DSLR Scanner: Light Sources

[Peter J. De Smidt]

DIYS (Do It Yourself Scanner—pronounced like ‘dice’)--Light Source Thread

Frank Pertronio started this project by suggesting that someone come up with an affordable and contemporary drum scanner, as there is currently huge gap in price and quality between consumer and professional scanners. Domaz suggested using APS-C sensors and using them to take samples of the film, similar to what Gigapan does with large stitched mosaic images. This lead to talk about making a copy stand scanning system using a dslr, a light source and a movable negative stage. Both horizontal and vertical prototypes have been made, or are in the process of being made.

The original thread (http://www.largeformatphotography.info/forum/showthread.php?84769-Making-a-scanner-with-a-DSLR) has become very long and unwieldy. As a result, I’m creating some new specialized threads for future project development.

The new build threads are:
Camera-Supports-and-Positioning (http://www.largeformatphotography.info/forum/showthread.php?87537-DSLR-Scanner-Camera-Supports-and-Positioning),
Lenses (http://www.largeformatphotography.info/forum/showthread.php?87538-DSLR-Scanner-Lenses),
Negative-Stages (http://www.largeformatphotography.info/forum/showthread.php?87539-DSLR-Scanner-Negative-Stages),
Light Sources (http://www.largeformatphotography.info/forum/showthread.php?87536-DSLR-Scanner-Light-Sources),
Stitching-and-Blending-of-Images (http://www.largeformatphotography.info/forum/showthread.php?87540-DSLR-Scanner-Stitching-and-Blending-of-Images),
Cameras-and-Camera-Control-Software (http://www.largeformatphotography.info/forum/showthread.php?87541-DSLR-SCanner-Cameras-and-Camera-Control-Software).
Workflow (http://www.largeformatphotography.info/forum/showthread.php?87545-DSLR-Scanner-Workflow).

These threads are only for positive contributions to the development in the area in question. The project may not succeed, but we’re going to find that out by trying it. But we are not unkind. As the original thread showed, some people have an overpowering urge to say negative things about the project. I’ve created a thread just for this purpose. Please post your negative comments about the project here (http://www.largeformatphotography.info/forum/newthread.php?do=newthread&f=17).

I would like to thank everyone who makes, or has made, a positive contribution to this project!

I'll be summarizing the posts from the original thread about light sources here soon.

[Peter J. De Smidt]

The light source can range from diffuse to a collimated point source. Nathan Potter has pointed out that the latter might allow higher resolution and contrast. Ideally, the source has a high CRI, i.e. it's as similar to standard daylight as possible, high brightness and low heat. It could be based on flash, LED, plasma, fluorescent, or incandescent source.

My first Light Source Prototype 1 (LSP-1), a diffusion source, looks like:
http://i955.photobucket.com/albums/ae37/peterdesmidt/LSP1_NoLight.jpg

http://i955.photobucket.com/albums/ae37/peterdesmidt/LSP1_LED.jpg

http://i955.photobucket.com/albums/ae37/peterdesmidt/LSP1_Flash.jpg

http://i955.photobucket.com/albums/ae37/peterdesmidt/lightsourcedrawing.jpg

I had some trouble with evenness, and so I move on to LSP-2, which is based off of a De Vere light mixing box for a color head for a 504 enlarger.
Unlike LSP-1, there's not a 45* panel opposite the light source. Instead, the "reflector" panel is at a much smaller incline, just enough that on the end opposite the light source the reflector comes up exactly the to the level of the opening through which the light enters the box. The idea, I expect, is to spread the light from the source over as much of the reflector as possible. The De Vere uses a profiled piece of diffusion plastic to even out the light. Instead of that, about 1 inch above the high point of the reflector sheet, I put a sheet of diffusion. Another inch up (or so) there's the final sheet of diffusion plastic. The lighted opening is bigger than before, and I've fixed everything in place, so that the whole apparatus can't shift when I attach the flash.

Some time soon I'll add some pictures and a diagram.

An interesting high CRI led source is: http://www.cree.com/products/modules_lmr4.asp

Brian Miller suggested that we could take a picture of the light source without film and apply it as a differential map to the image files. This would potentially even out the response from non-perfect source.

[Peter J. De Smidt]

http://i955.photobucket.com/albums/ae37/peterdesmidt/LSP-2.jpg

[Drew Wiley]

Why not just use a colorhead with its own diffusion chamber? This was done for copy work
in days of yore. An additive head would be preferable, one with feedback circuitry to keep
the illumination level constant. Color quality and the ability to fine-tune it would be superior
to any conventional light source. The trickier part of this is "low heat". There are ways to
do this, but a bit complicated. Not simply a matter of a fan. Or just project the light from
an enlarger, thru the image onto your capture device. Or an HMI source if one of these
turned up used at reasonable price. LED's are still pretty bad for color quality, way behind
color matching fluorescent tubes which can be found up to CRI 98.

[Peter J. De Smidt]

As you say, heat is the reason that I don't use my color heads. And in the De Vere's case (I have a De Vere and a Philips color head) vibration due to the fan would be an issue.

Vibration and heat would be less of an issue with a horizontal configuration, ala Mr. Denney's approach.

The LED module I linked to earlier is available in a module that has a CRI of greater than 90.

[Drew Wiley]

I'm rather interested in how these LED and CFL sources evolve, but mainly from a print
display standpoint, particularly since they are now required by local bldg codes in commercial applications. Not happy about that, but hopefully the display market will drive
better color performance. Getting a lightbox to even out might require more than a diffuser.
Multiple diffusers help but also reduce illumination as the Lambertian factor goes up, and
also affect color temp. One secret weapon I have found is called a linear array fresnel
(completely different animal from a conventional fresnel), which will diffuse far more evenly
with less loss than simple frosted panels, but at greater cost.

[Peter J. De Smidt]

I wasn't able to find any pictures of that type of linear array Fresnel with a brief search. Is it similar to the type of add-on diffusers that were available for something like a Vivitar 283 flash? I have a set with ones of various colors. The have a grid like array of small pyramids on them.

[jon.oman]

It would be nice for theses threads to be made 'sticky'.........

[Mike Anderson]

It would be nice for theses threads to be made 'sticky'.........

Or just a master index thread with links to the others to not take up so much room at the top.

[AJSJones]

Just noticed this set of threads!
I have a Rosco LitePad 6x6 and it is described as ~6000K with a CRI of 93 (http://www.rosco.com/canada/images/LitepadthetrphotoCDN1.pdf). I put a 3mm sheet of "white" plastic from TapPlastics (eyeballed as "neutral") as a diffuser and got very nice uniformity of both luminance and chrominance. I don't know how to measure CRI but I do have an IT8 from Agfa from years ago - it's piece of 35mm slide (shot with a Canon 5D2 and 100 macro). With an image of only the litepad, the values of a full exposure were from 240-251 in all three channels after WB on the middle of gray wedge. See the image below. I'm sure I can build a profile out of this :D The whole thing sits on a XY positioner (here (http://www.ebay.com/itm/4W-Macro-Focusing-Rail-Slider-Quick-Release-Plate-K0G-/160637691388?pt=AU_TripodsMonopods&hash=item2566c0a5fc#ht_3757wt_1141)is an example). The litepad has a bubble level (it stays centered as the XY go through their range) and the camera a double bubble, but I've not measured anything yet.
I have just started playing with combining exposures as per this tutorial, (http://jtrujillo.net/qpix/) as an early attempt to get the shadow noise down and it works well. Just need to set up an action that reproducibly combines each pair into one before stitching them together.
I have not finished the project but it seems on track so far, and I thought I'd share some info now. If I finish it, I will write it up a bit more.... (and maybe put the various parts in the appropriate threads)
http://www.fototime.com/FFDF6638FC4F455/standard.jpgLarger version is here (http://www.fototime.com/FFDF6638FC4F455/orig.jpg)

[Peter J. De Smidt]

The Rosco LitePad looks interesting. A 3"x3" version is about $50. The main thing will be if you get enough light. With a continuous light exposure, we probably want to avoid shutter speeds from about 1/15th of a second down to a second, as these are the most susceptible to vibration.

Near the beginning of the main thread, I showed a stepwedge photographed with a dslr+macro lens made with one exposure, and one with blended exposures. There are certainly gains to be had there, if needed. While most bw negatives would fit within the dynamic range of one exposure, I just scanned a tech pan negative with no problems, digital images are noisiest in the shadows, and with bw negatives this means that the noisiest areas will be the highlights when the image is inverted. As a result, some type of multi-exposure blending could be very useful, especially in images with lots of very delicate highlights.

[AJSJones]

I use mirror lock-up and 10 sec timer - my camera stand is not the greatest and exposures are indeed down in that range where I'd normally do that for taking pictures. I blended a 1 sec exposure with a 1/15 sec exposure and that dramatically reduced the shadow noise on the test (from a Provia 4x5 slide where the highlights were just not blown in the DSLR and were almost, but not quite blown on the slide). The result was way better dynamic range than a single scan from my AgfaScan Duo T2500 that broke down, and the detail captured depends on how many shots I stitch - I can get it all with about a 12 shot stitch, possibly fewer. That was the next stage in the project...

[Peter J. De Smidt]

Sounds good. I haven't down any work with slides yet.

[rdenney]

Well, I had a "duh!" moment, or two.

I reported last week that I had had no luck with the slide projector as a light source. I was giving a presentation in another city this week, and looked into the projector light, and, well, DUH!, it was a pinpoint of light. So, I think to use the projector as a source, it needs to be a good distance away (to make it as much a pinpoint as possible), and then a condenser needs to be close to the film, maybe right behind it. Well, DUH!, that's the way condenser enlargers do it. I had thought the condenser in the slide projector would be adequate to achieve that, but (1) it's not nearly enough of a condenser and (2) it's small and close to the bulb. When looking into the front of the (turned-off) projector, I can see the ELM bulb reflector in its entirety. And the condenser is too weak. It really does need to be a point source with enough of a condenser to collimate the light.

So, in response to the above, I looked into the Rosco Litepads. For $50, the 3"x3" panel looked interesting. Then, when perusing reviews of it, someone commented that it looked like the standard LED back-light arrangement for computer monitor.

Well, DUH! I have computer monitors hanging from the ceiling. So, I took my work laptop, which has a very bright LED backlight, and set it on edge behind my negative carrier, opened Notepad and maximized it to get a white display (or white enough for testing), and made some tests. And the tests look a lot better. I'm strongly tempted to task an old computer to displaying a fall-off correction pattern on the screen of an old monitor.

At the f/11 that works best with my best lens, I get about 3 or 4% falloff with an exposure of about half a second. I suspect the fall-off is sensor-related--the falloff pattern looks the same on all lenses, being a little off-center no matter how I position the monitor panel.

I learned a little about my lenses, too--in the lens thread. And I learned a little about sharpening and the effects of the anti-aliasing filter, also in the lens thread.

Rick "not quite ready to attempt stitching just yet" Denney

[Peter J. De Smidt]

Sounds like you're making some good progress.

[peter ramm]

Rick, have you considered trying this type of LED back light?

http://www.edmundoptics.com/products/displayproduct.cfm?productid=3358

Can't remember if it has come up here or not. They work pretty well imaging film, if you can live with the wavelength emission. We played with these and with fiber optic plates trying to get even lighting. In the end, we used a custom fluorescent that we had on hand anyway.

Otherwise, looks like you are heading towards Kohler with the projection system.

[Peter J. De Smidt]

Those are nice but a little pricy, and I wish they listed a CRI rating. Daniel has been investigating a similar, and much cheaper, approach. His current panel should work well for BW, and a high cri panel is in the works.

With continuous lighting, we really want to avoid exposure times in the 2-1/15th of a second range, as those are most susceptible to vibration.

Note that a very experienced coin photographer has found that using a canon model with an electronic first curtain shutter significantly helps with image sharpness even at a low 1x magnification. For more on Canon's low vibration shutter, see: http://krebsmicro.com/Canon_EFSC/index.html

[rdenney]

Rick, have you considered trying this type of LED back light?

A little pricey for me, too.

But I'm sorta liking the idea of using a computer display, because of the possibility of projecting a correction pattern on it. And the correction pattern can be made by inverting a photo with no negative in the stage and displaying it at the appropriate size on the display used for illumination. Should be able to dial in perfect illumination.

It does not solve the problem of shutter speeds in the danger range. My Canon doesn't have the electronic first curtain, but the opening curtain makes no discernible movement that I can detect. It's a custom function on my camera to turn the self-timer into a timer that raises the mirror, waits three seconds, and then releases the shutter. I see no indication of camera movement as a cause of fuzziness.

Instead of setting a laptop on edge, I need to take a spare monitor and do something a little more stable and repeatable, so that I can experiment further. For a diffused light source, though, I think this is the way to go.

For a collimated source, that's another matter. There, I think an enlarger head is the cheap way to go. But I'm not yet motivated to dig that deeply in my storage room--I know how many "might as wells" will be attached to that effort, unless I do it on the sly, which isn't easy.

I also need to experiment with exposure levels. I've been photographing to put the peak in the middle of the histogram, which is fine for negatives, but which may exaggerate lighting variability when photographing the light source without a negative. I will eventually need to set the exposure to just sit under 255 (on an 8-bit scale) with no negative, or, at most, through film base+fog. That will provide the maximum dynamic range, I think.

Rick "whose experiments are continuing" Denney

[ludvig friberg]

Perhaps this is making it more complicated that it needs to be, but if we use a flying spot perhaps with a white laser or a fast smooth wipe with a line laser we would get rid of flaring. I have no idea how difficult or expensive that would be. But In a way it would turn the areasensor in the dslr to a linesensor.

Ludvig

[rdenney]

Perhaps this is making it more complicated that it needs to be, but if we use a flying spot perhaps with a white laser or a fast smooth wipe with a line laser we would get rid of flaring. I have no idea how difficult or expensive that would be. But In a way it would turn the areasensor in the dslr to a linesensor.

Ludvig

Yes, I think that would be best. But now we'd have to have two movement apparatuses, one for the film and another for the laser. And the laser apparatus would have to be tightly controlled at the resolution of the scan, and also constrained by an aperture. Very soon, we are at the mechanical complexity of a drum scanner, which works similarly.

We need to be sure that our diffusion sources are truly limiting before it would seem worth trying something else, and then we should try to concoct a proper collimated source. Those are much cheaper and easier alternatives, it seems to me, and cheap and easy are objectives given that people will probably have to build their own.

Rick "back in travel mode and having to delay further progress for a while" Denney

[Peter J. De Smidt]

We need to be sure that our diffusion sources are truly limiting before it would seem worth trying something else, and then we should try to concoct a proper collimated source. Denney

I agree.

[ludvig friberg]

Yes all that make a lot of sense. If one used a a computerscreen as backlight as I have seen some interest for one could at least easily try it by drawing a white line one pixel row at a time. It could be interesting to try I am just throwing some ideas around

[Peter J. De Smidt]

Throwing ideas out there is a very good thing! Even if it's something the people don't check out immediately, it could turn out to be very useful down-the-road.

[ludvig friberg]

I have built a lightbox for flash. It works ok and is almost even but I am having trouble getting anything larger than 40x40mm even with my design. Ideally I would want something bigger so I can use the same backlight for full frame captures of a 6x7 negative or slide. Then If I need to i just move the camera closer and take a series of tiles. I started to look at dichroic light boxes. Could thhese be altered to have a flash as a lightsource?

What would be the best way to get a flash like speedlite really even across 6x7 is the question I guess.

[Peter J. De Smidt]

I based my source on a De Vere light mixing box. The lighted area is approximately 250mm x 250mm.

[Nathan Potter]

I have built a lightbox for flash. It works ok and is almost even but I am having trouble getting anything larger than 40x40mm even with my design. Ideally I would want something bigger so I can use the same backlight for full frame captures of a 6x7 negative or slide. Then If I need to i just move the camera closer and take a series of tiles. I started to look at dichroic light boxes. Could thhese be altered to have a flash as a lightsource?

What would be the best way to get a flash like speedlite really even across 6x7 is the question I guess.

ludvig, I think you would need what is commonly called a beam expander. It would be a simple condenser lens, or set, at some distance from the flash face - say a few hundred mm - and a diameter sufficient to easily cover the largest size image frame you are likely to use. The condenser set from a 4X5 D2 enlarger comes to mind. That condenser is a pair of convex/plano 150 mm diameter lenses face to face. This set might be placed 100 to 200 mm from the face of the flash; you'll need to experiment a bit with this. Also you will have to work on the uniformity of the light hitting the film. The further the condenser is from the flash the more even the light but the lower the intensity.

A second approach without using condensers is to just increase the distance between the flash and the film. I do this by bouncing the flash off a white card which sits at a 45 degree angle right under the film. The flash is at the end of a cardboard tube about 1 meter away from the white card. The white card is a good diffusion source so reduces the dust and particle imaging compared to the condenser technique.

Nate Potter, Austin TX.

[ludvig friberg]

I googled quite a bit and could not find a good description on how a de vere box is constructed. It seems to involve two lightsources and a mirror plus some diffusor? Do you have a picture I could look at?
I am looking into a good way to do the flash at a distance but the best would be to have it a bit more compact than 1 meter away!

Thanks for the suggestions.

[Peter J. De Smidt]

Ludvig,

I'll take some pictures soon, both of a De Vere unit and of my light source.

[Serge S]

Why can't we just adapt a standard lightbox previuosly used for viewing slides as a scanning platform mated together with a simple copy stand. They are color calibrated with a decent CRI number.

Previously, I had purchased a couple of Bowen Ilumitran side dupers used on ebay with the idea of using it to scan (with a DSLR), but never got around to setting it up.

[Peter J. De Smidt]

Why can't we just adapt a standard lightbox previuosly used for viewing slides as a scanning platform mated together with a simple copy stand. They are color calibrated with a decent CRI number.

Previously, I had purchased a couple of Bowen Ilumitran side dupers used on ebay with the idea of using it to scan (with a DSLR), but never got around to setting it up.

Standard light boxes aren't ideal for two reasons. First, most have very noticeable hot spots. My Porta-Trace one certainly does. Second, at these magnifications vibration can be a real problem. If you have a Canon with the EFSC shutter mode (See: http://krebsmicro.com/Canon_EFSC/index.html), then you can probably do what you like, but if you don't have that super-low vibration shutter, then you probably want to avoid timed exposures in the 1 second to 1/15th of a second range. Using flash correctly minimizes the problems with vibration, but there might be issues with consistency, although that hasn't been a problem for me so far.

[Serge S]

Standard light boxes aren't ideal for two reasons. First, most have very noticeable hot spots. My Porta-Trace one certainly does. Second, at these magnifications vibration can be a real problem. If you have a Canon with the EFSC shutter mode (See: http://krebsmicro.com/Canon_EFSC/index.html), then you can probably do what you like, but if you don't have that super-low vibration shutter, then you probably want to avoid timed exposures in the 1 second to 1/15th of a second range. Using flash correctly minimizes the problems with vibration, but there might be issues with consistency, although that hasn't been a problem for me so far.

Thanks Peter. I am planning to upgrading to the 5D2 at some point. I see that it has that shutter you are referencing:)
Sounds like the Illumitran Idea was not half bad, as it is electronic flash

[rdenney]

I just managed to score a nice Illumitran, with the contrast device. I'm thinking a mirror might be a useful thing with that and my setup. I'll know more when it arrives.

Rick "who recently put eyes on some old work with 35mm Kodachrome that will see action again" Denney

[ludvig friberg]

The uneveness is not a big issue at least not for me. I have done some testing where I take a picture of the lightsource and use to divide the uneveness from the lightsource. This also takes out any vignetting and dust on sensor. If I have a piece of the film with no exposure I can also divide out all colorshifts from the lightsource in combination with the orange mask. I haven't tested how important CRI is yet. I have one some exposures with a Canon speedlite flash and some LED sources but I don't know the CRI rating of any of them. Xenon bulbs is far from perfect I believe and LED even worse most of the time. I get pretty decent results anyway. I am going to try out sunlight when it stops raining and see if there is any big difference. Sunlight of my test neg should provide a nice perfect reference. I have looked at some new halogen bulbs called solux that have a very good CRI and are used by kodak in the Pakon line of scanners.

[Peter J. De Smidt]

On an email list, Ernst Dinkla, a very knowledgeable scanner, printer and tinkerer, said that with BW film his Coolscan does best using only the blue LED. (Using Vuescan he can turn off the red and green leds.) He theorized that due to the shorter wavelength the scanner could capture more detail than when using all of the leds. He posited that with a dslr scanner using green light for BW film might be best, as most dslrs have twice the number of green sensors than blue or red, and it would avoid the longest wavelength, i.e. red. Since I wanted our scanner to work in color, I've been working with a white source. I also thought that getting info to all of the camera's sensors might be a good idea, but Ernst's idea is definitely worth looking into for BW scanning.

[ludvig friberg]

I have done quite a lot of testing on color neg lately. I have tested different LEDs, flash, sunlight, tungsten in my dslr scanner up against a hand tuned scan in Nikon coolscan 8000ed. With handtuned I mean I have dialed in the orange mask by changing the exposure of the red, green and blue LEDs in the scanner to get the best possible AD conversion from the sensor. This is my gold standard right now. I get pretty good separation between the cyan, magenta an yellow inverse dyes from the Nikon butI have never scanned on anything better than the NIkon so I dont know how good separation I should try to achive. I have no idea what wavelenghts the leds in in the 8000ed has but I assume it is as wide as possible to cover the very varying dyes in different films.

The conclusion of my tests so far is that tungsten is by far the best artificial lightsource for negative color film. This is quite obvious if you think about it, tungsten is full spectrum so it covers all parts of the dyes. With the LED and flash I get much poorer separation between C,M and Y. It is of course much to warm(Kelvin) so I use full CTB filter, this takes a lot of light so I need a lot of Watt. Either that or one exposure for the red channel and another one for the green and blue. Probably I will need something like 1500 lumen from the lightsource. I will have a relay so it turns the light on only for a second or so when its exposing. Also any fans will turn off a second before exposure. And then dimmed for setting focus and general setup, the live view autoexposes and shows a nice image with some more iso and full aperture. I will try in the future to do the Infrared too and then tungsten is very handy, lots of IR is emitted.

Another solution I have looked at is seperate leds in a matrix and a diffusor. Either rgbleds plus IR or led in a pattern of: red, blue, green, IR, red, blue etc. Here you have to really choose your wavelenghts with precision I think most leds are quite narrow.

This one is a cheap and very customizable platform for something like this.

http://evilmadscience.com/productsmenu/tinykitlist/75

or this one for a smaller lightsource like mine
http://evilmadscience.com/productsmenu/tinykitlist/157

I have built the first stage of motion now and I have solved all the programming and control of the machine. Are working on the negative holder now, waiting for parts to finish the stage.


best regards

Ludvig

[Peter J. De Smidt]

It's great to hear about your progress, Ludvig. Nice work! Are you using an Arduino to control the motion? (Although we should probably switch threads if we discuss that aspect of the project.)

[Daniel Moore]

Very exciting developments, Ludvig. Please tell us more about your design so far.

[Peter J. De Smidt]

Heat reflecting or heat absorbing glass could be helpful with a tungsten source, ala many enlarger heads. If someone goes this route, using a real enlarger head could be a very viable alternative, especially if they have one laying around.

[cosmicexplosion]

i have absolutely nothing to contribute to this fred, except maybe a bit of cheering from the side, GO YOU MAD BASTARDS!!!!!

[Peter J. De Smidt]

:)




.

[David_5527]

Hi. Has anyone tried using a large computer screen? I have an Apple 27"that gives off a lot of even light. I am trying to figure out a way to do cheap scans of my 14x17" negs before I make expensive scans or platinum print.

It seems like the issue will be to get it a bit away from the screen, so you dont see the screen pixels...anyone trying this?

[David_5527]

Actually, my computer screen is just shy of 14", so wont work...oh well....other sources?

[Amedeus]

Ludvig,

Getting white out of RGB LEDs is a difficult and expensive proposition if high CRI is a requirement. Getting white with high CRI out of a yellow phosphor driven by a blue or UV LED is equally difficult. You can match the LED wavelengths with the RGB sensor and the rest becomes a matter of software.

Xenon and tungsten are both blackbody emittors/radiators like the sun. When properly driven, Xenon is closest to sunlight and is preferable when a lot of blue is required. Xenon flashes are driven by very short high current pulses and as such the color temperature is often too high (lack of red, overdose of blue) and the latter doesn't necessarily match with a digital sensor.

Evenness is easy to deal with through holographic diffusers if the light source is not tooo hot ;)

I'm the CTO for a lighting company dealing with LED and Xenon with quite a few applications crossing over in photography lighting ... I'm open to specific questions.


The uneveness is not a big issue at least not for me. I have done some testing where I take a picture of the lightsource and use to divide the uneveness from the lightsource. This also takes out any vignetting and dust on sensor. If I have a piece of the film with no exposure I can also divide out all colorshifts from the lightsource in combination with the orange mask. I haven't tested how important CRI is yet. I have one some exposures with a Canon speedlite flash and some LED sources but I don't know the CRI rating of any of them. Xenon bulbs is far from perfect I believe and LED even worse most of the time. I get pretty decent results anyway. I am going to try out sunlight when it stops raining and see if there is any big difference. Sunlight of my test neg should provide a nice perfect reference. I have looked at some new halogen bulbs called solux that have a very good CRI and are used by kodak in the Pakon line of scanners.

[Daniel Moore]

David, admittedly, this is off the top of my head, but I'd think about stitching flatbed scans for this size. It could be possible to find a sheet of anti newton ring glass to prop up the neg to about the right height off the table, and perhaps another on top to sandwich it and keep it flat.

[Daniel Moore]

Amedeus, many of us have spent a good deal of R&D dollars (given that Time=Money : ) looking for a high CRI/high EI solution. You are in a unique position to comment on this. What would you chose personally?

[Amedeus]

Let me ask some boundary condition questions first:

Area to light ?
Sensor technology ?
Line sensor or area sensor ?


Amedeus, many of us have spent a good deal of R&D dollars (given that Time=Money : ) looking for a high CRI/high EI solution. You are in a unique position to comment on this. What would you chose personally?

[ludvig friberg]

To David:

I have built a setup to make digital "contact sheets" of all my archive of negatives. These are mostly 120 film in A4 plastic sheets. I used a 24inch screen for this with a diffusion filter, the type used on lights usually bought on wide roll. I had a small part of the screen not covered in diffusion. In this square I had a number displayed, this way I could easily index all my old negatives. It was 10+ binders full of negs, thousands of exposures. I managed to get digital contacts of them all in 2 days. Took a few seconds per sheet. I just have the raw files and a negative lightroom setting, works fine for preview and was MUCH faster than any scanning method.

Ludvig

[ludvig friberg]

CRI not important? Finding and matching sensors sensitivity to r, g and b respectively better?

I have done a lot of test on negative color film and my best results so far comes from RGB LED. I can easily dial in whitebalance for the negative and further getting the best separation between the CMY layers. It seems to me that the sensor has a much broader sensitivity around each of the r, g and b. Leds seem to be often VERY narrow in their color.
But still, the camera only sees the world as red, green and blue. And to me this seems to be the best way.

Also please tell me more about holographic diffusers!

[Daniel Moore]

For several reasons I think illumination to cover 4"x5" is ideal, it's the smaller formats that will benefit most from this project. The capture device will be a DSLR so in all likelihood a CMOS sensor. Since there is no single widely accepted design as yet, both stationary and moving light (integrated light and negative carrier) sources are viable. Favor given to higher shutter speeds if possible.

[Peter J. De Smidt]

My light source currently has a light emitting area of 4.25" square. I wouldn't want to go smaller than that.

[Amedeus]

OK, read a few posts deeper in this project and I'm getting it ... "scanning" negatives with DSLR rather than with a scanner. So you gain processing speed and there are quite a few DSLR's out there that will beat scanners in IQ and definitely in processing speed ;-)

RGB LEDs are indeed narrowband light sources, average ~ 30nm and the CMOS or CCD sensors of DSLR or digital backs to have a significant wider spectral performance and that is how the internal software of the camera is adjusted for. This said, it is possible to get good to decent performance with RGB LED and CMOS sensors if one can adjust the curves a bit ;-) ... this is what happens in cameras or post processing software anyhow when one corrects for different CT's and even different spectral performance for mercury based lights (fluorescent, mercury-vapor or HMI)

It's even possible with RGB lasers to create a white light suitable for full color projection and the bandwidth is only 2-3nm per wavelength. Not claiming the CRI is perfect but the eye can be fooled.

While LED's have equally a challenging CRI proposition but some manufacturers get creative and mix warm white (~3200K) and cool white (~5000K) together to create a more pleasing blend by dynamically adjusting the current for each sector. In some cases cyan/green/red LEDs are added to ensure sufficient CRI performance. Keep in mind that the average white LED has peaks in blue and yellow and green/red performance is not too stellar.

Brings us to the next challenge ... blending it all together. Space and diffusers are your best friend. You will be loosing intensity but gain intensity and color uniformity. There is no magic bullet that offers it all, just the plain nature of physics. Some LCD backpanels are doing a decent job on first sight but as soon as you remove the LCD you'll find out that uniformity kinda sucks. There is just not enough space in an LCD sandwich to create great diffusion without throwing most of the light away.

Fast shutter speed ? This translates to lots of light and that in turn requires more power feeding the LED's and at that point you have to keep an eye on the thermal performance of the LED. Although the latter don't produce any IR out of the front, they produce a decent amount of IR out of the back and that needs to be dissipated one or other way. If not, the LEDs will become less efficient (droop) and you'll have to deal with a color shift ... more so for while LEDs.

It's of course all possible but will require choosing the correct operating point/compromise for the application. Brightness, uniformity, size, cost ...

I'm personally somewhat familiar with the immediate challenges as I've built on multiple occasions diffuse light sources to illuminate CCD and linescan sensors during their calibration cycle, so "perfect" uniformity was high on the list of desirable characteristics. In all recent cases a combination of LEDs were used with a plurality of spaced holographic diffusers in a large integrating rod type of assembly (I need to find a link to a company who will supply holographic diffusers to individuals, they are widely used in LCD displays and in LED fixtures ... ) In the days of no suitable LED's, we would use tungsten lamps to achieve same and have the benefit or blackbody CRI performance. Instead of holographic diffusers we used anything from ground glass, milk glass etc ... you name it, we probably tried it. You can use Xenon but it is too much of a point source to diffuse easily.

Let me know if I'm way of topic here or misunderstood what you are trying to achieve. So 4x5 makes sense for those wanting to scan 4x5 negatives but the majority will be using 6x6, 6x7, 6x8 and 6x9 format ... 35mm can use same also ... of course ;-)

[Amedeus]

Also please tell me more about holographic diffusers!

Take a look at http://www.luminitco.com/

[Amedeus]

CRI not important? Finding and matching sensors sensitivity to r, g and b respectively better?



Since RGB LEDs are so narrowband you can generate a tri-stimulus with no overlap when it comes to the sensor part (depends on the Bayer filter on the sensor of course, they are not all created equally)

Without "overlap" one can intuitively see that one can adjust the power to the individual light sources to hit the correct coordinates into RGB space to achieve a wide range of CCT (correlated color temperature ... as we're not taking black body spectral performance here) This is how the laser projections systems work also. Strategically chosen wavelengths to fool the eye we're seeing white. CRI is poor but the eye doesn't seem to mind. Although in the latter cases saturation can be an issue.

If a narrowband light sources stimulates two colors in the CCD/CMOS sensor at the same time then it is much harder to create a clean profile as changing one, affects the other ...

[Struan Gray]

I usually recommend Edmund Optics as a source of optical components for individuals buying lots of one. Not only are they willing to deal with individual amateurs, but they do so in a professional and friendly way. It's always worth checking their surplus outlet, Anchor Optics, to see if what you want is available more cheaply. For photographic use, the difference between their research grade and commercial grade optics is (almost always) irrelevant.

They have holographic diffusers.

Worth noting though, is that if you are going to scan-and-stitch (rather than shoot the whole neg as a single frame) there are advantages to restricting the illuminated part of the negative to only the part being photographed by the DSLR. Flare will be less, and all your illumination optics become smaller, and so can be cheaper and/or higher quality.

There is a fundamental tradeoff in scanning between diffuse lighting for convenience and minimising the intrusiveness of scratches and dust, and collimated lighting (Kohler illumination, if you can manage it) for maximum sharpness at the resolution limit. These were issues debated ad-infinitum in the days of enlargers, and they are still relevant now. There is no best answer - it depends on your way of working, and your goals.

Finally, the traditional argument against using narrowband light sources is that they make it hard to distinguish subtle colour variations - which for me at least is one of the main attractions of LF). It's the same problem as print metamerism. However, here it becomes a question of calibration, but it might mean that you need to build your LED panel with hardware adjustments of the colour balance, rather than relying only on post-capture software adjustments.

[Peter J. De Smidt]

<snip.
Worth noting though, is that if you are going to scan-and-stitch (rather than shoot the whole neg as a single frame) there are advantages to restricting the illuminated part of the negative to only the part being photographed by the DSLR. Flare will be less, and all your illumination optics become smaller, and so can be cheaper and/or higher quality. <snip>


I agree, as long as the masking material is between the negative and the lens. The light source, at least with a diffuse source, needs to be much bigger than the area being sample, since otherwise the edges of the negative will get darker than the middle of the frame.

Flash is a viable lighting option. It's been working well for me.

Edmund optics is very good. So is Thorlabs.

[Amedeus]

Flash is a viable lighting option. It's been working well for me.


Xenon flash is definitely a viable option. Takes a little more work/optics to generate a flat illumination field (or overfill the aperture ...) but the key advantage is that the CRI is high and CCT is well defined (mostly blackbody radiation) and software is readily available for correction. Simply built in into the camera or post production. And easy to get short shutter times ;-)

[Amedeus]

There is a fundamental tradeoff in scanning between diffuse lighting for convenience and minimising the intrusiveness of scratches and dust, and collimated lighting (Kohler illumination, if you can manage it) for maximum sharpness at the resolution limit. These were issues debated ad-infinitum in the days of enlargers, and they are still relevant now. There is no best answer - it depends on your way of working, and your goals.

There's a lot going on for collimated lighting of course, the downside is a more difficult set-up and often a higher cost in optics. Definitely a winner in combination with a point source when it comes to obtaining a flat field




Finally, the traditional argument against using narrowband light sources is that they make it hard to distinguish subtle colour variations - which for me at least is one of the main attractions of LF). It's the same problem as print metamerism. However, here it becomes a question of calibration, but it might mean that you need to build your LED panel with hardware adjustments of the colour balance, rather than relying only on post-capture software adjustments.

Agreed with the need for hardware adjustments. Also, there's no need to be limited to 3 LED colors ... Many designers of high CRI LED fixtures utilize 5-6 or even more colors to get the spectral response they need, each with their own drivers and abiltiy to adjust operating current.

[Daniel Moore]

I'm not sure how various colors are created in LED technology (is it really just power?) but how about adding alternating gel filters to an LED array panel such as this Back-lit LED Panel shown further down this page. (http://www.luminousfilm.com/led.htm)?

Is there a way in which a digital capture can be used to analyze an illuminant for spectral information beyond the basic RGB histogram?
Would an image taken of a holographic diffraction grating film (http://www.anchoroptics.com/catalog/product.cfm?id=247) placed over the illuminant be instructive?

Does anyone have a notion on what it would take to wire an array of lower power flash bulbs? Specifically, how to get say 2 or 4 bulbs to agree they are charged similarly? The plan would be to arrive at an averaged output to help rule out exposure variations, while allowing the illumination to be built into the negative stage thus moving with it.

[Struan Gray]

.... there's no need to be limited to 3 LED colors ... Many designers of high CRI LED fixtures utilize 5-6 or even more colors to get the spectral response they need, each with their own drivers and abiltiy to adjust operating current.

That's an interesting idea. It's not as if this project needs another sub-project though :-)

Would an integrating sphere (mixing box for old timer darkroom colour types) make more sense than layers of diffusion if you have that many different colours to even out? It would make it easy to switch between flash and LED lighting if you wanted that.

[Struan Gray]

I'm not sure how various colors are created in LED technology (is it really just power?) but how about adding alternating gel filters to an LED array panel such as this Back-lit LED Panel shown further down this page. (http://www.luminousfilm.com/led.htm)?

LEDs are pretty-well monochromatic, certainly by photography's standards of spectral resolution. What that means is that the spectrum is just a narrow peak centred on a particular wavelength. There are ways of nudging the peak to higher or lower values, but you can't move very far - not far enough to change the apparent colour as seen by the eye - and most commercial LEDs are mounted and packaged so as to prevent this happening as much as possible.

So adding a filter won't shape the spectrum in the way that, say, using a 85B on incandescent light will boost the blue over the reds and yellows. All any filter will do is cut the peak to a greater or lesser extent (including perhaps extinguishing it altogether).

So for colour shifting you tend to combine three or more LEDs with different peaks corresponding to some set of primary colours. The eye-brain system then does the work of turning the three monochromatic peaks into a perceived colour. Varying the intensity of the individual LEDs (usually by adjusting the average current in various ways) allows you to mix the three primaries and create colours in the usual additive colour wheel way.

There are now 'white' LEDs which have a high enough power to then put a filter in front of for spectral tuning. They are made by combining LEDs of different colours exactly as described above (and sometimes a phosphor or two), but in such a way as to create a colour perceived as white. In some ways, the manufacturer has done the job for you, but in a way that cannot be adjusted or altered.



Is there a way in which a digital capture can be used to analyze an illuminant for spectral information beyond the basic RGB histogram?
Would an image taken of a holographic diffraction grating film (http://www.anchoroptics.com/catalog/product.cfm?id=247) placed over the illuminant be instructive?

If you know what you are looking for, just looking at the spray of coloured light from a CD or DVD is instructive. Anything which diffracts relatively evenly across the visible spectrum will do, including those cheesy effects filters which surround streetlights with rainbow stars.

If you want to get into quantitative measurements cheaply there are ways to build cheap spectrometers (a cereal packet and a CD are popular school science projects) for direct viewing with the eye, whose spectrum you could photograph with your camera. Otherwise there are cheap (one-to-several hundred dollars) USB spectrometers which will do the job, or if you have a colour calibration puck for your screen and/or print output some of them have shareware programs available to give you a raw spectrum from their built-in spectrometer.

Often though, interpreting the spectrum is more difficult than obtaining it. This is doubly so if you wish to interpret it in terms of perceived colour rather than scientific spectrum weightings.



Does anyone have a notion on what it would take to wire an array of lower power flash bulbs? Specifically, how to get say 2 or 4 bulbs to agree they are charged similarly? The plan would be to arrive at an averaged output to help rule out exposure variations, while allowing the illumination to be built into the negative stage thus moving with it.

My own tinkering has lead me to the conclusion that it will be easier to build a small area illuminator and move the film through the patch of light it produces, than to create a wide-area light source which is even in intensity and colour. However, there are no rules, and nobody is handing out prizes :-)

[Daniel Moore]

Thanks Struan, great info.

The concept I'm working on is based on a stacked XY sliding mechanism so it would have to carry the light with it, as opposed to an open base/sliding rail arrangement with the light in the center. There is a deluge of LED lights on the market now, some claiming high CRI values and all intended for photography. I wouldn't mind testing some of them if I had a scientifically sound way to do it.

"shareware programs available to give you a raw spectrum from their built-in spectrometer." Do you have one in mind? Still seaching on that one. Being able to determine CRI would be handy.

[SURF]

Spectroscope is the key world. There is a lot and cheap.
http://www.ebay.com/sch/i.html?_sacat=0&_from=R40&_nkw=spectroscope&rt=nc

Want to add some obvious and not so things:
1. For BW originals color doesn't matter.
2. For negative and positive color originals the color of the light source should be different.
3. It is not obvious that the lamp with the best CRI will be the best for digital camera scans. For me at least.
4. The LCD monitor can be 3-4 stops brighter if the TFT panel is removed. But it is not straightforward task. If the color is OK then it makes a perfect light table.
5. LEDs for LCD panels often are not narrow peaks seen through spectroscope. Fluorescents for LCD are mostly narrow peaks.
6. One of the best high CRI lamps are made of incandescents with blue filter.

[Peter J. De Smidt]

I'm not so sure that color doesn't matter for BW film "scans". It would be interesting to try different colored sources and compare results. My guess is that green light would be best, but this is the type of thing is best investigated by experiment. I could use Rosco filters (from a swatch book) on my flash as a source.

Another idea would be based on a Philips PCS2000 tri-color enlarger. This one had three halogen bulbs, each with it's own filter; a red, a green and a blue. The intensity of each light could be varied independently of the others. It was a very good setup for printing color, and not bad for VC papers. The only thing I don't like about halogen bulbs is the heat.

Various LED mixes could also be interesting, but the biggest hurdle right now for me is automation.

[Amedeus]

That's an interesting idea. It's not as if this project needs another sub-project though :-)

Would an integrating sphere (mixing box for old timer darkroom colour types) make more sense than layers of diffusion if you have that many different colours to even out? It would make it easy to switch between flash and LED lighting if you wanted that.

I've used large integrating rods (rectangular and cylindrical) and diffusion and even full "real" lab integrating spheres and diffusion for the camera calibration projects. Needed diffusion in all cases for this particalur application where the camera is seeing the lightsource directly. We were looking at 2-3% intensity flatness ... pretty much approximating the noise of the sensors we had/have access to.

Most of the high CRI lightsources I'm involved with use more than 3 wavelengths or phosphors to get there ... especially when you need a decent R9 value.

Regardless of the solution, you can still switch out Xenon versus LED, especially with a mixing box approach

[Struan Gray]

"shareware programs available to give you a raw spectrum from their built-in spectrometer." Do you have one in mind? Still seaching on that one. Being able to determine CRI would be handy.

There is an open source colour managment system called Argyll which will read an illuminant spectrum if the colorimeter/spectrophotometer allows it - which means it needs to be one of the higher end models. See here:

http://www.argyllcms.com/doc/illumread.html

(project home page here: http://www.argyllcms.com/)


If you don't already own one of the suitable colour calibration devices, you can homebrew a crude spectrometer fairly easily, but you can save time and effort by piggybacking off a project like this one:

http://publiclaboratory.org/tool/spectrometer


Naturally, there are more advanced options too, but for measuring a lamp's CRI or the CIE coordinates of particular LED combinations you can get by with a relatively crude instrument.

[Struan Gray]

... snip...

Thanks Rudi. Nice to know how the pros do it :-)

[AGChicago]

(Disclaimer: new forum member, first time post.) I've been developing a DSLR digital capture rig for 35mm film for 4 months or so. Ok, so a bit smaller in scale but the principles are the same and I'd like to do a 120 unit also (so XY, stitching, etc. are also an issue). I'd like to share a few things I've learned thus far, though I can't say I've found the magic bullet yet regarding lighting for all film types. I apologize for the length. I'd also like to say that this endeavor is awesome, and I believe we're only a year away from a film scanning revolution.

-As SURF states just above, the light source for positive and negative (color at least) shouldn't be the same, ideally. I started with positives, and after much researching and trials, settled on a Xenon lamp (not strobe). Xenon lamps run cooler then Halogen, have a CRI of 100, and a color temp of 3100K. I have tried multiple sources, but none produce the color rendition and saturated image of the Xenon lamp. Even close. Why? Plainly put, to me anyway, it is because a positive film exposure is already perfect. It should be lit in a neutral temp, with as high a CRI as you can muster, or you will shift the colors from what your eye sees and the film has in it. Digital positive captures using a high temp LED diffused source created a washed out image with no shadow detail. You can WB it out, and adjust color curves and saturation, but all that manipulation creates noise in the image. Shine a 5100K LED lamp at a flower, take a picture, then notice how crappy it looks. Same idea. I think the Solux Halogens will work also, as someone earlier stated, a 4700K bulb is used in some high end industrial film scanners. But I think the 4700K is a compromise between positive and negative ideal sources, and they run hot.

The setup and cal was simply to white balance on the diffusion plate with no film, capture a frame, and done. I've never had to touch the color curves or contrast. I can post a sample untouched jpg if there is interest.

-The aforementioned 3100K/100CRI Xenon bulb, when used on a color negative (I've tried various Kodaks, but no Fuji's), looks, to put it mildly, like dog doo. The orange mask is lit into a gloriously bright and highly saturated glow. When reversed, the blue channel is screaming out of control and requires significant curve adjustment. After you're done, the noise levels are through the roof and the resulting image is garbage. The first light source I bought to try was a cheap $30 76 LED "video light". With some diffusion the results with negative color are far superior. My theory so far is that it works better because the higher color temp is diminishing the saturation of the orange mask level in the capture, hence subduing the blue channel after reversal. I think similar results might be achieved using a blue 80A or 80B filter (or gel) in front of the Xenon lamp. Albeit at the expense of lumens at the frame. The idea of using some type of blue filter on the light source (or very high temp source) seems a common one when scanning or digitally capturing negative color film. The RAW file captures (pre-reversal) between the Xenon and LED are really night and day, they look completely different. I think the light source, so far, is the most challenging issue.

I plan on trying various blue filters on the Xenon tonight, just to experiment and see what I get. And negative B&W, I haven't even started with yet. I suspect that the mask colors (if the film has one) will cause issues.

[Daniel Moore]

AG, care to share some of your design details for the light source? And I'm very curious to see more of how you approached the negative stage as well, which there is a thread for here (http://www.largeformatphotography.info/forum/showthread.php?87539-DSLR-Scanner-Negative-Stages). Thanks for sharing your valuable empirical evidence.

[Peter J. De Smidt]

First off, Welcome to the project AG!

We've got a lot of good people working on this, and I expect that progress will increase once the snow flies.

What Xenon bulb have you been using? It's likely, as you say, that different media types will do best with unique lighting. I plan on playing with a rosco gel swatch book on my flash source.

[AGChicago]

Sure, I'll elaborate. The attached shows the prototype (sorry for handheld low rez shots). The light is a 20W undercabinet "puck" light, mounted inside of 3" thinwall PVC tube (with standard end caps). The diffuser is 1/8" opaque white cast acrylic, and the light is mounted right to it (with a 2" hole cut in the mounted end cap and front plate). You can get 35W Xenon bulbs of the same format, I'm going to try that also. In the end, a standard MR16 would be better, and the PVC will be replaced by stainless steel (with some holes to allow some air flow). When I try different lights, I just pull off the light assembly and direct the light into the mounted end cap. The opaque diffuser works fine, but more might help. Using opaque eliminates the hot spots from the light source, at the expense of light output.

A small trick I've used is to mount a 4" diameter tube to the front (PVC, but again will be stainless on the real one). This keeps any stray light from getting to the DSLR lens. The inside of the tube, and back wall in front of the film is covered in telescope flocked material. Protostar.biz sells it very cheaply, and it's pretty nice quality. So basically you're shooting into a non-reflective black hole.

Because this is for 35mm, there is the luxury in getting off the shelf film holders. I've simply created a fixed opening with UHMW polyethylene slide rails inside (it's nice and slippery and very wear resistant). I don't know if any of that is applicable to large format capture designs. Unless you fix the camera, and design a film holder that can move and index up/down between the light source and "shooting tube" front plate. Side to side is not an issue. The forum seems to be focusing on keeping the film fixed and moving the camera though.

[Peter J. De Smidt]

Thanks for the updates.

All of the designs I remember here keep the camera fixed and move the negative.

[Amedeus]

Hi AG,

Thanks for sharing, quite a download of information here.

One tidbit of information caught my eye ... "Xenon lamps run cooler then Halogen, have a CRI of 100, and a color temp of 3100K."

I would like to know which Xenon lamp you are referring to with a CRI of 100 and a CCT of 3100K

I also tend to agree with your observation that the light source is indeed the most challenging part as any compromise there has ramifications for the post processing.

[Struan Gray]

A quick heads up on safety: bare Xenon lamps can put out a lot of UV, enough to give you snowblindness if you are unwary during alignment or use of the capture system.

I have only worked with research grade xenon lamps (up to 1 kW - great brownouts as you start those up :-). For measurements in the visual spectrum we always filtered out the intense IR with a cold mirror, and the UV with a filter. I assume that bulbs intended for domestic use have filters integrated into the glass, but it may be that the enclosure is intended to do the filtering, in which case homebrewers will need to take UV eye saftey into consideration. The old Ilford colour gel packs for enlarging ilfochrome go for pennies on the dollar and include a useful UV filter (and a bunch of colour correction gels to boot).

AG, nice to see your setup. The rationalist in me says that you should be able to capture any piece of film with three monochromatic sources of the right wavelength. You are, after all, trying to measure the density of only three different dyes, and suitably chosen wavelengths should allow you to tease out their individual densities without cross-contamination. But, and it's a big but, both the dyes in film and the colour filters in digital cameras are explicitly designed to be used with continuous light sources, and fighting that fact leads you - as you found - into signal to noise problems.

Once upon a time there was talk of developing a film whose actual colours after development bore no relationship to the colour bands they were recording. The idea was that the freedom to manipulate the final colour would give you more freedom in chosing sensitiser dyes. Such a film might be more useful in the long term of analogue photographic capture, but I suspect the budget to develop it has long since gone.

FWIW, one advantage of an LED-based setup would be the ease with which you could take separate colour frames for RBG (or CMY), each optimised for the signal to noise of that particular channel. It would be a way of getting round the unavoidable fact that colour negs have little blue in them.

[Daniel Moore]

"FWIW, one advantage of an LED-based setup would be the ease with which you could take separate colour frames for RBG (or CMY), each optimised for the signal to noise of that particular channel. It would be a way of getting round the unavoidable fact that colour negs have little blue in them."

Struan, this begs to be expounded upon! And thanks as always for sharing your expertise.

[Struan Gray]

Daniel, the problem, as AG pointed out, is that it is very easy to get an underexposed blue channel if you do one-shot 'scans' of colour negative film with a DSLR. When you adjust the colour balance to restore the right proportions between the channels, the blue channel ends up with significantly less signal to noise than the red and green.

The problem is that colour negatives simply are not very blue if you hold them up to the light. Your eyes, and DSLR colour filter arrays, are made to sense all colours at once, and if you adjust the total gain so that you don't blow out the strongest channel, the weak one ends up underexposed.

Papers made for analogue printing from colour negatives could be adjusted so that the blue-forming layer had the right combination of sensitivity and gamma that the strong orange colour of the negative did not matter. You could make a DSLR with a higher gain for the blue pixels and achieve the same end, but nobody does that because it would be useless for taking pictures of the real world. You could make a scanner with a blue-sensitive sensor with a higher gain, but there is then the engineering problem that the sensor itself will have some optimal gain for best signal to noise, and it makes much, much more sense to keep the on-chip gains the same and have a brighter light for the blue channel.

In principle you can do this with blue-filtered incandescent or xenon light, but the spectral absorbances of the dyes in negative film are quite broad, so it is much easier to do with a narrow band LEDs whose wavelengths are chosen to minimise absorbances by the dyes they are not tuned to. The problem is usually the magenta dye, since there is no such thing as pure spectral magenta - it is a colour invented by our trichromatic vision.

Kodak and Fuji do not, as far as I know, publish the spectral absorbances of the dyes in developed negative film. They do publish the spectral sensitivities of the unexposed film, and the resulting absorbances you get from a grey exposure, which both act as reasonable surrogates to find starting points.

[Struan Gray]

Sorry, typing a bit quick to make a deadline, and missed the conclusion:

The beauty of LEDs is that you can choose three wavelengths which as much as possible avoid cross-contamination by avoiding the portions of the dye absorbance spectra which overlap. You then adjust the brightness of each type of LED (with modulation, or by adding more or less LEDs to the source) so that each channel gets as much light as it can take without being over-exposed. Post-capture you adjust the strength of the relevant channels *downwards* to balance for grey and overall gamma.

If the spectral absorbances of the dyes in the film overlap at all wavelengths, there will always be *some* cross contamination. But with better signal to noise in the blue channel you are in a better position to do something about that with further processing.

The central point is that just because colour negative film was optimised for analogue printing with an incandescent light source, it doesn't mean that will be the best light source for capture with a DSLR.

[AGChicago]

The bulbs: http://www.lightbulbsdirect.com/page/001/PROD/XenonBiPin/LVPX20BP

The CRI (go all the way to the bottom): http://www.pegasuslighting.com/xenon-light-bulbs.html

Struan- Xenon arc lamps have, from what I read, huge UV. Xenon low pressure household bulbs are incandescent. http://en.wikipedia.org/wiki/Xenon_arc_lamp .

Optimizing the three film channels individually against an LED source would be great. Can't quite wrap my head around how you would go about tuning them to a particular film.

My mistake on the XY being for the camera. I read through the posts, but after a few pages, it starts to be confusing what the question was!

[Peter J. De Smidt]

For color negatives, couldn't you use something like a Cree 3w red, green and blue leds, each color being adjustable, and shining all of them into a light mixing box. Now place a piece of orange film leader from the film you want to scan on the light source, and adjust the leds so that the readings through the film base for each channel are equal. For instance, in 8 bit RGB, make the reading through the film base equal to 250, 250, 250. Now the relative intensities of the three colors should remain constant for that film, although the overall luminosity might have to be adjusted some.

[Peter J. De Smidt]

Here's some pictures of my current light source, LS2:

http://i955.photobucket.com/albums/ae37/peterdesmidt/LS2_LED.jpg

http://i955.photobucket.com/albums/ae37/peterdesmidt/LS2_Flash.jpg

http://i955.photobucket.com/albums/ae37/peterdesmidt/LS2_Light_Souce_Entry.jpg

There are two sheets of diffusing plastic. The top sheet was removed for the flash picture. The interior is lined with white styrofoam. The bottom of the light mixing chamber is slanted towards the opening, such that the far end is as high as the opening. I use the LED for focusing and positioning and the flash for doing scans.

[Kirk Gittings]

Similar to the Beseler 4x5 color head no?

[Peter J. De Smidt]

Similar to the Beseler 4x5 color head no?

Probably, although I haven't seen a Beseler in 20 years.

[AGChicago]

Peter- adjusting the RGB LEDs to eliminate the mask was what I was thinking would be easier also. It's a bit iterative, but I think can be done. I've ordered some 10W RGB diodes and three PWM LED drivers (with knob adjusters). We'll see how it goes.

What are you using for your diffuser plates?

Allen

[Peter J. De Smidt]

Wow, 10w led's would be bright! (A lot of people are using LEDs for lighting reef aquariums. My fixtures use 3watt cree's, and you can't look at them when they're on full.) Make sure to have them properly heat-sinked.

A few years ago I bought a 4x4ft sheet of white Plexiglas. I still have some left, and that's what I use.

Peter

[Daniel Moore]

Cemil Purut, who is making custom LED enlarger lamps for sale here advised me to use 3/16" #2447 white acrylic, there are a few different translucencies available and this is the most. For the sake of thoroughness, Peter has suggested using non-glare P99 acrylic to mount the neg to.

I just mocked up a light source using the LuminousFilm panel I linked to earlier in this thread. It's an array of 49 LEDs over a total 5 ¼" square area. Placing the 3/16" acrylic 2 ½" above it resulted in very even illumination. The difference between the center and outer edges measured -4 points in photoshop, a simple matter to correct in the final stitched scan. Though I don't hold much hope for it regarding optimal color spectrum I will test that aspect shortly. The manufacturer says the panel reaches a CRI of about 80 and is between 4,000K and 5,000K. Exposure was 1/100' @f/8 ISO 100. Incidentally, this thing barely gets warm.

[Peter J. De Smidt]

That's a nice shutter speed. You might try using a second diffusion panel, ideally with some space between the two.

[Daniel Moore]

A second similar diffuser will help, I think it's 40% transmission so should come close to 1/30th sec. at that point, given some space and drop off. I made the world's worst spectrometer tonight and had a look at this light. It was encouraging not seeing sharp bands but rather smooth blends, similar to but not entirely like an incandescent spectrum. The blues were less smooth and seemed deficient. I plan to build another spectrometer (though really, they should have a different name, the way I build them : P) such that it can be attached to a camera, I'll put up the spectrum when I have it. The encouraging part was trying my splectromodor on other light sources at home and seeing in CFL's and other bulbs what I didn't want to see from my DSLR scanner lamp, narrow clearly defined bands surrounded by gaps. I'm very interested to see what my Epson V700 transarency lamp shows as well. Here's a link (http://publiclaboratory.org/sites/default/files/Make%20a%20Spectrometer.pdf) to the goofy job I plan to construct tomorrow.

In case it's of any interest, here's a sample of a 5"x5" area of the lamp as described at this stage.

82709In case it's not clear to any folks who haven't had the time to read everything posted previously, my present intent is to move the light source along with the negative to sidestep stitching artifacts due to less than perfect illumination or variations in illumination, repeated over individual 'scans' from a fixed light source as Peter's invaluable empirical evidence has shown can be an issue.

Struan, thanks so much for the spectrometer info!

Note to any splectromodor builders: two pieces of blue painter's tape iis very nearly .2mm, a preferred slit width for DIY jobs.

[David_5527]

It seems to me that an x-ray viewing box may be good for this...Also, the standard size if 14x17 "

[Peter J. De Smidt]

It seems to me that an x-ray viewing box may be good for this...Also, the standard size if 14x17 "

Are they more even than standard light boxes? My Porta-trace light box is no where near even enough.

[Peter J. De Smidt]

I finally had a few minutes to take some readings. I used an X-rite 811 densitometer in Status M mode, which is for color negatives. Taking a reading through the un-exposed but processed base of 35mm Fuji 100 color negative film, gave RGB numbers of R.22, G.64, and B.86. I then sandwiched Rosco gels with the negative strip and took readings to find the most equal R, G, B numbers. The winner was #70 Nile Blue. Sandwiched with the Fuji film, it gave R0.92, G1.01, B1.02.

Hopefully, I'll have some time to run some scan and see if this equalization of the RGB numbers helps.