Are CC values linear?

[BetterSense]

I'm designing an additive enlarger light source with Red Green and Blue LEDs.

I'm trying to decide if 8 bits (255 levels) of LED brightness is enough.

Subtractive color heads often have knobs with values 0-130. The question is, are those 130 values distributed somewhat linearly? I will calibrate my light source once I have it built, but I need to know how much precision I will need for dimming each channel.

 

[polyglot]

Sorry, no. It depends on the enlarger but they are often geometric and approximately 0.01D each (10-0.01 = 0.977). That means that 30 of them (10-0.3 = 0.5) is about one stop of attenuation and 120 is about 4 stops. So 8 bits is certainly enough to represent the dynamic range, but you need to figure out whether you will have sufficient precision at the lowest levels of illumination in each channel, i.e. at extreme filtration settings with big colour ratios.

Presumably you're also building the timer which will control the 3 intensities and durations of illumination? Using PWM for the intensity control? If have only 8 bit PWM and you need more resolution at the lowest exposures, you can expose one channel at higher intensity (therefore more precision in the PWM) for a shorter time. However, that will cause problems with dodge/burn operations.

Most cheap micros have 16-bit counters that would give you high precision PWM if you wanted it. Even an Arduino (ATmega328) has one and you can PWM with it if you write a bit of native AVR code - I can help you with that if you want - which gets you 16-bit PWM at 244Hz but only one or two channels.

 

[Kawaiithulhu]

You're already looking at one bazillion color combinations (256^3 = 16,777,216 to be pedantic) so I'm not sure what good extra precision will do. Though Polyglot sounds like he has some pertinent experience!

The tricky part will be to give the software a good, usable user interface.

 

[polyglot]

The total number of available colours is not the problem, it's how far apart they are in the final output gamut. If a minimal change in the colour produces too much change in the image, it's got insufficient precision.

On further thought though, assuming colour printing, you tend to have at most 30CC difference so (assuming 8 bit PWM) that leaves you with 7 bits of precision in the lowest intensity channel, which is about 0.5CC precision, which is plenty.

 

[AgX]

It depends on the enlarger but they are often geometric and approximately 0.01D each

Those enlarger filter values are logarithmic not geometric.

 

[Chan Tran]

They are logarithmic but instead of doing brightness level you can do with time. Exposure for the R,G,B lights of different times.

 

[AgX]

I'm designing an additive enlarger light source with Red Green and Blue LEDs.

Wouldn't just buying that 35mm Philips additive enlarger with incandescant lighting yield an outcome?

 

[BetterSense]

The tricky part will be to give the software a good, usable user interface[/QUOTE

I don't do GUI's. I do knobs and gages and buttons and switches.

I know the atmega 328 has dual output compare registers on timer1. I can use those for green and blue. But I would rather use bytes internally if I can. If a cc is .1D I can calculate from there. For color printing it seems like I will be in the middle or upper range all the time...not like a subtractive enlarger at all where you leave one channel at 0.

 

[Rudeofus]

Since you will most likely need more than one LED per color, you could trivially increase dynamic range by turning on/off the whole array, or only a subset of your LED array. 8 bits of PWM plus 4 bits from switching all or only every 16th LED will give you an 1:4096 range, or a density difference of 3.6, which is way more than any subtractive color head will ever give you.

One thing you may want to look out for is actual light output from individual LEDs. There is a good chance that you will see quite a variation in light output between different parts, even from the same batch.

 

[polyglot]

Chan: that makes dodging & burning unreasonably difficult because you can't get nice smooth blends by moving the tool around.

The tricky part will be to give the software a good, usable user interface

I don't do GUI's. I do knobs and gages and buttons and switches.

I know the atmega 328 has dual output compare registers on timer1. I can use those for green and blue. But I would rather use bytes internally if I can. If a cc is .1D I can calculate from there. For color printing it seems like I will be in the middle or upper range all the time...not like a subtractive enlarger at all where you leave one channel at 0.

0.01D. You would definitely have too-coarse control with 0.1D (1/3 stop!).

Knobs & text displays are easy on modern micros, certainly easier than driving a gauge accurately. I would suggest that a rotary encoder (for changing values like channel intensities or exposure times), small LCD (to display current settings), a few buttons (to select which channel and/or time you're chagning) is an easy approach. There's absolutely no need to keep things in just one byte though because 16 bit, 32 bit and even floating-point arithmetic is supported by gcc with zero extra effort on your part. It runs a little slower but for the purposes you have here, the micro is still about 10000x faster than necessary. Hell, you could write the whole program using double (64-bit IEEE floating point) and it would still run fine unless you trip yourself up on rounding modes.

 

[David Grenet]

There's absolutely no need to keep things in just one byte though because 16 bit, 32 bit and even floating-point arithmetic is supported by gcc with zero extra effort on your part. It runs a little slower but for the purposes you have here, the micro is still about 10000x faster than necessary. Hell, you could write the whole program using double (64-bit IEEE floating point) and it would still run fine unless you trip yourself up on rounding modes.

This. Exactly this. Let go of the temptation to keep track of every register and put some faith in your compiler.

 

[ME Super]

Exactly. On modern compilers, a signed integer (the default for an integer) has a range of -2,147,483,648 to +2,147,483,647. An unsigned integer has a range of 0 to 4,294,967,296. Both signed and unsigned integers are 4 bytes long. One of these for each of RGB is more than fine enough control. Your average computer monitor uses 8 bits (one byte) for each of RGB, so really a 32-bit integer for each channel is overkill.

 

[DREW WILEY]

Good luck. LED's aren't anywhere near close enough to correct spectral options to make an additive colorhead with at this point in time.
Maybe someday. Additive heads are a tricky engineering project. The rheostatic model like that little toy Phillips of yesteryear probably didn't
put out enough light to wake up an aphid. I know folks who have built huge rheostatic additive heads with banks of halogen bulbs and the main issue was extreme heat and very long exp times - a bad combination which just feeds its own pitfall. Beseler-Minolta marketing an additive head using small xenon flashtubes - OK for little stuff, but sequential, so risks color fringing with dodge/burn, and too weak for many pro applications. Durst was planning on marketing a high-output additive halogen head for their pro 8x10 enlargers, but then shut down the
true pro enlarger division completely. Again, the problem was extreme heat combined with very high utility bills and frequent expensive filter
replacements. My own additive enlargers run cool and give very consistent clean color, but rely on some temperamental complex feedback
circuitry. But that was kinda state of the art when I built them. Nowadays you could hypothetically control the schizophrenia of EMI on
multiple circuits operating at once by employing what is termed sine wave current pulsing, just like they do for fancy stage productions and
rock concert light shows, but then you'd be tethered to computer and software control, and the risk of that going obsolete. No easy game.
I had to work on a very tight budget, doing all the machine shop work myself, and that last colorhead still cost me 15K in materials and three
years on n' off labor. The colorhead is so big that I have to use a block and tackle to replace any of the bulbs. But it does give wonderful
color with my 8x10 film.

 

[BetterSense]

Don't they have red green and blue LEDS?

 

[DREW WILEY]

To get anything resembling accurate color reproduction the colors would have to be spot on. This isn't anything like taking white light and selectively filtering it as in ordinary subtractive printing. Your RGB's have to each be narrow bandwidth, and then be capable of being controlled for balance. If your LEDS are a bit wide in spectral properties, then you could filter trim then, but if so, you're right back to a complicated expensive project, so why bother with them to begin with? If somebody has the time to experiment, more power to them! But I think it's rather premature to consider LED's as an acceptable alternative for color printing, at least simultaneously, all at once. You might be able to rig up some source and filter pack that would allow you to do three successive exposures. But then you can't dodge or burn, and it is hell to establish color balance to begin with. But the hypothetical possibility of using LED's shouldn't be abandoned. I just might be awhile.

 

[BetterSense]

Do you know the specific frequencies that must be used for the R, G, and B LEDs, and what an acceptable spectral linewidth would be?

Does color paper have very narrow absorbtion spectra (i.e, it would be good to know the speed of each layer, as a function of wavelength, and see how broad or narrow those R, G, B speed spikes are).

In subtractive printing, the R, G, and B components coming through the filters are each quite broad-band, so I didn't think that it would be so important to hit the exact wavelength with the LEDs. They are going to be calibrated anyhow.

 

[DREW WILEY]

You can't calibrate something that's not there to begin with. You'd need to study precise published spectrograms for each type of bulb, along
with any change in performance due to temp change. What appears blue or red or green to the eye might not be the same way a printing
paper see it. As far as exactly how narrow things need to be, that a much bigger subject than I can give a brief answer to. But you have to essentially be as tight as a good set of RGB separation filters, if not better. For example, if you use a combination of any two of three 47B blue, 29 red, and 58 green filters, you get sheer black. Substitute 23 red and a bit of light might get thru. With subtractive CMY filtration there is always some leftoever white light spilling past, hence contaminating all your respective dye layers to a slight degree. That'w why additive printing is cleaner. But beyond that, the subject get very complicated very fast. Additive printing has long been the holy grail of
color enlarging. But nobody spends any R&D money anymore seeking the holy grail. They've got "improved" ideas, like expensive fussy laser
printers which do use RGB, and complicated inkjet patents already running out their ears. The appeal of LED is that it can run relatively cool.
High temps are the traditional curse of large additive colorheads, and even big subtractive pro colorheads. My electric bill for my good ole
Durst color mural enlarger was more than for the rest of my property combined. I won't go into detail here how I got around that with my own colorhead design, but it involved a lot of circuits which don't always get along with each other. That's a routine problem in complex
lighting situations, even in large buildings where different branches have to be programmed independently. Worse, anything in the vicinity
that is solid state can go bonkers too. I'm certainly no electrical engineer - not even close. But I did have some help from them.

 

[DREW WILEY]

Well ... I realize that answer probably didn't satisfy you, BetterSense, so let me simply add that if you print narrowband, you get a very different visual result than you do being broader, or than with any subtractive colorhead. Color balancing your paper is even different, and in
the end, a lot more accurate. And for additive to be worth it, you're outright wasting your time if you are not dead-on in terms of spectral
positions. Why bother? Just buy an ordinary CMY colorhead.

 

[BetterSense]

You'd need to study precise published spectrograms for each type of bulb

I'm talking about using LEDs, which have easily obtainable spectrograms published. I'm not sure where to get data on the speed vs. wavelength characteristics of color paper, however.

The fact that they do use colored lasers in LightJet printing is good indication that LEDs will also work. LEDs aren't quite as narrowband as lasers, but still closer to lasers than RGB filters (I would think).

you have to essentially be as tight as a good set of RGB separation filters, if not better

That should be very easy with LEDs.

 

[DREW WILEY]

They can't even make a high enough color compound LED bulbs at this point in time, using multiple colored bulbs inside, to equal good color
matching fluorescents - which aren't any good for color printing at all. No it should NOT be easy with LED's. And NO... lasers are not a good
comparison, because there has already been mountains of R&D done to make them suitable for tricolor printing, and some very expensive
devices to complete the performance. As far as anything goes that I'm aware of, there is simply not a suitable LED set out there. I really wouldn't worry about specific paper sensitivities at the beginning. Printing true additive analog is just like printing from RGB lasers in that respect. But there's a whole set of engineering steps involved to getting a truly workable unit, so expect a lot of serious prototyping with some failures. It's just a matter of time till it happens. Figure out the electronics, diffusion, etc etc. in advance. Test it. But this is a much
fusssier game than making and LED blue-green VC printing paper source.

 

[Mr Bill]

I'm talking about using LEDs, which have easily obtainable spectrograms published. I'm not sure where to get data on the speed vs. wavelength characteristics of color paper, however.

You could try some color paper data sheets, such as: http://www.kodak.com/global/en/professional/support/techPubs/e4070/e4070.pdf

The fact that they do use colored lasers in LightJet printing is good indication that LEDs will also work. LEDs aren't quite as narrowband as lasers, but still closer to lasers than RGB filters (I would think).

In the photofinishing industry, some of the early silver-halide digital printers I worked with were LED units from Kodak. But this was a different time, using different papers. Years went into the optimization of color papers for exposure by digital devices.

Something else worth pointing out: when you are making exposures from film, you want a light source which can be readily attenuated by the film dyes; this is a vastly different situation than exposing a digital image.

That should be very easy with LEDs.

One has to wonder, if it's so easy, why aren't they commercially made? It may simply be that there's not much of a market, but perhaps more than that.

I don't have direct experience in choosing light sources for printing, but I have a few thoughts on the matter, if you're interested.

 

[polyglot]

They can't even make a high enough color compound LED bulbs at this point in time, using multiple colored bulbs inside, to equal good color
matching fluorescents - which aren't any good for color printing at all. No it should NOT be easy with LED's. And NO... lasers are not a good

You're referring to LEDs having low CRI, which is an issue only for the initial capture - because photo subjects like people are broadband in their colouring - and completely irrelevant to printing because a paper has only 3 or 4 layers of dye in it. The final image quality is a matter of balancing those 3 or 4 channels and that is ALL. It doesn't matter how you expose them, as long as they are exposed in the correct ratio according to the information in the negative (or digital file).

This is why digital printing with LEDs or lasers is relatively easy; you have one LED per layer and you calibrate your LEDs against the paper. Once you know how much light from each is required, that's a simple lookup table to apply from the source RGB or CMYK file, and you can make perfect exposures every time.

Doing LED exposures via a C41 negative is slightly more complicated: you still have one LED per colour channel on the paper, but you need to make sure that the LED is right in the middle of the absorption band of the relevant dye in the C41 film. If you get that right though and calibrate your system, it should not be a huge problem. You're performing a linear translation between three dye layers in a negative to three other dye layers in the paper, using three monochromatic lights to carry the information between the two.

Problems arise though where the negative has two dyes targeted at a single layer in the paper, for example the orange mask. The blue-sensitive yellow-coloured dye notches out one wavelength and the mask notches out a slightly different wavelength, both of which have an effect on the formation of blue parts in the print. Your LED spectral form must take BOTH of these dyes into account otherwise you will get uncorrectable colour contamination.

Still. I reckon the OP should just bloody try it and see how it goes. It stands a good chance of working, even if experimentation with a handful of different brands of LEDs is required.

 

[MartinP]

As someone who has never tried additive RA4 printing [sound-effect: hollow laughter], only subtractive, and having read that major problems with traditional style light-sources are heat and power . . . how about broadband ("white") leds behind RGB filters? It couldn't be worse than halogen bulbs for domestic-sized prints, not murals, or could it? Presumably one would make the light, filter the light, mix the light and then use the light - in the usual way?

The poor-man's version could have one light-source, probably multiple big led's in one fitting, and move the filters in-and-out for three exposures while the deluxe version (allowing burning in etc) could have simultaneous exposures through the three filters using one light source per filter. The tricks may include matching the "white" led's to the requirements of the narrow filter and also consistency of the output of the led's. The idea almost demands a device for measuring, and graphing, the colour spectrum of a light source but, without even trying very hard, I know at least one company who do that as part of their standard testing work so it is not insurmountable.

It would make an interesting project, but personally, I'll continue with my occasional bursts of enthusiasm for subtractive RA4 . . .

 

[DREW WILEY]

Talk comes cheap. Ever head that old saying? Now go build something that really works, and you might discover that some of you guys don't even recognize more than 5% of the questions you need to ask before presuming to provide the solution. But basically, what is your motive?
Save money? You're sure not going to do that. Get a nice flat panel instead of a big box-like colorhead? Not so fast. You still have to diffuse'
the thing. Additive printing takes a LOT of light. If you want a panel, you'll need a whole bank of them. LED's are not continuous spectrum,
so you'd have to trim filter ever damn one of the things. Do you realize green printing lasers are not primarily green, but are strongly
filtered to allow only a very narrow band of green to pass. Similarly, once you trim a near-miss LED, it might have just a fraction of its original output. Now if you just want a challenge, or a fun project, I'm rooting for you the whole way. But based upon a few of the previous
posts, you don't even understand the difference between correcting additive versus subtractive light. I've been printing additive for over twenty five years. Yeah, you might get lucky and land on something that will print this or that film/paper combination so-so; but will it be
good for anything else? Different film/paper combinations behave very differently. My additive heads have been used for direct positive
Cibachromes, for RA4 prints from all kinds of different negs, for very critical dupes, even for fussy color separation negs for dye transfer
printing or things like that - with almost ideal performance in every case, and far less work in calibration than an ordinary CMY subtractive source. The penalty? An engineering headache.

 

[BetterSense]

Drew,

I didn't want to get in an argument, but you are either wrong or not-even-wrong in several ways. For what it's worth, I am an engineer for Cree, and I know just a little bit about LEDs.

Additive printing takes a LOT of light.

Actually, it takes exactly as much light as any other way of printing. The amount of light needed depends on the speed of the paper.

The stone-age technique of using incandescent bulbs and then filtering out most of the light to obtain R, G, B light sources is definitely extremely inefficient, so you do need to start with a huge amount of light so you have some left after you throw most of it away. This does not apply when using LEDs, which do not waste power by outputting unneeded wavelengths, and are very much more efficient, even for the wavelengths they do produce.

If you want a panel, you'll need a whole bank of them.

Probably; so what? The good thing is that LEDs are cheap, so you can use lots of them. The ability to use a "whole bank of them" is a feature vs. mixing chambers and condenser lenses and other complicated optics used with incandescent light sources.

LED's are not continuous spectrum

Yes, and that's also the point. That's the *whole reason* to use LEDs. The impetus behind additive printing vs. subtractive printing in the first place is to obtain *less continuous* light with *more narrowband* R, G, B light sources, to avoid crossover. You aren't wrong when you point out that LEDs are narrowband but your point escapes me.

Do you realize green printing lasers are not primarily green, but are strongly
filtered to allow only a very narrow band of green to pass.

This is sort of BS. A laser is a laser. The noteworthy characteristics of laser light are that the output is 1) narrow-band and 2) coherent. It's true that some green lasers are frequency-doubled from lower frequencies, but that's neither here nor there.

The primary, interesting problems with building a RGB LED colorhead for use in C41/RA4 printing will have nothing to do with the hardware. The hardware implementation will only be easier due to the use of LEDs instead of continuous light sources. The interesting problems will be related to the dye absorbtion curves of the negative film+mask and how those curves will interact with a narrowband LED RGB light source, compared to the "continuous, but RGB-adjustable" spectrum from a subtractive enlarger, or an additive enlarger that achieves R, G, B light sources with output spectra narrower than subtractive enlargers but nevertheless broader than LEDs. If existing additive enlargers are better than subtractive ones due to having "purer" colors, then an LED additive enlarger should be better still. But it would have to be tested to see, since the C41/RA4 system can be assumed to be engineered for subtractive light filtering techniques (such as minilabs that sequentially expose white light successively filtered of R, G, B by way of C, M, Y filters).