Hey all, I am planning on starting a little experiment to see how capable the current generation of super bright LED's is at replacing a comercial projector lamp in a DLP projector. I know a lot of you are LCD all the way and that is cool, but given the cheapness of second hand DLP projectors with burned out bulbs, I now think this is a viable way to go for DIY. I would also be interested in seeing any results if someone wants to mod a DLP projector to use a typical LL lamp and ballast if anyone wants to try that experiment.

Anyway, I am posting this as my plog so here are the details:

I bought a 1024x768 DLP projector second hand on EBAY for $300, working, just needs bulb. I plan on creating a cicular array of 25 of these LED's:

http://www.luxeonstar.com/item.php?id=1811...o=LXK2-PW14-U00

With these 3 degree lenses on them:

http://www.luxeonstar.com/item.php?id=2214...rtno=OPK2-1-003

For a total cost of $202 for the LED's and lenses, then add about $50 for solder, circuit boards, heatsinks, etc.

This will give me approximately 3250 total lumens of which > 2763 lumens will utimately be directed at the DLP chip. I am hoping to harness at least 60% of the directed light at the output of the projector, ending up with a minimum brightness of 1658 lumens.

Total project cost: $552 + lots of fun and pictures.

Anyone want to call BS on this? Well, I don't care, I plan on doing it anyway. I will post results as I achieve them.

Remy

We look forward to your adventure ! Good Luck !

cheers,
gs

So doing a little research, I found this article written by DJ Segler from TI (developers of DLP technology):

http://www.digitaltvdesignline.com/howto/s...3069&pgno=2

It looks like even the creators of DLP are experimenting with this option. If you look at the references at the bottom of the article it appears that they are using Luxeon (aka lumiled) LED's for their test implementation. Pretty interesting stuff, just thought I would share.

Remy

You may want to look into color LEDs, in regular DLP (not the high end ones) they use 1 DMD (Digital Micromirror Device) and one bulb. They seperate out the colors with a color wheel. If you take all the visible light as 100% then in a perfect world the blue filter would let through 33%, green 33% and red 33%. We don't live in a perfect world so those numbers are lower, I don't have specific numbers but they are lower. So this means that we are using enough power for 130 lumens (per LED) but only getting 43 lumens at a time (actually less). But they sell green, blue and red LEDs the Red and Green hit 130 lumens while the blue is about half that. So with RGB LEDs you could run each LED at a 33% duty cycle (removing the color wheel and get the same of better performance (filters seem to pass about 90% at best) at 1/3 the power. As the article you mention states the bigest problem is heat so only running the LEDs 1/3 of the time will greatly reduce the heat and improve the LED performance. You would still need the same number of LEDs but use only 1/3 the power and stop the LEDs from over heating. Control is more difficult but nothing worthwhile is ever easy. I've had similar ideas to this in the past, so I look forward to your results.

Yeh, I have looked into duty cycle reduction through the use of three seperate colors, the problem with that is then I need to create my own driver board for timing on the LED's. The way that Samsung, Phillips, and Sony are doing theres is to have three LED arrays, one array each for Red, Green and Blue. There are two dichroic filters that allow the different colored arrays to be reflected into the primary light pipe, through a condensor lense, then into an optical integrator that combines the colors as well as causing the light to be created in a square pattern. Here is a diagram:



You are right that decreasing the duty cycle is the key to long lasting and bright LEDs. Plus, as heat increases the LEDs will color shift as much as 10nm (which is visible). You can further decrease the duty cycles of the LEDs by modifying the driver board with logic for determining when you have to even turn on each LED array based on the input image. This is alot more complicated than I wanted for a first pass. Therefor my initial plan is to integrate a white LED array and try as much as possible to control the heat. If I need to I can create a simple timing board for pulsing the LED's rather than having a constant on state. This should decrease the heat alot. This is far more doable than creating an entire driver architecture to control the three sets of arrays in optimal form.

The next step would be to develop three RGB arrays with a simple driver architecture (your example above) with 33% duty cycle on each array. This is a pretty simple timing board, maybe just a modification of the board from the first step, but still requires me to reverse engineer some the electronics on the projector.

The last step (if I ever get there) is to create an optimized driver architecture for shortening duty cycle and increasing LED life.

Remy

Add another person on the list looking forward to seeing the results of this experiment. If it works out this could spawn another branch of the DIY projector effort. Besides... led powered projectors are just plain cool anyway...

I have been in an engineering bubble for the past couple of weeks studying the patents and technology used by several different vendors for LED light source TV's. I have gained a lot of information and am making progress in my own design, but it's taking time.

In researching some of this technology, I have found some information that indicates that the color wheel bleeds a lot of light from a DLP engine. I have only found one measured reference to it and it is sketchy at best. Does anyone else know of any research that details where the all of the light loss in a DLP light engine goes?

If not, I will find the answer eventually, even if it's through my own experimentation.

Thanks,
Drak

Hi,

If it is a single chip (color wheeled) DLP then you are automatically giving up on 2/3'ds of your white LED's total output. This is because the white LED is emmiting equal parts of Red Green and Blue light to make white. At any moment the color wheel will be filtering out 2 of those color components to leave just the one component color of interest.

If it is a 3 chip DLP then you can get full output in white at the screen.

However, it is fairly simple to deduce that you cannot have the full light output if you are projecting a pure Red Green or Blue picture with iether single chip or 3 chip DLP's. Same for any color for that matter. Fortunately our eyes are used to this as the sun suffers the same problem . And no worries since projector Lumens are measured as the total white level anyway.

In terms of projecting a single color component (Red Green or Blue) with a 3 chip VS a single chip DLP, the 3 chip will still be 3 times brighter. This is for the same reason as projecting white: the color wheel will only be displaying red for 1/3'd of the time as the rotation of the filter wheel is constant and the red area only represents 1/3'd of its area. A 3 chip DLP (no colorwheel) projects that component constantly.

Of course, there are numerous other sources of light loss, but you have potentially nailed down 2 huge factors by having a. an efficiently collimated light source (assuming collimation is pretty good) and b. no polarizers in the system.

In terms of making a brightness prediction, I would try and research what the lumens are for that unit's replacement bulb.

Mark

What would happen if you took out the color wheel all together on a 3 Chip Dlp, and just used Red, Blue, and Green LEDs ?

There is no color wheel on a 3 chip DLP.

On a single chip DLP, removing the color wheel and using a contant white light will give you a flickery innacurate monochrome projection. If you could have the LED's trigger just Red Green or Blue at the exact moment the color wheel would have been filtering to that color then you would have a proper full color image again. But getting that down would not be easy. The only benefit to just having a color wheel and all three on at the same time is electrical power savings.

Mark

First of all, please accept my admiration for your efforts to use the dead projector lamp-projector to come up with LED lighting modification. I read through all the posts and noticed here that you are concerned about sychronizing the Individual colorLEDs (If you use that approach) with the DLP. I would like to propose a simple approach to synchronize the LED's:

1. If you were to move the color wheel out of the optical path, you should be able to use it as an encoder by putting stripes on it for the three colors and triggering from your encoder marks. this is assuming that they don't already have some form of encoding for their circuit. Each time you trigger at the beginning of a color, you can fire an adjustable monostable whose timing you adjust with a pot. You therefore have three such pots that you adjust the duty cycle for each color for best color rendition.
(Each pot controls the contribution of each color duration). This way you have not made any mods to their circuit. The system still acts as if the color wheel is still in place.

2. It is further possible that the color wheel already has encoder signals going to their circuit to tell the DLP which color filter is in place. In that case you won't need to add the encoder marks and sensor.

Good Luck

mpa

I too would love to offer encouragement in this pursuit!

One note: I think you will end up spending more then the anticipated $50 on electronics, circuit boards, heatsinks, etc. Luxeon LEDs are notoriosly high current LEDs, and the versions you have picked out are 1.5A LEDs to get that brightness. Building a constant current power supply for all those LEDs is going to be tricky.

I don't know what type of arrangement you plan to put the LEDs in, but I imagine with that tight a lens you are going to get hotspots.

But anyway, good luck! Look forward to the pics! I was going to do Luxeons in a 15", but I never got started on that projector. So, whenever my eVo arrives, that is now going to become my test bed

I would persue the RGB LED route if I were you. Wether you replace the Colour filters or just shoot through them you will end up with much better quality light than if you used White LEd's. Filtering White LED's for RGB ends up with crap distribution and a Lot of lost lumens. Combining them wont be easy. I suppose you could make up a Colour wheel of LEDs that flash on when in the correct position (Mechanical wiper?) This would give you the PWM and the timming.

Keep us posted.

Kirk

A Red Green Blue LED color wheel.

you can glue the LED's or sum thing to the color wheel and the LED's can be lit with
Brushis all the time or whin needid.

Novalux seems like a better choice to me.

hello, I thought I had registered on lumenlabs before. Anywhozel, I'm interested in this LED projector expirement as well and subscribing to this thread.

to bad this topic died anyone done an interesting LED experiment since ?

Fry.

Yes, I am very interested as well. Replacing a UHP with an LED array would rock in terms of dollar savings on bulb replacements. I have a burned out UHP housing that I recently replaced in my Samsung DLP rear projection downstairs. It is basically a very reflective parabolic with a thin quartz filament element positioned in the center on thin wires.

I don't know how many lumens a brand new UHP puts out, but I would think it would be possible to duplicate that column of light with LEDs. I have often thought about popping that old UHP open, slipping in some kind of light element in place of the now broken quartz filament, then seeing how it performs. The UHP overcomes the shadow problems of normal parabolic reflection by having a transparent (almost) filament vertically hung at the focal point, rather than probing through the back of the reflector.

I've been doing some over here...
Rgb Leds As A Light Source? A Spectrum Advantage?!?
The thread is very long, I started on post #963.

I'd say the results so far are promising, but not incredibly amazing.

Wow long topic.. looks interesting ! As far as my own research goes so far (only into all this DIY projector building for 2 days haha.. straight tho i'm an addict now), I havent found a competing LED solution yet compared to bulbs. But that conclusion is based on a not to solid way of comparing delivered lumens by LED's and bulbs. Comparing the both of them all stands and falls with calculating the exact lumens that go through the screen.. I think that's where DIY projector research currently lacks. Either we must find out practical (building more -different- LED solution and trying them out) or describing the theories more precise, taking all the laws of light into consideration. As I'm personally interested in trying to use LED as backlighting I'm currently collecting all the info I can find on LED's and plan on writing a FAQ.

Fry.

hello there i have also been looking into this. i want to use a camara lense a fresnel lens and a ps1 screen in a projector. being run on small 12 volt batteries i will use leds for most light per watt. lexeon are not the way to go if you have some money spare go for these. http://www.dealextreme.com/details.dx/sku.2394 i have used these in a torch that i have nik names "pocket sun"
but if money is tighter you could use these http://www.dealextreme.com/details.dx/sku.4282 an array of 6 or these would give aproxamitaly 1200 lumens and only use 2 amps at 12 volts. i might make an array of 9 for a 3 amp bulb with 1800 lumen even this more expensive option will ony cost about 40 dollars
do you guys reckon that this setup could project a half good picture in day light if i restrickted the screen size to maybe 30 inches?