Newbie doing a DIY homebrew HDTV projector, need advice please!

First post here. I am designing and building a homebuilt HDTV projector on the http://www.lumenlab.com/forums/ ... An simple example schematic follows:

http://www.distantvisions.net/images/projector/proj-diagram.png

I have done all the research I need to build one of these. The only thing I need to do is to come up with a light source. Since I am trying to make it as compact as possible, I am interested in using LED lighting due to the lower heat output and directional nature of the light they throw.

Typically, I would throw a 400W metal halide light in the box and use a few fans to cool it down. I have always been interested in LED lighting, and thought that this would be a great hobby project for me, LEDs and all.

The LEDs are my stumbling block though, as I am not too knowledgeable about them. I only really know that they are cooler burning, longer lasting, use less electricity, and are more space efficient. Sounds ideal. Except I have no idea where to start.

I think that I should just get this made for me, and tell people about how and who made it. I am interested in LED lighting, but the learning curve is kinda steep... I have played with a breadboard in the past, but this exceeds my technical abilities.

So my questions are as follows:

Can this array be made for me for less than $400.00US?

Who do you recommend making it for me?

What kind of LEDS should I use?

Here is the specs for the LCD panel I will be using:
344.5mm max(W) x 222.5mm max (H) x 6.5maxmm (D)

Aparently, from what I have read, an LCD panel will only pass light through it at a limited angle (6-9 degrees). However, the ultimate goal is to get the most possible lumens through the triplet lens as is possible and onto the wall in as even a manner possible. Also, the desired light would be about 6500K 90CRI.

You guys (and gals) seem to know a heck of a lot about this stuff, so I thought you might have some fun with this one. Any help is appreciated!

Cheers, Jerald!

Hello Jerald,

Firstly :welcome:, you sound like you have an extremely interesting project under way. However, I don't think you will achieve your aim of using LEDs for the projector. Modern LEDs can be a little more efficient than a metal halide lamp but they can't compete when it comes to output. An example (Philips) 400w lamp I just looked at turns out almost 30000 lumens on average. Some of the upcoming multi-die LEDs are putting around 800 each so you would need quite a large array of them to produce usable light. Such light would also be from a large area rather than the 'point-source' of the metal halide lamp (making it difficult to focus the light coming from the LEDs).

When it comes to sheer output LEDs can't match the discharge lamps and whilst they are smaller I don't think you would have saved any space by the time you have a large enough array of them and the heatsinking/cooling infrastructure to match.

Good luck with the projector though.

Andrew

Thanks for your reply. Actually, I was interested in using this type of array not as a point light source, but as an almost direct contact light source. I wanted to make an LED array that was placed as close as possible to the LCD panel. My design differs from the standard single point light source design, and I was hoping to do away with the first fresnel lens as having the LEDs that close to the screen would be more efficient.

Partly I am considering this because:

I don't think it has been done on the Lumenlab forum in the last couple of years and LEDs have made a lot of progress. They would be interested in this if it was possible.

I am interested in LED arrays and have played with breadboards and making circuits in the past.

I think I can replace the 400W lamp with a much lower wattage lamp.

I can cool this array with less fans and :. less noise.

I am trying for sure to one up major projector manufacturers.

I am convinced that if designed and executed correctly, it can be done.

I am a curious person, and would go through with this if the cost wasn't prohibitive.


With all that in mind, who do you recommend making it for me? What kind of LEDS should I use?

Thanks for the reply!

Cheers, Jerald!

PS, by that stat of 30,000 lumens, I would need about an array of 40 LEDs. I was thinking that a custom array would be able to fit quite a few more than that in there based on my desired array size of 344.5mm W) x 222.5mm (H) x 6.5mm(D)... Not sure though... thats why i'm here!

There are likely a lot of issues to consider here.

1) LEDS have a different spectra compared to Metal Halide. If you use most white LEDS, you may find that they are very lacking in red output and hence the colors that you get out are washed out. Though both the LED and MH may have the same color temp, the MH has a richer output across the whole spectrum. You may be able to use warm\neutral whites, but not sure.

2) When you put the LEDs close to the LCD, you will need to ensure that your LEDS are very consistent in terms of light output and color. At a minimum, you will need to ensure they are all from the same bin... binned for color and light output.

3) You WILL NOT one up the projector manufacturers. To get any sort of good light output, you need to concentrate your wide panel down to a small spot....pretty much need the fresnel for that. With just a projection lens, say a 7" lens imaging your screen, you would lose already 70-80% of your light since you have a big panel. Maybe if you got a 9" lens you could do better, but still would lose a lot. Whole host of issues w.r.t. backscatter, internal reflections, etc. that make getting a high level contrast very difficult. The best projector use lots of tricks including irises that change with the scenes to enhance contrast ratios well beyond what panels can do.

That said, it will be a fun project. When you look at high end LCD panels that are LED backlit, they actually use RGB LEDS, not white. They are not as efficient currently as white, but it is not just about efficiency, but color range. They are able to display a much much wider color gamut with very rich colors.

Semiman

Hey nobody has tried doing a DIY projector with a phlatlight evaluation kit yet. If you can get them to send you one it would really be an interesting first. Just to be clear, this is an LED source specifically designed for projectors.

Here is where to look. Best part for you is that they'll send out a complete solution with electronics and thermal management. So just be sure to come up with some sort of corporate identity, fake if necessary, in order to get them to sell it to you.

When the light passes through the panel, any evidence of its point nature is destroyed.

So I think it can be done. I think I would look to have about 8k-10k lumens.

Some MC-Es (12) in an array would do it. I'd mount them to something from here.

Driving them will be a little bit of a challenge. Someone else can chime in on how to do that efficiently from AC line voltage.

When the light passes through the panel, any evidence of its point nature is destroyed.

Click to expand...

This is true of an LCD television, but not an LCD projector. If the panel diffused the light then how could the light be projected onto the wall as a coherent image?

snarfer said:

This is true of an LCD television, but not an LCD projector. If the panel diffused the light then how could the light be projected onto the wall as a coherent image?

Click to expand...

A converging lens creates a real image from any real object at or beyond its focal length.

If the panel did not diffuse the light, then the only image the projector would project would be the small circle representing the portion of the panel between the light source and the lens.

An infinitely small point source at the FL is only necessary to focus the lens at infinity, i.e. to create a parallel series of light rays for a light beam. A projector focuses at a finite distance and from an object of finite size (the LCD panel).

Thanks for all the replies! Since this has gotten some attention, and you all seem to understand what I am after, I will put some more detailed information here about the projector itself.

First, the standard box:

http://diyprojectorkits.com/welcome.../product/3f2b04572fc97982832b340cb93611f4.jpg

http://diyprojectorkits.com/welcome.../product/3f2b04572fc97982832b340cb93611f4.jpg

http://www.hacknmod.com/pics/497-1.jpg


I want to mod this design by chopping off the back 6 to 8 inches and instead of using point lighting, using multipoint lighting. I was hoping to have an LED panel that was the same size as the LCD panel I was going to use. Here is a very rough schematic of the panel I am proposing:

http://img222.imageshack.us/my.php?image=diyledboardfk5.png

I'm not sure how realistic that is as far as a design goes, taking into account thermal management in between the LEDs, but that is the gist of what I am trying to do (i actually designed that one myself, a bit crude but shows what i am after)...

I plan on using this lens:

http://diyprojectorkits.com/welcome...category_id=9&option=com_virtuemart&Itemid=85

as well as a standard fresnel lens that is the same size as the lcd panel.

I looked at the Phlatlight, and it is used in actual projectors, and wouldn't be suitable for a DIY project if I understood what I am looking at.

LukeA, that would be a great site for finding the right heatsink or heatsinks...

Once again, thanks for your support!

Cheers, Jerald!

A converging lens creates a real image from any real object at or beyond its focal length.

If the panel did not diffuse the light, then the only image the projector would project would be the small circle representing the portion of the panel between the light source and the lens.

An infinitely small point source at the FL is only necessary to focus the lens at infinity, i.e. to create a parallel series of light rays for a light beam. A projector focuses at a finite distance and from an object of finite size (the LCD panel).

Click to expand...

Maybe we are trying to say the same thing in a different way. All I'm trying to point out is that the eventual contrast ratio and brightness of the projected image will be proportional to the degree to which the light hitting the panel is collimated. This is why it is extremely difficult to use an LED array for this application - hard to collimate light from multiple sources.

The reason why the phlatlight was developed was in order to provide a high brightness LED light source that was also highly collimated, specifically for projection applications. As far as I know there is no other highly collimated LED light source commercially available, and no other LED light source used in professional projection applications. But I've been wrong before and anticipate being wrong again in the future!

May I suggest that you refer to this extremely thorough 14 page long thread on lumenlab for some more specific information on exactly how to go about this project from people who have actually executed it. Strangely enough it appears that the most satisfactory results were obtained not with an array of LEDs at all, but with a single P7! But please, don't take my word for it. Read the whole thread.

snarfer said:

Maybe we are trying to say the same thing in a different way. All I'm trying to point out is that the eventual contrast ratio and brightness of the projected image will be proportional to the degree to which the light hitting the panel is collimated. This is why it is extremely difficult to use an LED array for this application - hard to collimate light from multiple sources.

The reason why the phlatlight was developed was in order to provide a high brightness LED light source that was also highly collimated, specifically for projection applications. As far as I know there is no other highly collimated LED light source commercially available, and no other LED light source used in professional projection applications. But I've been wrong before and anticipate being wrong again in the future!

May I suggest that you refer to this extremely thorough 14 page long thread on lumenlab for some more specific information on exactly how to go about this project from people who have actually executed it. Strangely enough it appears that the most satisfactory results were obtained not with an array of LEDs at all, but with a single P7! But please, don't take my word for it. Read the whole thread.

Click to expand...

If we were trying to collimate the light, then, yes, a point source would be ideal. But we are not trying to collimate light for this project. We are focusing the light at a finite distance, not at infinity. No matter where all the light comes from, the image that the projection lens projects is that of the LCD panel. An evenly lit panel will project well. An unevenly lit panel will not. A poorly lit panel will not project well.

It seemed to me that the lumenlab guys were having trouble with their LEDs due to a lack of flux. Those crappy multi-5mm junction LEDs just don't put out enough light. The P7 worked better because it was their brightest solution.

Try this experiment: Take a flashlight and put a piece of paper over its bezel. Write something on the paper. Then take a large lens and focus the paper's image on the ceiling. It focuses sharply. Note that the effective light source is the entire circle of the reflector, not just the LED die. For large-panel projection, you want a brightly, evenly lit panel. It doesn't matter how you get that.

Well LEDs aren't remarkably more efficient than metal halide. It's kind of a toss-up, you have to take LED temp, ballast/driver losses and unused light into account. It could easily be less efficient.

On the nature of the panel: the LCD panel does not diffuse light. The first fresnel aims to focus the light into a plane of parallel light the same size as the panel and second fresnel. It is not diffused when transmitting through the panel in the way it would going through glass shower doors.

As such I don't think the plan to make a planar "backlight" of LEDs will work, because they're emitting light in all directions. This creates a very poor capture ratio on the lens.

As was said earlier, the ability of the lens to give a clear focus on the light depends largely on how much of a point the light source is. Unfortunately an array of LEDs doesn't act optimally for that purpose.

And if you read through the lumenlab thread you will find that the single P7 performed better than a much higher powered array, precisely because of the effect Oznog has pointed out.

As was said earlier, the ability of the lens to give a clear focus on the light depends largely on how much of a point the light source is. Unfortunately an array of LEDs doesn't act optimally for that purpose.

Click to expand...

Also, a point source is useful because it can be collimated. So actually a planar source radiating collimated light, such as the phlatlight, will also be optimal. But, again, an array of LEDs will neither produce collimated light, nor act as a point source, so it is not going to be very useful in this application.

This is not just theory but result of actual experimentation. People on lumenlab forum actually built these things.

There is a solution though, and that is not to use a few extremely powerful LEDS passing through a lens, but about 190 5mm LEDS to form a planar panel of forward emitting light in extremely close proximity to the LCD panel (less than a 10mm gap between the LEDS and LCD). Also, just to point out, that the thread that is referred to here is not a "DIY Projector", but rather a mod of a commercial projector. That is a TOTALLY different animal. I am designing not to do a mod for a commercial machine with limited space constraints, but to simply shine a light through a 15.4" WUXGA LCD panel as intensely as possible. Uniformity is a bonus, but having the light pass through the condenser fresnel lens into the triplet lens should make the light more uniform. Not that big of an issue though, as I will have both the collimator and condenser lenses on hand to experiment with.

The higher powered array used in that projector mod is using different principles, mainly that it is trying to pass the light through multiple lenses before it even hits the lcd in the projector. My array will probably be lensless and as close to the lcd panel as I can have it without melting the LCD panel. It should look like this, but with a single color led and higher density:


or like this, but once again, higher density:



or like my reference design image:

http://img222.imageshack.us/my.php?image=diyledboardfk5.png


I'd be willing to make a mistake as long as it didn't set me back too much money. You guys have the knowledge to help me minimize my error, and potentially produce some kickass results, which I would be proud to show you.

Things I have liked from your responses include:

"When you put the LEDs close to the LCD, you will need to ensure that your LEDS are very consistent in terms of light output and color. At a minimum, you will need to ensure they are all from the same bin... binned for color and light output." SemiMan

"Some MC-Es (12) in an array would do it. I'd mount them to something from here..." LukeA

"we are not trying to collimate light for this project. We are focusing the light at a finite distance, not at infinity. No matter where all the light comes from, the image that the projection lens projects is that of the LCD panel. An evenly lit panel will project well." LukeA

and Oznogs post makes some good points, but seems based on the idea that there will be a first fresnel lens to collimate the light. I'll have one on hand and will play with it, but think 190 points of 9 - 12 degree light would make the uneven distribution less of an issue.


So back to my original questions:

Can this array be made for me for less than $400.00US?

Who do you recommend making it for me?

What kind of LEDS should I use?

Like I said, the worst case scenario is that I end up blowing some money on a hobby project that produces mediocre results. I can use it as a light for my driveway if this doesn't work (another craftworking project), and throw in a 400w MH light.

Cheers, and thanks for your interest and advice, Jerald!

Have you seen Lamina's RGB Titan? They are bright and small plus I think they have 10 degree optics for them..

http://www.laminalighting.com/ProductsandSolutions/ProductsandSolutions/TitanSeries/Titan.aspx

They are $80 a piece though...

http://www.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=521-1045-ND

They also have single colors which are ridiculously bright...
Red: http://parts.digikey.com/1/parts/1381846-light-engine-titan-red-nt-51a0-0468.html

Green: http://parts.digikey.ca/1/1/1207894-light-engine-titan-green-nt-52c1-0470.html

Blue: http://parts.digikey.com/1/parts/1381848-light-engine-titan-blue-nt-52b1-0469.html

There is another company called Enfis that makes led arrays that might be even brighter.

http://www.enfis.com/products/index.htm

I don't have any experience with those..

There is a solution though, and that is not to use a few extremely powerful LEDS passing through a lens, but about 190 5mm LEDS to form a planar panel of forward emitting light in extremely close proximity to the LCD panel (less than a 10mm gap between the LEDS and LCD). Also, just to point out, that the thread that is referred to here is not a "DIY Projector", but rather a mod of a commercial projector. That is a TOTALLY different animal. I am designing not to do a mod for a commercial machine with limited space constraints, but to simply shine a light through a 15.4" WUXGA LCD panel as intensely as possible. Uniformity is a bonus, but having the light pass through the condenser fresnel lens into the triplet lens should make the light more uniform. Not that big of an issue though, as I will have both the collimator and condenser lenses on hand to experiment with.

Click to expand...

Do some ray tracing and you will see why this will produce poor results. DIY projectors work on the same principle as commercial projectors. The only light that will end up in the triplet lens at the proper angle to produce an image will be the tiny percentage of the light from the LEDs that travels directly forward. All the rest of the light will just be contaminating the image. Also your image will probably not have uniform brightness. Good luck.

I referred vaguely to this thread in an earlier post:

http://www.lumenlab.com/forums/index.php?showtopic=9136

This has been done before guys. In 2005. So here we are, 2008, LEDs have come along nicely, and I am asking for advice in the right place. I don't think that the guy that did this had the same idea as me, and ask for expert help.

Here is what we know:

Collimating the light is a challenge, if you actually decide to try to collimate it. A solution to this is probably needed, but the 2005 project got decent results without doing this. Possibilities include using lenses, shining the light through the LCD and hoping for the best, using small narrow angle LEDS, using large maximum power leds and than collimating them, using a diffuser before the collimator lens, and whatever else you guys can think of.

Not using a collimating fresnel works, because the guy that did the panel in 2005 did this and it worked. Efficiency is questionable. I think we can do better. Possibilities include using a different type of collimating lens that would be more compatible with multisource light.

We have MUCH better, brighter LEDS today than we did back than.

So here are my requirements:

Highest possible lumens hitting the LCD possible using LED lights. I am assuming this can be done using either a lot of small leds and creating a high density of lighting points, or using higher intensity, larger lights and some sort of higher efficiency lens to collimate the light.

SO let's go through the fundamentals.

1. Lots of smaller points of uniform light or fewer ultrabright ones?
2. Suggestions for lenses if it is determined they are needed?
3. Who would you get to fab this if it is feasable (and there is no doubt in my mind it is, as it has been done before).

Sound off here! Post up some recommendations for LEDs! Who would you have make this thing?

Cheers, Jerald!

PS, can you recommend a ray tracing program for me? Open source or freeware would be best...

I am still unclear on your lens setup.

I believe you wish to take a clear 15.4" WXGA panel and put a 15.4" backlight of perhaps thousands of T1-3/4 5mm LEDs behind it. How far behind it?

Where would the lenses be placed, and what size?

The lens setup is something I will experiment with, as changing lenses is not as hard to do as changing LED arrays is.

You are correct about my intentions Oznog. I wish to place this 15.4" LED array of possibly thousands of LEDs behind a 15.4" WUXGA (HDTV quality) LCD screen. I have actually calculated that 3mm LEDs deliver more raw lumens per square centimeter than the larger 5mm or 10mm LEDs.

I was thinking of placing the LEDs as close to the WUXGA panel as possible to maximize light intensity. Probably less than an inch.

The lenses are typically the same size as the panel to be used, so both 15.4 inch, with a 450mm triplet lens at the focal point of the light to project it onto the wall.

Finally, the setup (light path) would be:
LED array------->1st Fresnel Lens to collimate the light *if needed* (probably in direct contact with the LCD panel)------------->LCD Panel----------->2nd Fresnel Lens to collect the light to a point --------> Triplet lens to project the light------------->Wall

I am NOT interested in using a single point of light. The whole reason I am wanting to do this is because it is a challenge that has not been completely solved. Doing this right would be a huge step forward for the LumenLab forum. Your help is massively appreciated.

Cheers, Jerald!

PS:
http://s384.photobucket.com/albums/oo284/jeraldjunkmail/?action=view&current=TIANsLEDPanel.jpg

There it is, but we can do this better/right...

I think that you might have problems using epoxy-covered small-junction LEDs due to a lack of red in the emission spectrum and lack of heatsinking.

Also, 3, 5, and 10mm LEDs aren't as efficient as the newest high-flux LEDs. But in use, the delivered lumens/watt in an even pattern may be higher than say, MC-Es or XR-Es.

I think that you will still want to run a fan over the LEDs.