waveform LED light strips. The exposure times were 90 seconds in 365nm, and 360 seconds in 395nm (the difference of two stops is because though the LEDs have similar power, there are twice as many 365nm LEDs in that box, and 365nm is much closer to the peak sensitivity of the process). I then processed the strips together, washed and dried them. After dry down, looking at the strips the dynamic range of the 365nm test wedge, going from Zone II (one step above paper white) to Zone VIII (one step below dmax), the 365nm test goes from step 9 to step 15, and the 395nm test goes from step 7 to step 16. Reading the density of the negative wedge at those steps reveals a dynamic range of 1.29 needed for 365nm, and 1.59 needed for 395nm.

I also looked again into the other reason I theorized as to why one would appear more contrasty then the other: the self masking nature of van dyke brown. Looking at the data, dmax begins sooner on the 395nm test when compared to the 365nm test, indicating to me that if one of the tests was overexposed, it would be the 365nm test. So, if the 365nm test saw more of the faux self masking contrast, if anything one may be able to achieve a greater amount of contrast variation using this technique than my test indicates.

So based on this, with wavelengths of 365nm and 395nm, and using Pyrocat MC as your developer, it is possible to make a print from any negative within the range of dr1.29 to dr1.59, by adjusting the amounts of light emitted by each wavelength of LED.

[edit: another note, I'm not sure it will be of much use trying to graph the contrast from the image of the test wedges, as that will be influenced both by the digital scan of the strips. My professor has an Xrite reflective densitometer I may be ble to use to measure that later, but I'm not sure)

If one exposure is 90s and the other 360s, wouldn't there be more self-masking in the latter? And therefore lower contrast?

FotoD said:

If one exposure is 90s and the other 360s, wouldn't there be more self-masking in the latter? And therefore lower contrast?

Click to expand...

If the self-masking is related to the print-out density and that tracks fairly well between both exposures, then no. I mean, the exposure times were established to be in the same ballpark in terms of final print density, so logically the self-masking effect should be similar. With the exception that the self-masking effect might indeed be wavelength-dependent. I've never really looked into this.

I'm still digesting this experiment, but so far it seems compelling enough. The notion that the pyro stain will have a different opacity to different wavelengths makes good sense. I haven't yet scrutinized the experiment well enough to exclude the possibility of exogenous factors that might give the same impression, but constitute a different mechanism. However, this really looks pretty solid at first glance!

Coincidentally, I just ran into a wavelength issue as well, but it was a of a different nature and specific to DAS-sensitized carbon transfer, so it's a different kind of effect. However, in that situation, wavelength definitely did prove to matter!

FotoD said:

If one exposure is 90s and the other 360s, wouldn't there be more self-masking in the latter? And therefore lower contrast?

Click to expand...

koraks is correct. The longer exposure is due to two factors. The first is that the box I exposed the 365nm test in has twice the amount of LEDs per unit area, which halves the time once, and 365nm is closer to the peak sensitivity of iron salts from what I've read, which seems to halve the exposure time again, resulting in a roughly 4x faster exposure. The times were determined independantly to reach max density and go a bit beyond, but not by much.

Another thing I just realized is that the dynamic ranges I listed are innacurate as I forgot to subtract the film base + fog value, which is somewhat high on the x-ray film, and maybe even a bit higher than usual since the bottle of Pyrocat MC solution A I recieved had no seal, and I've been getting a higher fog value with it than with Pyrocat HD, which is the opposite of what I've read should be the case.

So, with the film base+fog value of 0.32 subtracted, the range of dynamic ranges this box works with are 0.97-1.27, so actually a fairly flat negative by silver print standards. (for Van Dyke Brown that is, for other processes the required range will be different)

koraks said:

If the self-masking is related to the print-out density and that tracks fairly well between both exposures, then no. I mean, the exposure times were established to be in the same ballpark in terms of final print density, so logically the self-masking effect should be similar.i

Click to expand...

I guess that would make sense if the print out was instantaneous and completely linear between different exposures.

It would also mean that all that are making a pause in the middle of their exposure to allow for print-out and lower contrast are wasting their time.

FotoD said:

I guess that would make sense if the print out was instantaneous and completely linear between different exposures.

Click to expand...

Those are safe to assume at least for processes like salted paper (evidently) and Van Dyke. I can't vouch for Pt/Pd as I haven't done any (only some pure Pd), but I'd be surprised if it were otherwise.

FotoD said:

making a pause in the middle of their exposure to allow for print-out and lower contrast

Click to expand...

Is that a thing? Alright, I can sort of see the sense in it, but I'd be very cautious in assuming this does anything at all. Sounds like it would be simple to do a little A/B test and verify if it works. I really don't expect so, though. All processes I've done so far seem to print out instantaneously, without any delay.

koraks said:

Is that a thing? Alright, I can sort of see the sense in it, but I'd be very cautious in assuming this does anything at all. Sounds like it would be simple to do a little A/B test and verify if it works. I really don't expect so, though.

Click to expand...

"Mike Ware suggests that this technique may allow for a greater degree of print-out, which may be slow to bulid, and therefore more self-masking." The Book of Alternative Photographic Processes, Christopher James.

koraks said:

Those are safe to assume at least for processes like salted paper (evidently) and Van Dyke. I can't vouch for Pt/Pd as I haven't done any (only some pure Pd), but I'd be surprised if it were otherwise.

Click to expand...

It seems odd to assume that an accepted effect simply doesn't exist, while trying to prove that a small difference in contrast is explained by a never before reported phenomenon.

It should be fairly easy to test. Place a 2 stop ND filter between the 365nm lights and the paper and see if the contrast difference is still there. The filter could be any translucent sheet that gives the same exposure time as the 395 nm lights.