Backlight for Telecine- LED RGB?

[Roving6]

I did a bit of searching on LED backlights here and kept coming across 40pg threads...spent some time wading through those and then thought I'd ask for help!

I like the idea of the adjustability offered by RGB LED backlight and I think I can devise a controller circuit... either an analog servo setup or PWM using a Basic Stamp. What kind of total wattage is appropriate for a backlight? Any specific recommendations for LEDs? (i.e. part numbers?) I was looking at some Cree RGBW LEDs... http://www.cree.com/products/pdf/XLampMC-Ecolor.pdf I was thinking that it might be nice to use an RGBW led...have the white section on full blast and then do the color trimming with the individual LEDs.

A stud mount LED might simplify mounting and heatsinking compared to these SMD devices-- does anyone make such a thing?

Does anyone still use the halogen lamp + a blue filter? Is it good enough?

Nathan

[RCBasher]

Nathan,

I would recommend reading the section concerning LEDs on my website...you will find it under the Various Ramblings section. (Fred has "reprinted" on his site too)

I've spent the last two or three years on LED backlighting...there are many pitfalls and problems, but also many benefits if one gets it right.

Mixing white plus red and blue is a possibilty, although white LEDs are made with a blue LED + white phospor, they turn out predominately green and require more blue and red to balance up. There is little advantage with this approach, better to stick with white only plus a filter (Fred may advise here as he found a good one) or just RGB.

The big problem with LEDs is their output drifts with temperature...and reds drift twice as much as green and blues. You would need to compensate for this. Avoid PWM and conventional switching LED drivers because you will get beat patterns. I use short exposures to control colour balance and to keep the heat down. If you intend to do anything like my LED solution, then forget about BASIC Stamps...nowhere near fast enough or precise. This kind of solution needs very accurate timing or you will get frame to frame flicker. I'm using the hardware CCP modules in a PIC microcontroller and the code is written assembler. I have 0.125us jitter with 500us and upwards pulses.

Frank

[BigWorm]

Nathan,

I purchased a Lighting Science Group (formerly Lamina) Development kit DK-04D0-0394. The kit comes with two LED light engines attached to heat sinks with lenses and the power supply. The RGB kit number is DK-04F0-0396. It's the same light source that's used in the moviestuff line of products. Either one can be sourced form Mouser Electronics. I hope this helps.

R/

[Roving6]

Frank, I did read that section. Your solutions are always very elegant, but perhaps too complicated for me! I could just do an analog servo system (no PWM)... run fiber optic from the LEDs to photo transistors for feedback and a simple op-amp based DC servo which varies the voltage to the LEDs. That would negate the thermal tracking issue and there wouldn't be any PWM related interference.

If I went for an "always on" solution, surely the LEDs would reach some sort of state of stasis after a warm up period? I could just turn the LEDs on, let them come up to temp, adjust my color outputs and away we go. I understand the problem with LED dissipation, but I'll probably have a right angle light path and can do some generous heatsinking.

What kind of wattage should I be looking at for the LEDs?

[Roving6]

Nathan,

I purchased a Lighting Science Group (formerly Lamina) Development kit DK-04D0-0394. The kit comes with two LED light engines attached to heat sinks with lenses and the power supply. The RGB kit number is DK-04F0-0396. It's the same light source that's used in the moviestuff line of products. Either one can be sourced form Mouser Electronics. I hope this helps.

R/

Thanks! I'm probably going to try to roll my own rather than go for a complete kit, but that looks like a nice solution assuming there aren't problems with however the LED output is modulated. Your tip got me looking at some of the Lamina LED arrays like this one: http://www.mouser.com/ProductDetail/Lig ... WDZA%3d%3d

Nathan

[christoph]

I could just do an analog servo system (no PWM)... run fiber optic from the LEDs to photo transistors for feedback and a simple op-amp based DC servo which varies the voltage to the LEDs. That would negate the thermal tracking issue and there wouldn't be any PWM related interference.

funny, as far i know frank did a prototype with fibre to an optical sensor too. didnt work too well for some reason though :/

If I went for an "always on" solution, surely the LEDs would reach some sort of state of stasis after a warm up period? I could just turn the LEDs on, let them come up to temp, adjust my color outputs and away we go. I understand the problem with LED dissipation, but I'll probably have a right angle light path and can do some generous heatsinking.

yes, you can try to get a stable temperature but there are two problems as far as i can see (well, most of it comes from my conversations with frank ;)
one is that if you need to adjust the backlight a lot, you'll always need to wait until it stabilizes again. then fine tune, then wait again. this can become quite slow.

also, if you want to build a circuit which keeps light output constant, you run into the risk of thermal runaway. ie if the red leds get hotter, they have lower light output, so the driver increases the current which in turn makes them hotter.. etc etc.. BOOOM :)
i guess with a big enough heatsink you could overcome the issue but it's another trap to watch out for.

What kind of wattage should I be looking at for the LEDs?

if you have efficient ones, like the latest rebels, then about 8 at 3W should give you enough output for nice diffusion.

all that said, there's also a non-trivial problem with LEDs color matching. since the light source is very pure and colors well separated, the output files will have quite high saturation, so for best results it's a good idea to introduce some crosstalk in post-processing. this is less of a problem with reversal film then with negative film.

white LEDs have a green bias , as frank mentioned. they can look quite nice but it's quite hard to get good skin tones.

halogen with some CTB gels (or a 80A filter) is a surprisingly good option, since you can take advantage of the manufacturers color matching of the ccd bayer filter.

++ christoph

[RCBasher]

Nathan,

I purchased a Lighting Science Group (formerly Lamina) Development kit DK-04D0-0394. The kit comes with two LED light engines attached to heat sinks with lenses and the power supply. The RGB kit number is DK-04F0-0396. It's the same light source that's used in the moviestuff line of products. Either one can be sourced form Mouser Electronics. I hope this helps.

R/

How well does it work? Looks expensive, but perhaps ok if it is "ready to go".

The specs' do not look too great, just 120 lm, no mention of stabilised colour balance, etc.

My solution cost about half the price but offers 1640 lm in an area about 1sqin, stabilised, computer controlled so will allow for automatic white balance based on the image histogram...but I have no plans to sell it. Just highlighting what *can* be acheived by DIY ;)

Frank

[RCBasher]

I could just do an analog servo system (no PWM)... run fiber optic from the LEDs to photo transistors for feedback and a simple op-amp based DC servo which varies the voltage to the LEDs. That would negate the thermal tracking issue and there wouldn't be any PWM related interference.

If you are going to go down this route, don't waste time with the optical approach...too much cross-talk when sampling the light and mis-match of detectors to LEDs. The responses of the LEDs and detectors are quite sharp when combined and a change in LED temperature changes the light output frequency which then makes for an unstable feedback solution. Monitoring temperature and providing a fixed correction factor (it is pretty much linear) per colour does work well. You should be able to acheive this in an analogue circuit...but watch out for thermal runaway, as Christoph has warned.

Switching the LEDs off bewteen exposures is a very good way to keep the temperature down, even if you do not use switching for colour balance/exposure. Let's say you want exposures around 5ms, and are capturing around 10fps, even if you light the LEDs for 10ms, heat generation will only be 10%.

Frank

[christoph]

Switching the LEDs off bewteen exposures is a very good way to keep the temperature down, even if you do not use switching for colour balance/exposure. Let's say you want exposures around 5ms, and are capturing around 10fps, even if you light the LEDs for 10ms, heat generation will only be 10%.

true, but it also makes temperature sensor and compensation an essential part of the system, as otherwise changing frame rate will change the frequency and thus the temperature and thus the light output of the LEDs.

so an always on solution would be easier to keep stabilized, but slower to adjust, and one probably couldnt run it as the same current.
++ christoph

[RCBasher]

Switching the LEDs off bewteen exposures is a very good way to keep the temperature down, even if you do not use switching for colour balance/exposure. Let's say you want exposures around 5ms, and are capturing around 10fps, even if you light the LEDs for 10ms, heat generation will only be 10%.

true, but it also makes temperature sensor and compensation an essential part of the system, as otherwise changing frame rate will change the frequency and thus the temperature and thus the light output of the LEDs.

so an always on solution would be easier to keep stabilized, but slower to adjust, and one probably couldnt run it as the same current.

Except I didn't mean instead of temperature compensation, I meant as well as. Switching will help to reduce the possibility of thermal runaway. Heatsinking acts like a capacitor and smooths out the thermal pulses due to switching, so compensation can act over a longer time period.

Frank

[aj]

If colour and temperature are not so easy to control it makes one wonder why LED are used in coldlightsources for photographic enlargers.
Like this one: http://www.heilandelectronic.de/html/de ... t_main.htm


I was even thinking of constructing such a thing for my Minox 8x11mm enlarger. Although at Heiland they expect there will not be enough light coming from such a small group of LED. Still they were kind enough to document the output wiring of the Splitgrade computer.

What would the chances be for a 24x36mm surface. Only suitable for 10x15cm prints?

[RCBasher]

If colour and temperature are not so easy to control it makes one wonder why LED are used in coldlightsources for photographic enlargers.

Drift with temperature is not a myth, it's a fact....it is colour dependant and is severe. Just take a look at a datasheet: http://www.philipslumileds.com/pdfs/DS65.pdf page 13 to see for yourself. Just a 10 degree C change of temperature will throw white balance significantly. Reds drift at twice the rate of blue/greens. All LEDs are the same, the drift depends on the colour because of the different technology used for the different colours.

Also, a single-shot photo has less need for stability compared with video as each frame will like be balanced per frame and then the exposure made.

The light output per area is also pretty low for that head, so there is plenty of scope for a large amount of heatsinking. Exposures would normally be in seconds, not ms!

Frank

[aj]

If colour and temperature are not so easy to control it makes one wonder why LED are used in coldlightsources for photographic enlargers.

Drift with temperature is not a myth, it's a fact....it is colour dependant and is severe. Just take a look at a datasheet: http://www.philipslumileds.com/pdfs/DS65.pdf page 13 to see for yourself. Just a 10 degree C change of temperature will throw white balance significantly. Reds drift at twice the rate of blue/greens. All LEDs are the same, the drift depends on the colour because of the different technology used for the different colours.

Also, a single-shot photo has less need for stability compared with video as each frame will like be balanced per frame and then the exposure made.

The light output per area is also pretty low for that head, so there is plenty of scope for a large amount of heatsinking. Exposures would normally be in seconds, not ms!

Frank

OK, I'll keep it in mind should I manage to get something working

[Roving6]

Monitoring temperature and providing a fixed correction factor (it is pretty much linear) per colour does work well. You should be able to acheive this in an analogue circuit...but watch out for thermal runaway, as Christoph has warned.

Switching the LEDs off bewteen exposures is a very good way to keep the temperature down, even if you do not use switching for colour balance/exposure. Let's say you want exposures around 5ms, and are capturing around 10fps, even if you light the LEDs for 10ms, heat generation will only be 10%.

Frank

OK, that's good info- sounds simpler too. I can rig thermal compensation and pulse the LEDs - even just crudely cutting them off during frame advance would help keep the dissapation down. It's interesting that the change in LED output with temp is linear, as most semiconductors have log changes in conductivity with temperature. Then again, light intensity follows a log rule (yes?) so perhaps the two log slopes somehow cancel each other out to make a linear function? I'm a bit out of my depth here!

I'll probably start out with halogen, then upgrade to LED for the MKII version. It's not too expensive to experiment with light sources and should be fun.

Nathan

[Roving6]

What kind of wattage should I be looking at for the LEDs?

if you have efficient ones, like the latest rebels, then about 8 at 3W should give you enough output for nice diffusion.

all that said, there's also a non-trivial problem with LEDs color matching. since the light source is very pure and colors well separated, the output files will have quite high saturation, so for best results it's a good idea to introduce some crosstalk in post-processing. this is less of a problem with reversal film then with negative film.

white LEDs have a green bias , as frank mentioned. they can look quite nice but it's quite hard to get good skin tones.

halogen with some CTB gels (or a 80A filter) is a surprisingly good option, since you can take advantage of the manufacturers color matching of the ccd bayer filter.

++ christoph

Maybe this is daft of me, but do you mean 8 LEDs for a total of 3w? Or 8 LEDs @ 3W apiece? The former sounds low from my earlier camcorder experiments and the latter sounds rather high.

Nathan