The Backlight


#1

It is mentioned that an RGBW LED array is better for the backlight where the color of the backlight can be adjusted. In practice, is there any advantage in having a color adjustable light source.?
Thanks.


#2

Good question. I’ve wondered about this too. If the LED is a pure white and the camera is white-balanced properly, shouldn’t the natural colors of the film come through accurately?

I have no experience in this part of the design so I’m definitely looking for others to help fill in the knowledge gaps.


#3

I’ll start by saying that the color spectrum of the back-light is important. To make an extreme analogy, you could eliminate the back-light altogether and capture using a night vision camera, but obviously the results would be terrible. So to a lesser extent, that is the case when we rely on electronics to make corrections. Begin with an image that is projected in a form that is as close to the desired result as possible, then “tweak” it via electronic means if necessary, but the goal should be to minimize the amount of post processing, and I would even include the color balance feature on your camera as a form of “post processing”. Some amount of post processing may be desirable (e.g. to clean up or even sharpen a dull image), though when you do that, you want to provide as much original information as possible from the native image.

So the first problem is that “white” LEDs typically have a very skewed color spectrum (see image below), for example, blue tends to be over represented and red tends to be under represented. Have you ever taken a dark image and tried to lighten it via image processing software and noticed that the result now looks grainy? The reason for that is that you can’t electronically add back missing information. So if a particular set of colors are under represented in intensity, then electronic compensation can only work with the available information. It will do its best to compensate, but it can’t add back missing information.

To even begin to compensate for a “white” LED using your camera’s color balance means that you would have to project a known white source (e.g. a white piece of vellum) to your lens using the unbalanced LED and calibrate your camera’s white balance based on that. At that point you are making the assumption that your camera’s white balance completely negates the affect of your off spectrum “white” LED source (i.e. “flattens” the curve in the attached diagram).

There’s also something called “scene by scene” color correction. High end scanning services provide this as an option. It means that the color balance can change from one scene to the next based on a variety of conditions such as the original lighting, original movie camera settings, and film degradation (e.g. the red “vinegar” aging that some 16mm films have). Since it is nearly impossible to know what the original color balance for the particular scene was, the operator must make a subjective determination of what it should look like. This is where having distinct RGB channels can be utilized by the operator to find the most natural and pleasing color balance for the given scene.

I’ll add that when considering a back-light design, considerations such as providing effective diffusion and an appropriate means of dimming are important. If PWM is used, it may cause an interference pattern with some cameras, so in that case, it’s best to rely on analog means rather than digital means to control intensity.

Having said all that, if you’re willing to forego the ability of scene by scene correction, it may be sufficient for your needs to use a “white” LED (better, as I stated above, if your white balance is calibrated to the LED). Ultimately, like anything else, it all depends on whether or not you are happy with the end result.

filter-spectrum


#4

Also, when you are using white LEDs and a RGB camera, you end up having three different color response curves fighting each other: That of the Film, that of the Light and that of the Sensor; since they will never match, you will lose color information. In that regard you are better off with doing multiple, monochrome frame captures of different spectra.

(Warning - this is at the heart of a research project I am working on. www.filmcolors.org )

If anyone on this list is coming to the 3rd Color Conference in March in London, let me know.


#5

Thanks Greg for the lengthy explanation.
I found the following while searching for a solution. (I am no electronics wizard). But it is based on PWM.



Anyone knows the analog way to control the intensity,

Martin, Hope we would get all the information once you finished the research project (if possible).


#6

Martin - Yes, I’ve seen where multiple images are taken of the same frame and later blended as a way to maximize the dynamic range.


#7

Udayarangi - LEDs are controlled by current, not voltage. One simple way is to use a voltage regulator such as an LM317 configured as a current regulator (you should be able to easily find this circuit online). I found an even simpler way that I have adopted that works well and only involves a resistor and a pot. You can find it here (scroll down about half way). https://electronics.stackexchange.com/questions/55477/using-a-variable-resistor-to-dim-an-led/55507


#8

Thanks Greg for the link.
After your explanation I completely understand why color balancing before capture is very important when scanning old, color faded footage,


#9

Greg, I came through that link while searching for a solution. But When I showed them to a friend who is an electronics expert he said I would not be able to control a higher watt LED with this method. What is the wattage of the LED you have tested?
Thanks.


#10

Udayarangi - I agree with your friend. Both methods that I suggested are wasteful of power (LM317 and resistor solutions dissipate power in the form of heat) and so are best applicable to low wattage LEDs. The particular LED that I am using (see below) is only 3W. A backlight does not need to be nearly as strong as a projection lamp so the 3W “star” works fine along with the resistor based circuit for my application. If you need something more powerful, then you will need to use a variable current power supply for each LED. Switching power supplies tend to be the most efficient, however their output should be appropriately filtered to minimize ripple. If your friend has a better solution in mind, I would be interested in knowing what that would be. https://www.ebay.com/itm/5x-3W-RGB-3-Chip-6pin-20mm-Star-LED-Color-Light-Bulb-Lamp-USA-/111795082778?hash=item1a0781961a

Also, you could still try using PWM and verify that it is or isn’t a problem. If it is, then you might be fine filtering the output. Here is a great article I found on the topic. https://www.allaboutcircuits.com/technical-articles/low-pass-filter-a-pwm-signal-into-an-analog-voltage/


#11

as leds turn on and off very quickly pwm is probably a bad idea, you will end up with beating between the shutter speed of your camera, the ccd refresh rate, film transport any number of things. you may end up with flicker or banding of the image depending on how the numbers come out.

on the subject of lighting. you really do want 3 red, green, and blue leds. the point is the dyes are not narrow spectrum, they overlap which mixes the colours and makes accurate colours hard.

broadcast telecine have clever matrices to compensate for this, but you can sidestep all these problems by just having narrow spectrum rgb lightsource.

back in the day, this required dichroic mirrors but now the solution is easy, rgb leds.


#12

Greg - He suggested me to go the PWM way. But as you have stated, a 3W LED will be quite sufficient. Since you have tested with a 3W LED, I think I should first follow your footsteps. Did you use an RGB LED or separate R, G, B LED s? How about RGBW LEDs?

quintile - Thaks for your input. If I understand correctly, you are suggesting to use separate R, G, B LEDs? Pl correct me if I am wrong.

I have received tons of valuable information form you all. Thanks again.


#13

quintile - I was generally advising against using PWM, but if you take a look at the link I posted, with some appropriate RC filters, it does seem possible to virtually eliminating the flicker. Although I haven’t tried it myself, it seems like it might be a viable approach. Here’s an image of the filtered signal.


#14

Udayarangi - I posted a link to the LED I am using in one of my previous messages. It is a “star” led package with individual RGB LEDs underneath the lens. I have no experience with RGBW, but I see no obvious problems in using one. I have read that RGB LEDs have trouble producing a true white light, so the extra white LED might possibly provide a better starting point for white light. Below is a link to a DIY telecine that uses an LED similar to mine. [http://telecine.godiskind.com/#Bright White Light](http://telecine.godiskind.com/#Bright White Light)


#15

PWM is the right way to do. A friend of mine wrote his thesis about LED and various colour conditions.
He used PWM.
Also I would not stop at RGB or RGBW. Most RGB and RGBW LEDs have no continuous spectrum in reproducing white.
In most cases it will be enough for colour film, because also filmprints in the analogue world are sensitive to narrow bands within the red, green and blue parts of the spectrum and actually produce only specific colour tones in the colour layers of the film.
So for a DI process with film in good condition it will work, BUT if you have to deal with film in bad condition (faded) it could be very important to have a continuous spectrum.
On faded film the narrow bands of the layers could be shifted outside the bands of a RGB or RGBW light source. So they would not be that sensitive to that backlight and will be recorded or digitised distorted. Compensating this digitally might be very hard or in some cases even impossible.
And that is a problem of most film scanners that exist. Most of them are designed for DI processes with colour film in brilliant shape. They are not an optimal solution for good digitisation and restoration of material with lot of problems.
So the right way to get the whole information, would be to use a light source with continuous spectrum. That can be achieved by using RGB and CMY LEDs in combination and separated illumination passes with a monochromatic sensor, to get all halftones.
Also a monochromatic sensor is superior in image reproducing to a bayer sensor, because of the effective filling factor and thereby less aliasing.


#16

Hallo Martin, maybe I will be at the conference in London.


#17

What do you all think of an LED like this one? RGBW via PWM. I’ve used them in other projects, but never for imaging. Seems quite versatile and more than powerful enough for our needs. Any red flags?

There are also Natural White (~4500K) and Warm White (~3000K) for the phosphorus element.

M


#18

Andreas - This is what I am referring to.https://www.youtube.com/watch?v=LQDS6z-90sE


#19

I would not recommend the RGBW leds as I suggest you want as narrow a spectra as possible,
ideally R, G, and B line spectra. Failing that the phosphors leds use are fairly narrow - it is (I believe,
all down to the purity of the phosphors used.).

This is similar (though not exactly the same as) what I used:

you need a diffuser next to the film path.

I experimented with a cardboard tube lined with self adhesive mirror foil,
hoping this would retain some of the light I would otherwise lose, but it ruined
the nice even diffuse light I get from the unmodified led so I have dropped the idea.


#20

re: faded film and white light illumination.

sorry this is completely incorrect.

If you have faded film, the three dyes in the film (the vast majority of films have 3 dyes) have faded unevenly.

There are two main processes in film fading - dark (in the can) and light(due to multiple projections).
These cause different losses but fundamentally the way to restore them colour of the film is to restore the
amplitude of the three colour channels.

If your R signal has bits of G and B in it them due to the broad spectrum light illuminating other colour dyes, when you restore the R you will also amplify the G and B which will give you the wrong colours. I am not saying it cannot be done, but its vastly easier to colour correct film (faded or not) where the R control just controls Red.

Broadcast telecines either use RGB illumination, or they use video matrices to unpick this crosstalk (much easier and cheaper before RGB leds where available), though these matrices had to be tailored to the specific film stock used.

You could do this if you like but you would need a precision calibration film to do it - a TAF (telecine alignment film) was made by Kodak for exactly this reason. You could use an it8 target if you can get one to fit your telecine though the maths behind generating and using the 3d lut it would produce is messy.

By far the easiest way is to just use an RGB led.

A TAF or it8 target and luts do allow you to correct the other source of intercolour interlayer effect - the overlapping sensitivity of the dyes during film exposure. Leds don’t help here at all, and if you are doing it really “properly” that would be the path to take.

-Steve