RGB LEDs for color enlargers

[bernard_L]

True additive blue and true additive green do not mix to produce yellow, but block each other and produce black.

??? I would have thought they produce teal, not black. I'm probably too naïve.

 

[Ethan Brossard]

Thank you for this great thread! I recently (at the beginning of 2022) built an 8x10 LED enlarger myself, and this would have been a great resource while I was designing it, though it seems I did about the right thing, though I didn't do nearly as much research as you did. I used 450, 525, and 660 nm LEDs as well, and have not had any issues with them.

So far I've just been doing black and white printing, without the red channel of course (though it is useful for planning exposures without having to hide light sensitive papers). After reading this, I definitely want to try RA4 sometime, maybe my college has a processer hidden away somewhere I can use.

Anyway, in response to the questions about LEDs being fast enough to match dichroic heads, in my experience that's not an issue at all. If anything, the opposite is my problem. When I designed my system, I hadn't done any tests to see how much LED power would be needed, so I assumed I would want to maximise the speed as much as possible. To do this, I opted for a holographic diffuser, which has somewhere around ~80% transmission, far more than the ~30% opal glass diffusers often have. with the holographic diffuser, and a negative developed for grade 3 papers, and a 16x20 print size I had to stop the lens all the way down to f/45, and even then, found the optimum exposure at f/45 was about 2 seconds.

So, far too fast/bright to be useful. Since then I've replaced the diffuser with a more traditional one I scavenged from a broken cold light I had lying around, but based on exposure times, I'd hazard to guess that it's still brighter than my DeVere 8x10 dichroic head.

Maybe I'll set them both up and measure the brightness with my spotmeter, I'll get back to you guys if I do that.

 

[Ethan Brossard]

??? I would have thought they produce teal, not black. I'm probably too naïve.

cyan I think, but yes, I would imagine that as well

 

[albada]

[...] so I assumed I would want to maximise the speed as much as possible. To do this, I opted for a holographic diffuser, which has somewhere around ~80% transmission, far more than the ~30% opal glass diffusers often have. with the holographic diffuser, and a negative developed for grade 3 papers, and a 16x20 print size I had to stop the lens all the way down to f/45, and even then, found the optimum exposure at f/45 was about 2 seconds.

So, far too fast/bright to be useful. [...]

Can you dim your LEDs using PWM? Using a PIC18 microcontroller, with 12-bit PWM, I can dim the LEDs by up to 6 stops (1/64th intensity), while maintaining good accuracy of exposure. If you want to stay with analog electronics, a 555 chip can probably get you 3-4 stops of dimming (that was my first controller).

Anyway, please post a description and pictures of your LED-head and controller in a new thread. It's always interesting to see what folks have done.

Mark Overton

 

[DREW WILEY]

I think it would really help if some of you would simply study up on basic color theory first. Or just buy a cheap set of RGB color separation filters from an outfit like Edmund, stack or screw them together, then look at a bright light through those. If they're pure enough color (like a 29 red, 61 green, and 47B blue set), you won't see a thing, just black.

 

[Ethan Brossard]

I think it would really help if some of you would simply study up on basic color theory first. Or just buy a cheap set of RGB color separation filters from an outfit like Edmund, stack or screw them together, then look at a bright light through those. If they're pure enough color (like a 29 red, 61 green, and 47B blue set), you won't see a thing, just black.

you said addative blue and green in your post, did you mean subtractive? If so, then I agree with you

 

[koraks]

Blue is the challenge due to its large overlap with green. Measuring with a ruler on the monitor, at 450 nm, the B/G ratio is 2.8. But at the blue-peak at 486 nm, the ratio is 4.0. So I think your idea of employing a blue LED at about 486 nm is promising, because for a given exposure of blue, it will expose green the least. BTW, I got that 486 nm with a ruler as well. That max ratio of 4.0 should yield the purest hue.

Exactly! Now, part of the reason why I didn't proceed down that path initially was also my concern about losing the highest B&W contrast grade; see part of the earlier discussion in this thread. But if we sacrifice half a grade, or maybe a full grade, of B&W, perhaps we can gain a lot in hue purity. Perhaps one day I'll take the time and configure an RGB set along those lines.

Cree makes a cyan LED that is very close to 486 nm. Here's the link to it in digikey. It's 1-amp, so it's "freaking bright."

The different links you gave appear a bit contradictory, but looking at the datasheet, I see some promising devices, indeed.

We could probably make a good RA-4 enlarger using only LEDs from this one family.

Quite possibly, and I did investigate Cree and the products they had available back then, but supply issues were a problem at that time.

Unfortunately, that Cree blue-green example looks like a poor candidate for a balanced color enlarging source. It might be nice for VC paper, however.

Well, in fact, while I have some doubts about color, I would expect it would almost certainly be not an optimal choice for VC paper in any case.

As to the other remarks, Drew - yes, clear, understood and agreed upon. I follow you there.

 

[Ethan Brossard]

Can you dim your LEDs using PWM? Using a PIC18 microcontroller, with 12-bit PWM, I can dim the LEDs by up to 6 stops (1/64th intensity), while maintaining good accuracy of exposure. If you want to stay with analog electronics, a 555 chip can probably get you 3-4 stops of dimming (that was my first controller).

Anyway, please post a description and pictures of your LED-head and controller in a new thread. It's always interesting to see what folks have done.

Mark Overton

Mine does have PWM dimming as well, I can input any value between 0 and 255 for each channel, so theoretically 8 stops, but I haven't tested uniformity of the light at low values. In practice I'd guess it's similar to yours.

For my controller, I used an arduino which sends signals to a DMX decoder in the head, which powers the LEDs. DMX is a standard used in stage lighting, and the decoders are made with many different power levels, which means I can build other heads in different sizes using other decoders, or someone else can build their own head design using a DMX decoder, and the same color controlller and timer (I built a combined unit which does both) can be used with any of them.

 

[koraks]

True additive blue and true additive green do not mix to produce yellow, but block each other and produce black.

True additive green and blue will of course add to each other and remain distinct blue and green, which we perceive as cyan due to the limitations of our visus system. I'm not familiar with any system in which they would add up to yellow.

 

[Ethan Brossard]

For a few days, I've been thinking about the first graph in your article, which shows the sensitivity curves for R-G-B.

Blue is the challenge due to its large overlap with green. Measuring with a ruler on the monitor, at 450 nm, the B/G ratio is 2.8. But at the blue-peak at 486 nm, the ratio is 4.0. So I think your idea of employing a blue LED at about 486 nm is promising, because for a given exposure of blue, it will expose green the least. BTW, I got that 486 nm with a ruler as well. That max ratio of 4.0 should yield the purest hue.

The graph does not show the threshold (inertia?) of each color. If green has a higher threshold than blue, then an exposure at a ratio of 4.0 (486 nm) could produce a pure yellow.

Cree makes a cyan LED that is very close to 486 nm. Here's the link to it in digikey. It's 1-amp, so it's "freaking bright." This LED is a member of Cree's XE-G family, and here's the link. This family includes a 660 nm "photo red", and several greens. We could probably make a good RA-4 enlarger using only LEDs from this one family.

Wow, that line of LEDs looks quite nice, I'll have to do some experimenting later on. It looks like they were just released earlier this year, so still quite new to the market

 

[koraks]

So far I've just been doing black and white printing, without the red channel of course (though it is useful for planning exposures without having to hide light sensitive papers). After reading this, I definitely want to try RA4 sometime, maybe my college has a processer hidden away somewhere I can use.

You should! And let me/us know how it pans out! In particular if you run into any issues, because that's how we can take this even further.

If anything, the opposite is my problem.

Yeah, I recognize that. Well, it's mostly not a problem, really, but it's certainly true that power is just not the limitation.

Concerning PWM, addressed by yourself and also @albada: PWM is evidently the way to go with LED, and my design of course also uses it for filtering. I found that 8 bit is far insufficient for color work. In the first successful design of my head (I had moved beyond RGB COB LEDs by then), I used 12 bit PWM as that was the resolution the PCA9685 I used offered. The current version runs on an ESP32 uC which will do happily 16 bit PWM at the frequency I selected, but for convenience's sake I'm running it at 12 bit.

There's a bit more to the PWM story than just bit resolution; you'll have to start looking at the pass current of the LED arrays on a 'scope to see what you're doing, and a lot will depend on the driver topology used. I doubt there's a lot to be gained to go beyond around 14 bit dimming resolution; I don't find it necessary in any case. But 8 bit was way insufficient and only OK for B&W work. @Ethan Brossard, you may have to address this aspect if you're going to do color on your machine.

 

[Ethan Brossard]

You should! And let me/us know how it pans out! In particular if you run into any issues, because that's how we can take this even further.



Yeah, I recognize that. Well, it's mostly not a problem, really, but it's certainly true that power is just not the limitation.

Concerning PWM, addressed by yourself and also @albada: PWM is evidently the way to go with LED, and my design of course also uses it for filtering. I found that 8 bit is far insufficient for color work. In the first successful design of my head (I had moved beyond RGB COB LEDs by then), I used 12 bit PWM as that was the resolution the PCA9685 I used offered. The current version runs on an ESP32 uC which will do happily 16 bit PWM at the frequency I selected, but for convenience's sake I'm running it at 12 bit.

There's a bit more to the PWM story than just bit resolution; you'll have to start looking at the pass current of the LED arrays on a 'scope to see what you're doing, and a lot will depend on the driver topology used. I doubt there's a lot to be gained to go beyond around 14 bit dimming resolution; I don't find it necessary in any case. But 8 bit was way insufficient and only OK for B&W work. @Ethan Brossard, you may have to address this aspect if you're going to do color on your machine.

Interesting. For black and white printing I've been using the split filter method, where I do individual exposures for each color. Dimming was just when I had the holographic diffuser and couldn't dim enough with the aperture. The timer I built is an f/stop timer which gives me down to 1/12 stop adjustment increments. For RA4, I imaginine I could do similar where I flash the paper for a specified amount of time with full brightness of each color. Would that work for gaining more control over color, or is there something I'm missing?

What RGB COB LEDs (wow thats a lot of acronyms) did you use? I'm building a condenser head for my enlarger for UV enlarging using a UV COB, but I haven't done much looking into RGB ones. If they make suitable RGB COBs, it would be nice to change between UV and RGB without having to replace the condensers with the diffuser, especially if it also would allow more control like you've indicated.

 

[koraks]

For RA4, I imaginine I could do similar where I flash the paper for a specified amount of time with full brightness of each color. Would that work for gaining more control over color, or is there something I'm missing?

Nope, that would work! It's conceptually the same as the consecutive and separate exposures that were done sometimes. I mostly discarded that option because I liked the possibility for some burning & dodging. I rarely do this with color, and overdoing it tends to result in color issues, but a slight nudge here and there is nice. But I end up doing that...well, rarely, if every. So might as well do away with that requirement...

What RGB COB LEDs (wow thats a lot of acronyms) did you use?

Generic Chinese junkyard ones. Which is to say, I got more expensive ones from a German seller on eBay which turned out to be identical apart from the heatsink backing, which was way better on the German units. But they didn't appear to be different in terms of actual chips. That's BTW always a gamble; you can get LEDs from whatever source you like, but you virtually never know what actual LED die you find inside the device, because that information is not released.

If they make suitable RGB COBs

I am not aware of any RGB COB LEDs with red and blue wavelengths suitable for RA4 color. It's the route I tried first because it seemed attractive, but it turned out to be a dead end. That's the essence of my blog post: all devices with integrated R, G and B LEDs within a single device turn out to be the wrong color set for RA4 printing. I have not yet found a single exception to this rule.

 

[Ethan Brossard]

Nope, that would work! It's conceptually the same as the consecutive and separate exposures that were done sometimes. I mostly discarded that option because I liked the possibility for some burning & dodging. I rarely do this with color, and overdoing it tends to result in color issues, but a slight nudge here and there is nice. But I end up doing that...well, rarely, if every. So might as well do away with that requirement...



Generic Chinese junkyard ones. Which is to say, I got more expensive ones from a German seller on eBay which turned out to be identical apart from the heatsink backing, which was way better on the German units. But they didn't appear to be different in terms of actual chips. That's BTW always a gamble; you can get LEDs from whatever source you like, but you virtually never know what actual LED die you find inside the device, because that information is not released.


I am not aware of any RGB COB LEDs with red and blue wavelengths suitable for RA4 color. It's the route I tried first because it seemed attractive, but it turned out to be a dead end. That's the essence of my blog post: all devices with integrated R, G and B LEDs within a single device turn out to be the wrong color set for RA4 printing. I have not yet found a single exception to this rule.

Good to know. Thinking about it, I can probably make my own little board of properly colored LEDs for the condenser. Because condensers are so much more efficient than diffusion heads, it might be possible to build a bright enough head with only 1 LED of each color

 

[albada]

But 8 bit was way insufficient and only OK for B&W work. @Ethan Brossard, you may have to address this aspect if you're going to do color on your machine.

Yes! In fact, I found 8-bit to be frustrating for B&W because it only provides about 4 stops of accurate dimming. 4 stops down is 1/16th intensity, so its PWM is about 256/16 = 16, so the step-factor of PWM of 1/16 is larger than I like. A 5th stop would have a step-size of 1/8, which is too coarse.

Good news: Arduino has one 16-bit PWM with two outputs, so you can get 12-bit PWM out of it (green and blue, red would still be 8-bit). But the Arduino library does not provide access to that, so you must write registers yourself. I can give you source-code if you're interested.

@Ethan Brossard: The optical system in a condenser-lamp creates an image of the white bulb on the paper. That's why it's important to have a bulb of the correct physical dimensions in there. But an image of a LED won't work; you'll just have a bright spot on the middle of the paper.

 

[Ethan Brossard]

Yes! In fact, I found 8-bit to be frustrating for B&W because it only provides about 4 stops of accurate dimming. 4 stops down is 1/16th intensity, so its PWM is about 256/16 = 16, so the step-factor of PWM of 1/16 is larger than I like. A 5th stop would have a step-size of 1/8, which is too coarse.

Good news: Arduino has one 16-bit PWM with two outputs, so you can get 12-bit PWM out of it (green and blue, red would still be 8-bit). But the Arduino library does not provide access to that, so you must write registers yourself. I can give you source-code if you're interested.

@Ethan Brossard: The optical system in a condenser-lamp creates an image of the white bulb on the paper. That's why it's important to have a bulb of the correct physical dimensions in there. But an image of a LED won't work; you'll just have a bright spot on the middle of the paper.

I’m a bit confused by what you're saying about condenser design.

By my understanding of condensers, the only two points where the lamp would produce a sharp image of the lamp itself are at the lamp itself, and at the lens focal point where the condenser is focused. By the time that light reaches the paper, isn’t it so out of focus that it’s just even illumination?

 

[koraks]

Because condensers are so much more efficient than diffusion heads, it might be possible to build a bright enough head with only 1 LED of each color

I considered this, but there's no way to get enough power in a small enough spot across 3 colors with LEDs for a condensor system to cover from 35mm up to 4x5 and still allow decent exposure times in color as well as VC B&W (especially at lower grades) up to somewhat decent enlargements, let's say 11x14" from 35mm. At least, that's where I got stuck in the theoretical exercise. I know the stumbling in the dark blog guy (google it) did a condensor VC light source with just two LEDs, so apparently it worked for him, but that didn't have to do any color.

Arduino has one 16-bit PWM with two outputs

Which do you mean? A Mega or so?
Anyway, there's a plethora of microcontrollers out there, many of which have at least some Arduino support. Have a look at for instance blue pill or black pill boards (STM32 based) or ESP32 boards. They're generally dirt cheap and usually offer way more performance than most Arduino's. For instance on ESP32 with their Arduino core, 16bit PWM is simply accessible through analogWrite(). Just set the desired configuration on the selected GPIO and go.

And of course there's no rule that binds one to Arduino, although it can be convenient to use their IDE and C++ dialect across different hardware platforms. There's so many options to choose from these days. Heck, I've even done some small projects based on the stm8s series with the Cosmic C (not ++) compiler. Fun, if you're into it - bloody annoying if you're not.

 

[albada]

I’m a bit confused by what you're saying about condenser design.

By my understanding of condensers, the only two points where the lamp would produce a sharp image of the lamp itself are at the lamp itself, and at the lens focal point where the condenser is focused. By the time that light reaches the paper, isn’t it so out of focus that it’s just even illumination?

I have not tested this, but I remember reading that condenser systems work that way, so it's worth checking before building something. I suppose you could remove the bulb, hold a small LED with your fingers (or tape?) where the bulb was and look at what's projected on the paper. I could not locate the "stumbling in the dark" article, but if he built a condenser lamp using two LEDs, then what I read must be wrong.

I was assuming an Arduino Uno (Atmega 328p), which was my 2nd controller. My 3rd controller uses a PIC18 because I wanted to (1) avoid 8-bit PWMs and (2) get the joy of programming down to the bare metal, with no library or OS in between. I've never heard of blue/black pill boards; I'll look into those.

 

[koraks]

I could not locate the "stumbling in the dark" article

That was my fault; my memory basically sucks

Variable Contrast LED Head for Durst 138S – Condensers | Tripping Through The Dark

www.trippingthroughthedark.com

Have fun, I found it very inspiring when I first read this years ago.

I've never done any work with PICs, but I see what you mean with the bare metal experience. Can be fun, indeed! My stm8s foray was like that.
PS: I still have a soft spot for the 328, even though it's been surpassed left and right by just about anything. It's just such a nice little piece of silicon.

 

[albada]

That was my fault; my memory basically sucks

Variable Contrast LED Head for Durst 138S – Condensers | Tripping Through The Dark

www.trippingthroughthedark.com

The author put an opal glass between the LEDs and condensers, thus simulating the effect of an opal bulb. He states that "the condenser setup relies on focusing an image of the light bulb at the nodal point of the lens", hence the need for the opal glass.
He also created a point-source system with no opal glass, which is interesting.

I've never done any work with PICs, but I see what you mean with the bare metal experience. Can be fun, indeed! My stm8s foray was like that.
PS: I still have a soft spot for the 328, even though it's been surpassed left and right by just about anything. It's just such a nice little piece of silicon.

I also needed EEPROM, and the STM32 and ESP32 have none, so it's emulated using software and flash, which makes me worry about longevity because flash wears out more quickly than EEPROM. The EEPROM is used for:

LED PWM calibration-tables (one per color).​

Paper curves (a range of contrasts for each paper).​

Storing miscellaneous settings.​

Storing exposure-settings for later recall (handy for multi-stage exposures, and for making copies).​

1024 bytes of EEPROM is cramped, but that's the largest I could find in these 8-bit microcontrollers.

 

[mshchem]

Thank you for this great thread! I recently (at the beginning of 2022) built an 8x10 LED enlarger myself, and this would have been a great resource while I was designing it, though it seems I did about the right thing, though I didn't do nearly as much research as you did. I used 450, 525, and 660 nm LEDs as well, and have not had any issues with them.

So far I've just been doing black and white printing, without the red channel of course (though it is useful for planning exposures without having to hide light sensitive papers). After reading this, I definitely want to try RA4 sometime, maybe my college has a processer hidden away somewhere I can use.

Anyway, in response to the questions about LEDs being fast enough to match dichroic heads, in my experience that's not an issue at all. If anything, the opposite is my problem. When I designed my system, I hadn't done any tests to see how much LED power would be needed, so I assumed I would want to maximise the speed as much as possible. To do this, I opted for a holographic diffuser, which has somewhere around ~80% transmission, far more than the ~30% opal glass diffusers often have. with the holographic diffuser, and a negative developed for grade 3 papers, and a 16x20 print size I had to stop the lens all the way down to f/45, and even then, found the optimum exposure at f/45 was about 2 seconds.

So, far too fast/bright to be useful. Since then I've replaced the diffuser with a more traditional one I scavenged from a broken cold light I had lying around, but based on exposure times, I'd hazard to guess that it's still brighter than my DeVere 8x10 dichroic head.

Maybe I'll set them both up and measure the brightness with my spotmeter, I'll get back to you guys if I do that.

I've seen your project on the Large format forum. I would be very interested in RA4 tests from your machine. Keep us informed. I think folks here would like to see a thread on your enlarger.
Best Regards Mike

 

[albada]

Here's an article in Shutterbug magazine that specifies the Wratten filter numbers and their wavelengths used in RA4. The article doesn't actually say it's RA4, but what else could it be? Green and blue agree with Koraks' article, but red is specified as Wratten 25 with a wavelength of 610 nm, which seems implausibly low to me.

 

[koraks]

I also needed EEPROM, and the STM32 and ESP32 have none, so it's emulated using software and flash

Yeah, that's true and it's what I use on the ESP32. Let's see how it pans out...I'm not too concerned at this point about the longevity issue. I don't do many writes to the flash and the ones that happen aren't very crucial. I use flash for basically the same as you. It's quite a sizeable chunk of data if I add it all up, to be honest, but my machine has a pretty broad set of variables to store for a silly enlarger!

red is specified as Wratten 25 with a wavelength of 610 nm, which seems implausibly low to me.

Yes, and if you trace the source for "my" wratten selection, you'll find that Photo Engineer specifically mentioned they did not prefer the #25, but he didn't mention why I think. Perhaps if you root around a bit through his posts you'll find some clues.

 

[bernard_L]

I think it would really help if some of you would simply study up on basic color theory first. Or just buy a cheap set of RGB color separation filters from an outfit like Edmund, stack or screw them together, then look at a bright light through those. If they're pure enough color (like a 29 red, 61 green, and 47B blue set), you won't see a thing, just black.

You wrote:

True additive blue and true additive green do not mix to produce yellow, but block each other and produce black. I

Had you written "blue and green filters" and instead of "mix", "stack" indeed I would understand without even needing to go back to basic theory.

On the other hand, your using in the same sentence "true additive" and "mixing" brings up the Philips additive system: PCS150, PCS2000.

Ollinger's Guide to Photographic Enlargers: Philips Enlargers

A short guide to and description of Philips enlargers.

www.jollinger.com

Finally, it takes true additive red and green mixed together to produce yellow; colorimetry 101.

 

[koraks]

brings up the Philips additive system: PCS150, PCS2000.

Yes, I briefly mentioned the 2000 in my blog post. I never inspected one, but AFAIK they used dichroic filters and a single (?) light source with a mirror system to create the R, G and B components. I've never used one, so I can't comment on how well they work. The company did have a reputation for decent quality products, at least back then.