Craziness - using LEDs to print RA4 (and B&W)

koraks -
fantastic stuff here ! thanks for sharing, and have fun printing

wyofilm said:

Why Photrio? Threads like this one!

Click to expand...

+1 !
john

koraks said:

Many thanks!
Yes, you could use a lot of small low-power leds instead of fewer high-power leds. It might make it a bit more difficult to make an effective diffuser; currently, my light source is of a similar area as the maximum negative size, which in practice works quite conveniently (and fits inside the 138s head). Lower-powered leds would either result in a larger surface area of the light source or lower total light output. I wouldn't myself want to compromise on the latter; in fact, the next light source I will construct will likely increase the number of green leds (so a ratio of 2:2:1 for R:G:B whereas I now have 2:1:1) to make b&w printing faster. The problem with lower power is that exposure times escalate; it's a powers-of-two relationship after all. Especially with smaller formats (e.g. 35mm) this can become problematic at some point.
If you were to cram a lot of small leds together on as small as possible surface area, there may be some advantage in terms of heating up of the led dies themselves, but you'd still have to get rid of the large overall amount of power, of course. But maybe that's easier than keeping the dies cool, since those are buried within the manufacturer's led component and therefore possibilities of cooling the dies directly are very limited.

Click to expand...

Small LED's are apparently the Heiland approach - their 5x7" head for the Durst 138 has 1600 LED's according to their web site.

Even though Heiland is offering a RGB LED head, they seem to be marketing it specifically for VC black and white printing, and seem hesitant to venture directly into color printing applications or recommendations in that respect. I'd want to see some very specific spectrograms and luminance values for anything like that.

DREW WILEY said:

Even though Heiland is offering a RGB LED head, they seem to be marketing it specifically for VC black and white printing, and seem hesitant to venture directly into color printing applications or recommendations in that respect. I'd want to see some very specific spectrograms and luminance values for anything like that.

Click to expand...

They do say it is ok for color on the German part (https://heilandelectronic.de/led_kaltlicht/lang:de) of their web site: "Die Lichtquellen eignen sich Farb- und SW-Papiere. Entsprechende Steuergeräte bieten wir an" - translation: The light sources are appropriate for color and b/w papers. The respective controllers are offered by us.
Interestingly, the English version (https://heilandelectronic.de/led_kaltlicht/lang:en) does not say that.

Thank you. Well, OK is OK, but might not meet the demands of a serious user in terms of either color accuracy or sufficient lumen output, versus traditional halogen colorhead options. In other words, the burden of proof is still on them until there's a serious track record, and that would mean someone gambling a fair amount of their own money and time to find out. Maybe they're just being cautious and wiggling their toes in the water first. German marketing conservatism, which I'm quite aware of, having specialized in distributing German machinery before I retired. I'm not shopping myself, already having more than my fair share of big color enlargers. But the continual search for the holy grail is always interesting - something low maintenance which runs cool, yet also capable of providing clean hues.

DREW WILEY said:

Thank you. Well, OK is OK, but might not meet the demands of a serious user in terms of either color accuracy or sufficient lumen output, versus traditional halogen colorhead options. In other words, the burden of proof is still on them until there's a serious track record, and that would mean someone gambling a fair amount of their own money and time to find out. Maybe they're just being cautious and wiggling their toes in the water first. German marketing conservatism, which I'm quite aware of, having specialized in distributing German machinery before I retired. I'm not shopping myself, already having more than my fair share of big color enlargers. But the continual search for the holy grail is always interesting - something low maintenance which runs cool, yet also capable of providing clean hues.

Click to expand...

True. They are a small company, so they might not have the capacity to test check everything 100% and playing it safe, as you say, especially since the main market is b/w splitgrade printing. I wouldn't be concerned about the brightness, after they first came out with their heads they had to offer an additional control box later to reduce brightness by up to two stops, and that was definitely needed (speaking from personal experience here). Btw, here is a link to their color controller: https://heilandelectronic.de/color_timer/lang:en

Not to discredit the Heiland system, as I obviously haven't used it, and I trust their engineering is of very high standing. But...I personally find the control electronics rather simplistic in terms of functionality. A separate control unit for color and B&W, no integrated timer functionality - I mean, as a potential customer, that wouldn't appeal much to me. It seems to me that Drew is at least partly right and that the light source was primarily conceptualized for B&W printing and color was added as almost an afterthought. But that's conjecture on my part. One thing is certain: I have not yet had the opportunity to talk to anyone who actually uses their product for color printing.

koraks said:

Not to discredit the Heiland system, as I obviously haven't used it, and I trust their engineering is of very high standing. But...I personally find the control electronics rather simplistic in terms of functionality. A separate control unit for color and B&W, no integrated timer functionality - I mean, as a potential customer, that wouldn't appeal much to me. It seems to me that Drew is at least partly right and that the light source was primarily conceptualized for B&W printing and color was added as almost an afterthought. But that's conjecture on my part. One thing is certain: I have not yet had the opportunity to talk to anyone who actually uses their product for color printing.

Click to expand...

Oh, you're most likely right about that. Initially, their LED heads could only be used with their splitgrade controller (which is an older development than the LED head and obviously just for b/w), and the "normal" controllers, e.g. with the grade settings in b/w or the RGB color one, came more recently.

Well, I applaud all the efforts this direction. I do see it as a reasonable if pricey improvement over the cold light concept. But it's hard to say about color applications yet. And with the diminishing market for new color enlargers, custom or individual production is probably going to be the primary route.

I built an LED-head using 128 LEDs RGB which should be used for color printing. The total power is around 20 Watt.
When enlarging a 6x9mm negative to 12x18 (enlarging factor 2) using a 105mm lens at aperture of 8,
exposure times exceed 1 minute at full power!
There is no way to make large formats, because you will have times in the 3-5 minute range.


The problem with color prints is that red sensitivity of the paper is just a small fraction of green an blue.
Even B/W is printable in the 5-10s range but RA-4 paper needs LOTS of red!
Red is not enough and at the same time you waste lots of available green and blue power because you have to bring it down.
I don't think that Heiland uses more red LEDs than green or blue ones. Color printing may be "OK" but I guess you will have veeeeeery long exposure times.
The other problem is that you have very limited range of intensity control at green and blue.
If all colors go from 0-200 for example, the maximum value for blue will be around 20. 20 steps is not enough to do a proper color filtration.
So intensity of R, G and B LEDs has to be pre-calibrated to for example 100%R 20%G and 10%B.
After this, electronic control can be 0-200 for each.

My color head was a test that failed. My next head will have more power at red plus much more overall power.
Now I plan about 40 Watts of red and 10 Watts of green and blue.
This will be 5x the power for R as I had at the last test and will hopefully bring times down to <1min for larger formats (>16 inch).

Joachim

That looks very nicely engineered for a one-off product, Joachim! Too bad it doesn't live up to its purpose. Yes, you need a *lot* of red in comparison to green and blue. If you also want to do variable contrast, I would recommend also increasing green, as with variable contrast emulsions, it turns out you need quite a lot of green in comparison with blue. For my next build I will therefore seriously consider the 2:2:1 (R:G:B) ratio I mentioned before. For color work, 2:1:1 would be a good start, although 3:1:1 is perhaps even closer to what you need, but then you may have to start capping the output of the red channel in some situations.

JoJo said:

The other problem is that you have very limited range of intensity control at green and blue.
If all colors go from 0-200 for example, the maximum value for blue will be around 20. 20 steps is not enough to do a proper color filtration.
So intensity of R, G and B LEDs has to be pre-calibrated to for example 100%R 20%G and 10%B.
After this, electronic control can be 0-200 for each.

Click to expand...

Yes, that general line of argumentation is about right. To make matters more complicated, a linear attenuation does not really approximate a true dichroic filter head; you're looking for a more exponential change between settings. So 200 steps of attenuation will actually require many more steps in software or electronics - or you'd have to 'hardwire' the desired exponential curve by means of discrete electronics, which is inflexible and unnecessarily complicated IMO.
To overcome all this, I did the following:
1: Use twice as many red lights as blue and green, so that 'automatically' already gives a 2:1:1 ratio without further measures.
2: I use a 12-bit PWM controller instead of relying on e.g. an Arduino's in-built 8-bit PWM. This gives me 4096 distinct power levels to play with, and that conveniently allows something like 300 brightness steps with exponential curves to be used, in which a single unit change in any of the channels still works out as a distinct light level. See the curve plots I posted before; in my software, they are translated into a range of 0-4095 that is written to the PWM controller.
3: The 2:1:1 ratio of the leds requires some further finetuning (quite a bit in fact), or in other words, blue and green need to be further capped. Empirically I reached an approximation, which turns out to be a maximum level of 4095 for red, 3689 for green and 2514 for blue. In other words, actual electrical power driving each of the channels is only 45% of the red power for green and just 30.7% for blue in comparison with red.

The story above is yet further complicated by the actual efficiency or luminous output per watt of each of the leds. For this, I could find no really good data, which has a few causes:
1: The leds I used like I said before don't come with any datasheet and only a few key parameters for use (max current, voltage drop range, peak wavelength and total lumens output).
2: While the total output in lumens is provided for these leds, the lumen range is non-linear and based on the sensitivity of the human eye. In other words, the same number of blue lumens is not the same absolute photon flux as the same number of green lumens, etc.
So lumens seem at first glance a nice approximation of output power, but when you think about how they are defined and how that relates to the sensitivity of paper (which evidently is completely different from the sensitivity of the human eye), lumens become nearly useless as a power measure.
For all intents and purposes, I therefore conclude that we're pretty much limited to experimentation and some rough estimations.

One way of trying to establish a neutral white output for the led head, i.e. R, G and B levels that correspond fairly well to a halogen light source in a regular enlarger, was to use the white balance of a digital camera for this purpose. Essentially I just shot a digital picture of a grey baseboard under a regular enlarger (halogen light source without any filtration) and then tuned the R, G and B channels on the led head until they produced a similar color temperature as 'measured' by the digital camera. This turned out to be a *very* rough approximation indeed, as it throws the sensitivity of a digital camera sensor (R, G and B...) into the mix of complexity and it proved to be impossible to get exactly matching results. However, it got me close enough - but never close enough to achieve perfect translation of filter values from a dichroic head to the led head. That wasn't so important to me, but initially I did try that, but evidently, was not successful in that respect.

If in your next try you are aiming for 40+10+10W I bet you would be very close or right on the money for enlarging up to 6x9cm. It may depend a bit on the light path after the leds. In my setup, I use the Durst 138 unmodified; I put the leds in an approx. 10x10x10cm enclosure with one open end (simplified a bit; there's also a fan on another side and ventilation slots on the opposing side to run cool air across the surface of the leds) and the open end holds a milky white plexiglass diffuser. That sits right in the place where the 138 normally has its heat filter installed, and after that it's just an unmodified condenser setup. I use the 240 condenser in the top position and the 200 for 4x5, the 130 for 6x6cm and either the 130 or the 85 for 35mm. So the top condenser is never changed, as opposed to when an old-style bulb is used as a light source. This way I get a lot more overall light at the same size print for larger film formats than for smaller film formats, but even with 35mm it's workable.

Exposure times are around 4.5 seconds for a 12x18cm print from 35mm negative at f/8 (Nikkor 75/4, 240+85 condenser). This is at something like 120W of nominal led power (in reality for a given print, actual consumed power will be probably something like 40-50W due to attenuation of blue and green). Your exposure time of ~1 minute for a 2x enlargement of 6x9cm at 20W nominal led power isn't all too far off from this, so sounds about right. You just want some more power for this. I based my 'somewhere around 100W' estimate also on the 250-1000W incandescent sources that are usual for the Durst 138, figuring that 100W of led power should get me pretty close (or even better) to the light level of a 300-400W halogen light source.

This is brilliant, I love it.

I do something similar, but with a LIFX RGB bulb in a condenser enlarger. Last week, I did a large volume of color prints with a new to me Durst roller transport. I did not infact notice any colour shift. Not sure if there is just better cooling, smaller LEDs or programming to correct for colour variance. I might suggest you try it. The only catch is that the bulb has to be raised about 1/2 to 3/4 inch as is sits too close to the condenser lens and bends the light. A roll of masking tape makes the perfect shim. The results, to my eye, match what I was getting with a minolta 45a. The exposure times are quick, being in a condenser. 8sec at f16 or f22 for a medium format negative. You can also control the brightness of the bulb itself. All my RA-4 times are exposed at 8sec, using aperture and brightness of the bulb to control light.

LIFX has an online API. I use set-state to crudely control the colour, which rarely needs to be changed. (https://api.developer.lifx.com/docs/set-state) BUT!!! If you wanted, you could easily program multigrade printing with the api. You can program transitions with api. On/Off so that you don't need a separate timer. I find that with brightness turned way down on my phone, I can use it in a darkroom without fogging the paper. I've experimented with split grade printing and found grade 0 to be around hue of 60 and saturation of 100 and grade 4-5 to be around hue of 330 and saturation of 0.6.

LIFX controls the light colour with Hue and Saturation. This is how I learned to print RA-4 and it is now intuitive to me. I found RGB on the minolta 45a difficult to balance. If you had the interest and time you could program something to convert Hue and Saturation to CMY or RGB and control it that way, but you might find that, like me you prefer Hue Saturation.

For Kodak Films on Endura Paper, developed with replenished kodak RA-RT chemicals, I am finding Hue of 336 and Saturation of 0.6 or so. See the HSB ring below. Since this is negative film, you move away from the colour you are trying to add. If you are trying to give it more magenta, print at a hue of 338 or 340 rather than 336. If more yellow, hue of 334 or 332. If you want more cyan (actually turqoise), then increase saturation by 0.1 or 0.2. If you want more red in the image, decrease saturation by similar margin. For Dense negatives, like long exposures with colour shifting I've had to use very low saturation of 0.1 in some cases to get acceptable results.

If you were really good at programming, you could use the phone to take a picture of the print, and make a 'digital' ring around on the phone. Automatically set Hue and Saturation and print from there without ever having to make a million prints. For me personally, I would rather spend my limited time printing. I do a pretty good job just from eyeing and having a feel for HSB. If Heiland or Beseler or Ilford anyone else is reading this, feel free to partner with LIFX and run with the idea!

cmyk-HSB by Aaron, on Flickr

Last year, before he passed, the great Photo Engineer complimented my results. I think that should give the skeptics some pause:

Gold 200 on Endura print 2 by Aaron, on Flickr

1kgcoffee said:

This is brilliant, I love it.

I do something similar, but with a LIFX RGB bulb in a condenser enlarger. Last week, I did a large volume of color prints with a new to me Durst roller transport. I did not infact notice any colour shift. Not sure if there is just better cooling, smaller LEDs or programming to correct for colour variance. I might suggest you try it. The only catch is that the bulb has to be raised about 1/2 to 3/4 inch as is sits too close to the condenser lens and bends the light. A roll of masking tape makes the perfect shim. The results, to my eye, match what I was getting with a minolta 45a. The exposure times are quick, being in a condenser. 8sec at f5.6 or f8 for a medium format negative. You can also control the brightness of the bulb itself. All my RA-4 times are exposed at 8sec, using aperture and brightness of the bulb to control light.

LIFX has an online API. I use set-state to crudely control the colour, which rarely needs to be changed. (https://api.developer.lifx.com/docs/set-state) BUT!!! If you wanted, you could easily program multigrade printing with the api. You can program transitions with api. On/Off so that you don't need a separate timer. I find that with brightness turned way down on my phone, I can use it in a darkroom without fogging the paper. I've experimented with split grade printing and found grade 0 to be around hue of 60 and saturation of 100 and grade 4-5 to be around hue of 330 and saturation of 0.6.

LIFX controls the light colour with Hue and Saturation. This is how I learned to print RA-4 and it is now intuitive to me. I found RGB on the minolta 45a difficult to balance. If you had the interest and time you could program something to convert Hue and Saturation to CMY or RGB and control it that way, but you might find that, like me you prefer Hue Saturation.

For Kodak Films on Endura Paper, developed with replenished kodak RA-RT chemicals, I am finding Hue of 336 and Saturation of 0.6 or so. See the HSB ring below. Since this is negative film, you move away from the colour you are trying to add. If you are trying to give it more magenta, print at a hue of 338 or 340 rather than 336. If more yellow, hue of 334 or 332. If you want more cyan (actually turqoise), then increase saturation by 0.1 or 0.2. If you want more red in the image, decrease saturation by similar margin. For Dense negatives, like long exposures with colour shifting I've had to use very low saturation of 0.1 in some cases to get acceptable results.

If you were really good at programming, you could use the phone to take a picture of the print, and make a 'digital' ring around on the phone. Automatically set Hue and Saturation and print from there without ever having to make a million prints. For me personally, I would rather spend my limited time printing. I do a pretty good job just from eyeing and having a feel for HSB. If Heiland or Beseler or Ilford anyone else is reading this, feel free to partner with LIFX and run with the idea!

cmyk-HSB by Aaron, on Flickr

Last year, before he passed, the great Photo Engineer complimented my results. I think that should give the skeptics some pause:

Gold 200 on Endura print 2 by Aaron, on Flickr

Click to expand...

Which specific LIFX bulb are you using?

A19 11watt 1100 lumens

model # LHA19E26UC10

I'll be doing more printing today with it.

Good to see you are pleased with the results of your lifx bulb. I dont think I'm going to try it, however. It's seriously underpowered for my purposes and to be frank, I dont expect the color rendition to be on par with what I want. Wavelengths for the individual colors are not specified (not a good sign to begin with) and the red and blue are most likely too close to green. This is the case with 99% of the red and blue leds intended for interior lighting and I've experimentally confirmed that this is not good enough for ra4 printing. I'm afraid your sample image very much confirms my concerns, although it's always hard to judge digital reproductions. But the dull and impure colors are exactly the problem I also saw with the first version of my light source.
However, if course, if it works for you, then that's great!

I'll post a couple more examples this evening if you wouldn't mind giving your opinion on deficiencies. Always appreciate the feedback if I can become better at the craft.

I’m a little surprised no one has jumped on this in a big way.
1kgcoffee, I remember when you started working with this. Frankly, the initial results looked dubious. But you got it dialed in and it clearly works. So you can screw this $50 bulb into just about any old b&w head with a little fiddling and voila, you now have a color head and a split grade b&w head. It would seem that if one had a color head to start with and a color analyzer you could quite easily come up with the equivalents between the two.
If I didn’t already have a bunch of color heads I would be tempted to spend the time to get into something like this. It appears brilliantly simple.

1kgcoffee said:

I'll post a couple more examples this evening if you wouldn't mind giving your opinion on deficiencies. Always appreciate the feedback if I can become better at the craft.

Click to expand...

Some skin tones and a gray card/chart would be great to see.

wyofilm said:

Why Photrio? Threads like this one!

Click to expand...

.
I'm beginning to see your point.

1kgcoffee said:

I'll post a couple more examples this evening if you wouldn't mind giving your opinion on deficiencies. Always appreciate the feedback if I can become better at the craft.

Click to expand...

Please, go ahead! Take my criticism with a grain of salt, or at least the side note that what matters in the end is if you are pleased with your results.

Ok, here are a few. The Both professionally processed and done in kodak chemicals. I cannot get the scan to look exactly the same as the prints. Highlights are more cyan in the scan:
Ektar on Vega 12b 90mm lens

Monkey Jungle by Aaron, on Flickr

Gold 200 on Pancolar 50mm

Birthday by Aaron, on Flickr

Blind Bay by Aaron, on Flickr

Well, I suppose it's maybe also a bit of a matter of taste, but here goes:
#1 is too cyan, with an odd oversaturated section of reds/maroons at the bottom. If you correct the cyan in the skin hues, the foliage in the background will shift to an unnatural, desaturated brownish green.
#2 is also a bit cyan, but it's not an unusual look for indoor flash photos. However, a similar issue with color crossover here: if you correct the skin tones, the browns in the chocolate and the background will end up too red.
#3 this one is a bit odd as the edges seem to lean towards yellow and magenta, while the center is shifted a bit towards cyan and blue. If I got a print like this, I'd probably make some test prints of an entirely neutral grey surface to determine if the problem is on the printing process (eg uneven illumination and color balance of the light source).
Particularly in #1 and #2 the color crossover is the kind of problem that made me switch to different types of leds.

One thing you could try is to print some greyscales and color checkers to see how close you can get to a reference and to determine if there are color crossovers between highlights and shadows (which I suspect is indeed the case).
Other than the light source there could of course be issues in the process of making the negatives; there are many variables to deal with.

Just depends on what you're after. Having a fun funky look is fine for some things, not for others. Kodak gold isn't even a film suitable for any kind of objective tricolor test, much less a web version of it. I still don't think LED color printing is anywhere near getting on first base yet. The correct RGB primary sources don't all exist yet. That doesn't mean a print will be unappealing in its own way. But actual hue objectivity is a whole other layer to the topic. In other words, by all means enjoy the creative niches made possible by these LED head experiments. Just don't claim they're a substitute for actual pro colorheads using long-proven technology.

To me, the images posted by 1kgcoffee look to have both a strong cyan cast and some real cross-over problems. However, the intervening scanning and downloading processes make any evaluation from this end unreliable at best.
The approach he is taking is interesting, but I still would be much more comfortable with a colour control system which matches the one that the paper and chemicals were designed for (and which I learned on), rather than a system based on hue and saturation adjustments.

It's outright impossible to do objective hue and saturation studies with a film in which exactly zero primaries or tertiaries are capable of being accurately reproduced. Kodak Gold was engineered to forgive every amateur error involving exposure and stereotypical skintones, yet at the expense of nearly everything else. But trying to prove anything one way or another via a web picture is largely futile. It's designedly just too crude a medium of visual communication. I don't want to rain on anyone's parade, but one needs to know the basics of color theory just to get to first base in a relevant technical discussion. In this case, it would be difficult to pin down where difficulty reproducing a standardized color chart would lie : in the initial exposure itself, in the inappropriateness of the film, in the deficiencies of the printing light, in the idiosyncrasies of the paper or process control, or in some scanning and web limitatons. Each one of these things would have to be ironed out in turn. A person can still have a lot of fun, and even achieve interesting pictures; but as the old saying goes, One man's medicine is another man's poison. And the idea that a phone can serve as some kind of color calibration device is ludicrous. Paint companies have spent millions of dollars developing and marketing that idea already, and it's about as effective as trying to chop down a redwood tree with a fingernail file, really just an advertising ruse. Not even a fifty thousand dollar color spectrophotometer can get every hue right. And it is indeed a very big step from simply balancing a print across the color wheel and learning how to saturate a whole ranges of colors while, at the same time, retaining the neutrality of others. All your ducks have to be in a row.
But the issue of excess cyan in the Ektar print might simply be lack of a warming filter like an 81A. Ektar isn't artificially warmed like Portra or other color neg films optimized for skintones, and it's difficult to post-correct for this.