LED Light project

[Michael Collier]

OK,

 

Some might remember my intervelometer project, that is almost done, but on hold. Today I got in the mail 130 white LEDs and have been putting together my panel. I will post some pics when I get home. This panel will differ from normal panels in its diming feature. First, it has a constant frequency. Not sure what that is yet (somewhere around 200Khz, but its flexible and I will set that when its ready) Secondly the dimmer won't be a simple 0-100 potentiometer style. I have built the circut to decrease or increase intensity exactly 1/3 stop each time the button is pressed. I figure that would be a bit more useful. After testing though I found I need a rotary encoder, since it has about a 7 stop useful range, which is 21 button presses to go from full on to nearly off. way too much. I plan to put 7 LEDs on top to indicate how many stops it has been cut by (one LED for each stop).

 

I had no color meter, but putting my camera on preset the LEDs looked very close to 5600K. for production runs I will be able to buy LEDs that are tested at the factory to guarantee a temp value.

 

 

After testing the LEDs I have found at 4ft they can throw about 2 footcandels. So I figure at 4ft average distance (for outdoor lighting of interviews) I can get up to 260 footcandles. That means on an average day with the direct sun at 500ft/cn I can achieve a 2:1 ratio, even if dark folliage is sucking up all the ambient light. I am also planning a 400 LED and 800 LED model comming soon. These will be availible for purchase sometime next year after I iron out all the details of the light. They should be cheaper than almost any light panel availible today. Pics to follow soon. I am happy to answer any and all questions.

 

....and yes this will definatley kill film....somehow. Its called blue.

[chuck colburn]

Hi Michael,

 

Wondering why you picked 200,000hz as your oscilator freq.

[Xavier Plaza]

After testing the LEDs I have found at 4ft they can throw about 2 footcandels. So I figure at 4ft average distance (for outdoor lighting of interviews) I can get up to 260 footcandles.

 

 

Hi Michael interesting project i would like to see it soon..., when you say LED produce 2fc at 4ft, you're talking about each LED or all...?, i guess you are talking about each LED, or i'm wrong?

 

 

Good luck

Edited June 29, 2007 by Xavier Plaza

[Michael Collier]

Hi,

 

Yes you are right, 2fc per LED at 4ft. so times 130 its 260.

 

And I should have said 200hz, not 200Khz, my mistake. I am not sure, but its around there. I actually haven't put much thought to the frequency yet, since the for the program it really doesn't matter. The dimming feature works by a variable that is set, and a seperate timing routine sets the frequency. Right now I just have arbitrary numbers in there to give it long enough for the LED to turn on, but short enough that I don't see a flicker by eye (or as I tested today at 29.97 fps) Once I get the prototype pannel set up, I will decide on frequency. Right now I am testing just a single LED with the circut.

 

Pictures will be avalible tommorow, but today, its a beautiful summer day in Alaska and we only get 4 months of summer, and 8 months of winter....so its time to play some frisbee golf.

[Luke Prendergast]

You will likely have to add some diffusion to clean up the beam, and suffer the stop loss, as focussed LEDs tend to project a very ugly beam.

[Hal Smith]

Watch out for heat build-up. I just built an LED exciter lamp retrofit kit for my Simplex projector using a Phillips/Sylvania Luxeon K2 Star from http://www.ledsupply.com/05027-pd12.php and was surprised at how hot it gets.

[chuck colburn]

Hi,

 

Yes you are right, 2fc per LED at 4ft. so times 130 its 260.

 

And I should have said 200hz, not 200Khz, my mistake. I am not sure, but its around there. I actually haven't put much thought to the frequency yet, since the for the program it really doesn't matter. The dimming feature works by a variable that is set, and a seperate timing routine sets the frequency. Right now I just have arbitrary numbers in there to give it long enough for the LED to turn on, but short enough that I don't see a flicker by eye (or as I tested today at 29.97 fps) Once I get the prototype pannel set up, I will decide on frequency. Right now I am testing just a single LED with the circut.

 

Pictures will be avalible tommorow, but today, its a beautiful summer day in Alaska and we only get 4 months of summer, and 8 months of winter....so its time to play some frisbee golf.

 

Hi Michael,

 

This months (June) issue of Electronic Products has a section on LED lighting that might be of intrest to you. Scroll down about half way for articles in that issue.

 

http://www.electronicproducts.com/

 

Chuck

[Richard Andrewski]

Watch out for heat build-up. I just built an LED exciter lamp retrofit kit for my Simplex projector using a Phillips/Sylvania Luxeon K2 Star from http://www.ledsupply.com/05027-pd12.php and was surprised at how hot it gets.

 

No question about it. Those Luxeons get hot and even some models of smaller LEDs too. The main problem is that LEDs simply don't have the lumen per watt performance yet to really surpass other forms of lighting but they are on the way. In labs they've achieved over 100 lpw which is super but in actual use, most LEDs available today come in at 30 LPW and that's barely better than the best tungsten at 20 LPW. So why switch? And it won't surpass fluorescent yet at around 50 to 80 LPW either.

 

Single, super high wattage LEDs are coming though. I have one in the 60w range that I got in Asia and I've seen a 150w model here in China too. Basically a whitish spot about 3/4" inch in diameter on a PCB with some metal around it to attach to a heatsink. Very interesting stuff but the main problem with them right now is they do need large heat sinks to keep them alive for their stated lifetime. Fans would probably be even better. The problem is, for our uses, fans don't come in too handy with microphones around! And this isn't like some electronic ballast you can put in another room because its noisy. The fan has to be on the light unit to be effective.

 

So we'll continue to wait and watch what's happening there but when LEDs approach 85 LPW and get to the 150w actual draw range with around 12000 lumen output, then they will start to rival metal halide (HMI) for performance and may overtake it. Because at that point, you are talking about a single point light source and that's what we really need for fresnels and pars--but that's a ways off. Till then, LEDs will mostly be usable for portable softlights especially great in situations where your power challenged and/or really need super portability.

 

I included a picture of the 60w I've got...

[Michael Collier]

Yes, K2 (and all luxeon lines) have a huge power dissapation. I think I read somewhere that on their 1watt models, something like 200mW-250mW must be dissapated through heat. Their data sheet specifies they MUST have a heat sink, or they'll die quickly. The plastic 5mm kinds are much more effecient and more robust (less heat dissapation anyway, K2s are more power effecient)

 

I have a plan for heat reduction in my 400 and 800 light models I have planned. Reluctant to say how though, hall can figure it out I'm sure. No fans of course

 

Also Rich said you need a single point source to make a par or frezzy, I have a design now to make a fresnel without needing a single point light. I won't say how, I think I can get it pattented. I will be attempting to build a 1.2K, 5K and 18K led fresnel light....down the road (equivalent HMI light output, they will actually be 1.5K, 6K and 21K lights delivering 110,000 lumens, 450,000 lumens, and 1.6M lumens respectivly.....no cooling fans needed)

 

The panel I am building now I want to test it with some light diffusion, I don't want to loose the punch these lights will have, but I do need the light to look a bit nicer. Also to that end each light will be optically aligned, something not done with other pannels. I think a light hampshire frost or opal might do the trick, and of course stronger diffusion can be put on if you need a softer output.

 

I like LEDs for the simplicity (of operation, not of construction) and the power savings. True, LEDs are only 100% more effecient than tungsten (the models I am running at least) and about 50% of an HMI, but lets face it, with an HMI or a tungsten, we are going to place a lamp more powerful than we need. I might throw up a 2K tungsten, but as I am lighting, I will probably scrim the light, wasting power. With LEDs when you need to cut the light a stop, you dial it into the controller, and the unit is instantly using 50% of the power.

 

Also LEDs offer a huge range of power input. 110, 220, 50hz, 60hz, 12vdc, 14vdc, 30vdc, it really doesn't matter. Building a converter is cheap and easy to bring it to the voltage the controler needs. I could even make a universal power adapter, that could accept all those voltages. Fine adjustments are fast and easy, I think LEDs are the wave of the future.....and I have some really really big plans....foundation shifting plans for LEDs on set. I can't talk too much about it but I have great ideas for this.

 

Those 60w and 150w lights are pretty interesting, I think I have some reading to do and decide weather to alter my plans any. Are they avalible in the US, or do I need to travel to china to pick one up?

[Richard Andrewski]

Yes, K2 (and all luxeon lines) have a huge power dissapation. I think I read somewhere that on their 1watt models, something like 200mW-250mW must be dissapated through heat. Their data sheet specifies they MUST have a heat sink, or they'll die quickly. The plastic 5mm kinds are much more effecient and more robust (less heat dissapation anyway, K2s are more power effecient)

 

I have a plan for heat reduction in my 400 and 800 light models I have planned. Reluctant to say how though, hall can figure it out I'm sure. No fans of course

 

Also Rich said you need a single point source to make a par or frezzy, I have a design now to make a fresnel without needing a single point light. I won't say how, I think I can get it pattented. I will be attempting to build a 1.2K, 5K and 18K led fresnel light....down the road (equivalent HMI light output, they will actually be 1.5K, 6K and 21K lights delivering 110,000 lumens, 450,000 lumens, and 1.6M lumens respectivly.....no cooling fans needed)

 

The panel I am building now I want to test it with some light diffusion, I don't want to loose the punch these lights will have, but I do need the light to look a bit nicer. Also to that end each light will be optically aligned, something not done with other pannels. I think a light hampshire frost or opal might do the trick, and of course stronger diffusion can be put on if you need a softer output.

 

I like LEDs for the simplicity (of operation, not of construction) and the power savings. True, LEDs are only 100% more effecient than tungsten (the models I am running at least) and about 50% of an HMI, but lets face it, with an HMI or a tungsten, we are going to place a lamp more powerful than we need. I might throw up a 2K tungsten, but as I am lighting, I will probably scrim the light, wasting power. With LEDs when you need to cut the light a stop, you dial it into the controller, and the unit is instantly using 50% of the power.

 

Also LEDs offer a huge range of power input. 110, 220, 50hz, 60hz, 12vdc, 14vdc, 30vdc, it really doesn't matter. Building a converter is cheap and easy to bring it to the voltage the controler needs. I could even make a universal power adapter, that could accept all those voltages. Fine adjustments are fast and easy, I think LEDs are the wave of the future.....and I have some really really big plans....foundation shifting plans for LEDs on set. I can't talk too much about it but I have great ideas for this.

 

Those 60w and 150w lights are pretty interesting, I think I have some reading to do and decide weather to alter my plans any. Are they avalible in the US, or do I need to travel to china to pick one up?

 

I haven't seen the single super wattage units in the USA yet. All the major LED development seems to be happening here in Asia right now. Also, for those super concerned about CRI, its hard to get an LED to do much over 80. While I believe this to be not too bad and certainly no problem for video use, many believe they have to have 90 or greater to have something usable for their purposes. You have to remember that most of these companies are developing these things for use in street lights and situations where there may be a solar panel driver too. Our market for film and video production is a tiny percentage of the world lighting market and very specialized and finicky in its needs. These companies don't always appreciate our needs for color rendering and color temperature. When it'll become more interesting will be after the street light ideas get developed somewhat and then they start perfecting their processes and can work on the color temp and CRI issues.

 

The really nice thing about LEDs though is for the person that needs to operate on rechargeable batteries for any semblance of a reasonable time, you really can't beat them. Great for use for filming in the car at night for instance; in those small cramped corners of a set; or for the guy doing documentaries in the backwoods of Africa, etc.

[Michael Collier]

Funny you should mention shooting in cars. I am DP for a shoot this coming month (just got upgraded from a digibeta shoot to a 16mm shoot!) I told the director I was building this specificaly for the car scene, and his eyes got really wide and he got really excited. No proccess trailer is avalible for the shoot, so hiding lights in the cluster and roof is the only option for lighting.

 

As for CRI, I couldn't descern any green shift on the LEDs I had, even after shooting them on preset daylight at various levels of over and underexposure, and checking the results with a color picker and vectroscope. Also all the LED panels avalible today don't have neg green attached to them. Where did you see an LED with a CRI spec? I haven't seen that in any data sheet.

[Richard Andrewski]

Also all the LED panels avalible today don't have neg green attached to them. Where did you see an LED with a CRI spec? I haven't seen that in any data sheet.

 

That's my point. The LED manufacturers don't even consider it a spec to quote for the most part. But its fully possible to measure it and one manufacturer I know of over here does measure it. The machines at the end of the line for quality checking sort the LEDs out into bins by color temperature and quality level. Mostly though, you are buying "white" or "warm white" which could be a variation of several hundred in Kelvin.

 

It'll be a while before they become as sophisticated as the fluorescent and metal halide markets and start measuring and quoting that regularly. Some really do have a shift toward blue especially and put out a very low quality of light. My first panel I built last year (with dimmer) had a bit of color shift to green when dimming. I still use it though sometimes when I need a fill light and its fine at the higher intensities. Not very pretty, I included a picture--it was in a shoe box lid ;-). The LEDs in that were not the best though. It was enclosed and sealed strips of LEDs and it was when I decided a panel of discretely mounted quality selected premium LEDs was the way to go.

 

I've seen all kinds of LED products over here from cheap to expensive. As for color temperatures, you pretty much are left to a range that they quote and you hope you can get within that. Its much harder to get a specific color temperature and stay within that tolerance. Those really are the "selected" premium group and carry a higher price.

[Phil Rhodes]

Hi,

 

Won't 200Hz be a bit slow? You could easily hit flicker problems with even fairly everyday amounts of overcranking.

 

They go purple as they age because the yellow conversion phosphors break down. It would be reasonable to presume that the warmer the colour, the worse this will be, since more of the output will be coming from the phosphors.

 

Phil

[Hal Smith]

Cutting (bleeding?) edge discussion gentlemen, thank you!

[Richard Andrewski]

Cutting (bleeding?) edge discussion gentlemen, thank you!

 

Hi Hal,

 

I was raised in OKC. Looks like you're in Edmond but we'll let it go ;-). Yes, this really is "bleeding edge" stuff and pretty interesting. As I'm sure you're aware since you've been playing with LEDs, LED dimming is not as easy as it might seem. It requires a special circuit akin to a motor speed controller. Some of the best ones have a 555 timer in them along with a power transistor, the choice of which will affect the amount of wattage you can drive.

Edited June 30, 2007 by Richard Andrewski

[Phil Rhodes]

Hi,

 

What we're talking about is this: http://en.wikipedia.org/wiki/Pulse_width_modulation

 

It isn't actually necessary to PWM drive LEDs - you can do it variable voltage, but really what we're talking about here is a switch-mode power supply in either case, either with or without output smoothing capacitors. This is by far the most efficient approach for a typically battery-powered item, especially when it needs to be small and portable.

 

If you want to achieve really accurate dimming of white (or other pumped-phosphor) LEDs, PWM may not be ideal since you will see an afterglow from the phosphor which will fill in the off-periods and increase apparent brightness, as well as bending colour reproduction since the blue component of the light will shut off more quickly than the yellow. This may explain our correspondent's reports of changes in colour at low dimming ratios.

 

These concerns are probably only of theoretical interest for anyone working outside an optical physics lab, but I'm not sure.

 

Phil

[Richard Andrewski]

You're right. You don't need PWM to drive LEDs--unless you want to dim them. In my research, there really isn't a better way though because it's not just a simple matter of turning down the voltage to dim the lights or turning it back up to make them bright.

[Phil Rhodes]

> it's not just a simple matter of turning down the voltage to dim the lights or turning it back up to make them bright.

 

Well, it is - it just won't be very linear, so you'll end up building a switch-mode power supply anyway - but there's no reason you shouldn't smooth it either way.

 

Phil

[Michael Collier]

In my tests this weekend (expanded from one LED test to now 48) it seemed that running the LEDs at a lower voltage produced a poor quality of light, and I think PWM without smoothing is the best way to go. The LEDs snap on very quickly, so PWM is perfect. Running at less than 3.6v @ 20mA seems to make them want to do odd things. Also my goal is to achieve perfect 1/3 stop incriments, something very hard to do with out PWM.

 

The problem I am having is the output amp just doesn't want to work. I even built a pre-amp stage since it seems the microcontroller doesn't want to saturate, but that transistor seems to be having problems too. Back to the drawing board. At full voltage it easily biases and my pannel snaps on and is REDICULOUSLY bright (I had to add two layers of 250 bc I was being blinded in tests) Hook the same amp to the uController and I get a dim output (barly biased at all)

 

I suspect the issue is in the microcontroller itself, and not in the output stage. Perhaps I need to have a pullup resistor set to get it to work.

 

 

Also running into problems with power. My test bench only has 5v@1A and 0-16v @ 300mA avalible to it, which really isn't enough. I have been trying to scavange any power adaptor that has a 20w capacity, but I haven't found anything laying around. In good news I found a power regulator chip (mosfet type) that can accept anything from 12vdc up to 40vdc, so I will have a wide range of input power to drive this thing.

 

Oh well, more experimenting needs to be done.

[Richard Andrewski]

Yes, this is exactly what I'm talking about and why just turning down the voltage is not enough to dim LEDs. You simply don't get a full range dimming and even may be hurting the LEDs by starving them of current with this method. What happens is exactly what you described where the LEDs simply turn off at some point and never go much dimmer. The PWM is the most universally accepted way to dim LEDs in commercial products and it's just a question of keeping the pulse width high enough to not become a flickering problem. The frequency is usually determined by however you set the 555 timer with its associated circuitry and the wattage of your dimmer is determined by the size of power transistor and how well you heat sink it.

[Phil Rhodes]

I'm not sure I've ever read of current starvation damaging LED junctions. Where'd you get that from?

 

I think you meant high frequency rather than pulse width to avoid flicker.

 

Making FETs saturate is notoriously tricky and I am not the slightest surprised that your uC won't do it. Modern SMPS-style chopper transistors have minute gate impedance and can draw whole amps for the microseconds it (should) take to push the device into minimum on-resistance (because you probably want a voltage amplifer to drive LEDs unless you have been very clever and want to use the FET in saturation mode as a constant-current source). You can get so-called "digital FETs" which have all the necessary buffering and amplification on board, which are very convenient. Or, get yourself a decent dedicated FET driver IC, preferably one advertised as matched to the FETs you're using. My experience is with driving cold-cathode lamps with the International Rectifier IRFIZ24N and ICL7667 FET driver, which are reasonably well matched..

 

In either case you will spend some time playing with resistance and capacitance to keep square waves square. Keep leads short and don't assume that what works on the bench will work on a radically different PCB layout. You will want to look at the FET gates with a 'scope and ensure they're not spending too much time in transconductance, especially where that's caused by inductive ripple. This can screw you up far more than poor heatsinking.

 

Then you have to figure out how you're going to stop it radiating wideband RFI and pissing-off the sound department.

 

Enjoy!

 

Phil

[Michael Collier]

Low current won't damage the LEDs, but if I were to keep current in the logarithmic nature I would like, it would not necisarily drive the LEDs output in that fassion, meaning I need to program a LUT to make up for the non-linearity. I need to ensure that if the LED is on, its at full power, and if its off its at no power, and if I need to dim, then I need to PWM the time its in each state to control brightness, rather than controling the current applied. That seems to be the most effecient way to regulate brightness in a controlable, predictable manner.

 

But yes, I did mean I need to up the frequency to prevent flicker. Also I am considering making two or four different output channels, so that although a particular part of the panel might be off, at least some part is on. 4 channels switching at different times would reduce EMI radiation.

 

It should be noted I am not making this with a 555 timer. All other products on the market use a 555 timer, and that is why they only have 0-100 dimming, without calibration. I am using a microcontroller to oscilate the dimming, because I can calibrate it in stops rather than percent.

 

I haven't really been looking at the FETs for switching, though if I were on the 555 I might be. the 555 can sink/source up to 200mA. The chip I am using can only sink/source 20mA, and I suspect that is my problem. To aleviate it, I am going to a tripple output Darlington array. Tripple output since I only have 3904 and 3906 on hand, but 3 or 4 together should be able to drive all 13-17 watts. Those digital FETs are interesting, but if the darlington works, then darlington it shall be.

 

Square waves will be square, as my uC determines that pretty well, and I should be able to input to the darlington without any more resistence than a saftey resistor, if that. A crystal controls the uC, so that is stable.

 

I will have completed panels avalible for purchase within 4 months. I am loving this 1/3 stop notion, it already seems more useful than the pannels I have used. We shall see though. My first real shoot will be in about a month, we shall see how well it works on film (no word yet on stock, the director just mentioned he may be able to get film, I don't know if stock is pre-determined or if I get to choose)

 

Oh well stay tuned.... (full scale panel should be assembled and tested as soon as this weekend, as soon as the 48 LED scale works out tonight)

[Phil Rhodes]

Hi,

 

> Square waves will be square, as my uC determines that pretty well,

 

This is the voltage at the gate of my CCFL driver, without the inductive components attached:

 

 

Feeding a square wave into it will emphatically not ensure you get a square wave out of it.

 

Phil

[Michael Collier]

NEW PICS!!! (or first pics)

 

 

 

I tested it today and it did put out a decent light compared to outdoor direct light (and it was BRIGHT and contrasty today) Its still not in a housing so I can't use it in the feild yet. Also the controller isn't fully working, so its an off or on only deal now. In those pics I have it hooked up to my bench power supply, which can't source more than 300mA. Accross 130 LEDs, thats about 2.3mA. The final power will be around 30mA per LED.

 

I'm very excited to get to use it, I will get about a months worth of practice in ENG shooting, and then next month I have a music video shoot, and a short film shoot (both likely to be 16mm). The latter shoot has me excited. We have a scene in a car with no chance of a proccess trailer, so having an LED in the instrument cluster would be excelent.

 

Point taken Phil. I meant I only need square wave of adequate integrity that it does not affect my LEDs. I meant that the square cannot be averaged (through capacatance or other means) into a lower forward voltage, since the LEDs would become innefficient and have poor color quality. It needs to reach full forward voltage in aproxamatley the time it takes the LED to turn on, which my output stage has done well enough.

 

I will probably build several pannels before I begin selling them off, just to find the perfect combination of LEDs and power and space requirements. Also I gotta find a good case I can sell them in. This pannel is destined for an old Omni housing I have around, but I can't sell them like that.

 

Oh well, first test complete.

[Jonathan Bowerbank]

Look excellent Michael! Can't wait to see it in its housing.

 

Just curious how much you might have spent in acquiring all the materials needed. Has it proven cost effective compared to just buying a brand LED light?