Hi again,
Is there a fresnel type lens which would do a fairly good job of collimating the light from an LED array when each LED (on the array) has a rather wide output angle of say 110 degrees?
What adverse effects might happen if the light source was not collimated for the correct LCD / DMD chip etendue? I'm still unclear of exactly what etendue is, and what problems it causes. I understand it has something to do with the "amount" of collimation you need compared to the area of the LCD panel or DMD chip in the projector?
I can understand that the smaller arcs of MH lamps are easier to collimate compared to a larger LED array, but why is it seemingly so difficult to collimate a small LED array with a simple lens? Does the display etendue put strict limits on the size of the light source, or is it possible to correct for this with lenses?
How would this apply to DLP projectors, because the light from the lamp is generally focused to a very small area on the colour wheel..... So, is it not possible to simply focus an LED array output onto a small area using a just simple lens?
Sorry for all the questions, but most of you guys are far more experienced with the optics stuff. I think I need to do some serious research on lens basics!
Is there a fresnel type lens which would do a fairly good job of collimating the light from an LED array when each LED (on the array) has a rather wide output angle of say 110 degrees?
What adverse effects might happen if the light source was not collimated for the correct LCD / DMD chip etendue? I'm still unclear of exactly what etendue is, and what problems it causes. I understand it has something to do with the "amount" of collimation you need compared to the area of the LCD panel or DMD chip in the projector?
I can understand that the smaller arcs of MH lamps are easier to collimate compared to a larger LED array, but why is it seemingly so difficult to collimate a small LED array with a simple lens? Does the display etendue put strict limits on the size of the light source, or is it possible to correct for this with lenses?
How would this apply to DLP projectors, because the light from the lamp is generally focused to a very small area on the colour wheel..... So, is it not possible to simply focus an LED array output onto a small area using a just simple lens?
Sorry for all the questions, but most of you guys are far more experienced with the optics stuff. I think I need to do some serious research on lens basics!
etendue
Etendue is the killer for LED arrays, and the laws of physics say you can't overcome it! With a point-source lamp like a short-arc MH, the condensor fresnel refracts most of the rays into a parallel beam. Then the field fresnel converges those rays into an arc image near the center of the projection lens. That gets most of the light through the lens and on the screen.
With an array of LEDs, even with cone spreads of only 10 degrees, there is no way to get all of those rays parallel. So there is no way to converge them into a small circle to get through a standard projection lens. You can do the math: Just find sin(5 degrees) * 2 * (distance from LEDs to lens). Each LED will make a circle of light with that diameter around the center of the lens.
This means that you will have to treat an LED array like it was a diffuse light source, ie. use a giant lens like a CRT projector to gather most of the light through each pixel. That is fine, as long as your LCD is smaller that the lens diameter. That limits you to combinations like an 8" Hami LCD with a 8" or 9" CRT projector lens. (There is no lens that would work with a 12" or larger LCD.) That also limits you to LED arrays that can light a small LCD.
If you tried to use a standard projection lens, it would miss most of the light through pixels that did not line up directly centered with an LED. So you would be able to see the LED array pattern in the screen image!
MH projectors, even with arc lengths of 25 mm, can converge an arc image into a standard projection lens. The arc image length is a product of arc length * (LCD-to-lens)/(arc-to-LCD), so a projector using a 25 mm arc lamp, 220 mm fl condensor fresnel, and 330 mm field fresnel will have an arc image length of about 37.5 mm. Very easy to find a good lens that can pass a 37.5 mm circle!
As for the efficiency of MH reflectors, that varies widely. With a non-reflector design, you can calculate the amount of light that will strike the condensor fresnel. It is the lamp output * fresnel area / sphere at that distance surface area. Pretty bad! Adding a spherical reflector will improve it only by about 20-30% (instead of the theoretical 100%) because that type of reflector sends all the reflected rays back through the fairly opaque arc chamber.
A good optical-quality silver or aluminum elliptical reflector with the output cone matched to a condensor fresnel can be about 50% efficient. (Most rays do not travel back through the arc chamber.) With a little spherical reflector in front of the arc, you can direct quite a lot of the otherwise unusable rays back into the elliptical reflector and get the efficiency even higher. 250 Watt and 400 Watt MH lamps put out about 20000 and 33000 lumens respectively, so with the right reflector you could get 10000-17000 lumens hitting the LCD with all rays on a useful path. That is a lot of light!
Etendue is the killer for LED arrays, and the laws of physics say you can't overcome it! With a point-source lamp like a short-arc MH, the condensor fresnel refracts most of the rays into a parallel beam. Then the field fresnel converges those rays into an arc image near the center of the projection lens. That gets most of the light through the lens and on the screen.
With an array of LEDs, even with cone spreads of only 10 degrees, there is no way to get all of those rays parallel. So there is no way to converge them into a small circle to get through a standard projection lens. You can do the math: Just find sin(5 degrees) * 2 * (distance from LEDs to lens). Each LED will make a circle of light with that diameter around the center of the lens.
This means that you will have to treat an LED array like it was a diffuse light source, ie. use a giant lens like a CRT projector to gather most of the light through each pixel. That is fine, as long as your LCD is smaller that the lens diameter. That limits you to combinations like an 8" Hami LCD with a 8" or 9" CRT projector lens. (There is no lens that would work with a 12" or larger LCD.) That also limits you to LED arrays that can light a small LCD.
If you tried to use a standard projection lens, it would miss most of the light through pixels that did not line up directly centered with an LED. So you would be able to see the LED array pattern in the screen image!
MH projectors, even with arc lengths of 25 mm, can converge an arc image into a standard projection lens. The arc image length is a product of arc length * (LCD-to-lens)/(arc-to-LCD), so a projector using a 25 mm arc lamp, 220 mm fl condensor fresnel, and 330 mm field fresnel will have an arc image length of about 37.5 mm. Very easy to find a good lens that can pass a 37.5 mm circle!
As for the efficiency of MH reflectors, that varies widely. With a non-reflector design, you can calculate the amount of light that will strike the condensor fresnel. It is the lamp output * fresnel area / sphere at that distance surface area. Pretty bad! Adding a spherical reflector will improve it only by about 20-30% (instead of the theoretical 100%) because that type of reflector sends all the reflected rays back through the fairly opaque arc chamber.
A good optical-quality silver or aluminum elliptical reflector with the output cone matched to a condensor fresnel can be about 50% efficient. (Most rays do not travel back through the arc chamber.) With a little spherical reflector in front of the arc, you can direct quite a lot of the otherwise unusable rays back into the elliptical reflector and get the efficiency even higher. 250 Watt and 400 Watt MH lamps put out about 20000 and 33000 lumens respectively, so with the right reflector you could get 10000-17000 lumens hitting the LCD with all rays on a useful path. That is a lot of light!
etendue explanation
Etendue is the "uncertainty of origin" that is built in to any light source. A true point-source would have zero etendue, even if you used reflectors and/or lenses to make a diverging cone, a parallel beam, or a converging cone. You could do all three and still converge it all back to a single point.
A real light source, like a MH lamp arc for example, has some size. If the original lamp arc has an oval shape, then you can use any combination of reflectors and lenses to make cones or beams, but you can't make the source look like a point from the point of view of anywhere in the optical system. It will always look like an oval. (No lens or reflector is perfect, so more optical elements means more distortion.) So even if you use lenses to try to colimate the beam, it will always spread to the same percentage as the original source size has to the first optical element: "The etendue is always conserved."
You can get an introductory optics book (or just learn online) so you can make some ray tracing drawings. If you follow the rays from an LED to an LCD or DLP, and then through the projection lens you will see the problem. It isn't enough to just illuminate the LCD or DLP: The rays through each pixel have to be going in a direction that will end up on the screen, or they are wasted.
Etendue is the "uncertainty of origin" that is built in to any light source. A true point-source would have zero etendue, even if you used reflectors and/or lenses to make a diverging cone, a parallel beam, or a converging cone. You could do all three and still converge it all back to a single point.
A real light source, like a MH lamp arc for example, has some size. If the original lamp arc has an oval shape, then you can use any combination of reflectors and lenses to make cones or beams, but you can't make the source look like a point from the point of view of anywhere in the optical system. It will always look like an oval. (No lens or reflector is perfect, so more optical elements means more distortion.) So even if you use lenses to try to colimate the beam, it will always spread to the same percentage as the original source size has to the first optical element: "The etendue is always conserved."
You can get an introductory optics book (or just learn online) so you can make some ray tracing drawings. If you follow the rays from an LED to an LCD or DLP, and then through the projection lens you will see the problem. It isn't enough to just illuminate the LCD or DLP: The rays through each pixel have to be going in a direction that will end up on the screen, or they are wasted.
LED array idea
If I was going to attempt an LED projector, I think I would build an array of white (RGB white, not blue with yellow phosphor) LEDs laid out on a hexagonal pattern. Then I would put an acrylic hexagonal lens above each LED, with the LED-to-lens distance equal to the lens focal length. This would let me use LEDs with a normal illumination cone instead of special narrow beam LEDs. Where each lens intersected the cone of the underlying LED, the light would be refracted into a fairly parallel beam. This would be like using 10 mm diameter hexagonal LEDs with 1 degree cones. The lens edges would all touch the adjacent lens edges, so it would create a mostly uniform parallel beam the size of the LCD.
On the far side of the LCD, I would use a field fresnel to converge the parallel rays into a small hexagon to get them through a standard projection lens. Like maybe an 18" fl opaque projector lens.
I would use the distance between the LED array and the LCD to adjust the picture uniformity versus the light lost by missing the projection lens. With the array close to the LCD, all the light should get through to the screen, but there might be some artifacts of the LED pattern on the screen. By increasing the array-to-LCD distance, I could smooth out the illumination around the hexagonal lens areas, but it would also make some of the light miss the projection lens.
If I was going to attempt an LED projector, I think I would build an array of white (RGB white, not blue with yellow phosphor) LEDs laid out on a hexagonal pattern. Then I would put an acrylic hexagonal lens above each LED, with the LED-to-lens distance equal to the lens focal length. This would let me use LEDs with a normal illumination cone instead of special narrow beam LEDs. Where each lens intersected the cone of the underlying LED, the light would be refracted into a fairly parallel beam. This would be like using 10 mm diameter hexagonal LEDs with 1 degree cones. The lens edges would all touch the adjacent lens edges, so it would create a mostly uniform parallel beam the size of the LCD.
On the far side of the LCD, I would use a field fresnel to converge the parallel rays into a small hexagon to get them through a standard projection lens. Like maybe an 18" fl opaque projector lens.
I would use the distance between the LED array and the LCD to adjust the picture uniformity versus the light lost by missing the projection lens. With the array close to the LCD, all the light should get through to the screen, but there might be some artifacts of the LED pattern on the screen. By increasing the array-to-LCD distance, I could smooth out the illumination around the hexagonal lens areas, but it would also make some of the light miss the projection lens.
Nice information about the LEDs and lenses, it's very helpful. A lot of people have also suggested that you'd need to collimate each LED in an array to get a decent output.
I'm still a bit lost on why the shape of the original light source puts limits on how well the LCD or DMD is "lit"? Is it not possible to use "beam shaping optics" to change the beam to a more rounded (or rectangle) output? There are many light guides available now which appear to shape the beam very effectively. There was an image of the Philips Ostar with a nice light guide somewhere, but I can't find it now.
I've attached a photo of an 80mW diode laser which uses beam-shaping prisms to produce a more rounded output beam. I know this would be expensive for commercial projectors, but would something like this work (in theory) for improving on the etendue match?
When there is an etendue mismatch, does this just cause the light collection to be less efficient, or does it mess up the uniformity of the displayed image?
I have to say, some of the white LEDs I've tried with my projector don't seem too bad at all with regards to the colour accuracy of the output. The picture does have a slightly blue tint to it compared to the MH lamp, but an RGB array would of course solve many of these problems. The main hurdle for us seems to be just collimating the light effectively?
I'm still a bit lost on why the shape of the original light source puts limits on how well the LCD or DMD is "lit"? Is it not possible to use "beam shaping optics" to change the beam to a more rounded (or rectangle) output? There are many light guides available now which appear to shape the beam very effectively. There was an image of the Philips Ostar with a nice light guide somewhere, but I can't find it now.
I've attached a photo of an 80mW diode laser which uses beam-shaping prisms to produce a more rounded output beam. I know this would be expensive for commercial projectors, but would something like this work (in theory) for improving on the etendue match?
When there is an etendue mismatch, does this just cause the light collection to be less efficient, or does it mess up the uniformity of the displayed image?
I have to say, some of the white LEDs I've tried with my projector don't seem too bad at all with regards to the colour accuracy of the output. The picture does have a slightly blue tint to it compared to the MH lamp, but an RGB array would of course solve many of these problems. The main hurdle for us seems to be just collimating the light effectively?
Attachments
etendue or the non point factor
It is not the shape offending, but the fact that the source isn't a point. Points do not exist in reality. They are just a geometry concept.
Etendue of a light beam is its section area times its solid angle. Solid angle roughly averages the ordinary angle between the most diverging component rays along the section periphery. So it isn't constant as the light beam travels, but it cannot decrease vs its value in the origin "point", the source.
I guess a single led has smaller etendue than discharge lamps. Too bad one isn't enough. But there's a catch: we can control each led aiming individually, unlike the ordinary emitting sources in classical theory. I would try use an uncut lens for glasses, the largest dioptry power and diameter you can get. Then cluster the leds you want to use so that their focused image on the other side of the lens is about the size of the dlp chip, and in plane. Aim each to not miss the lens. To combat light uniformity problems use the old defocus trick (at the glass lens, not projection objective). With the jumbos you should be easily able to collect light from 50 leds.
So, you have two projection systems: one projects the cluster to the dlp surface, the other the dlp to the wall.
The etendues need not match. It's just that if you get too high one from the source than the max of the rest of the system, you'll lose some light. DLPs have extra danger: the off mirror light ending to the wall as well (the flip mirror angle is small, like 8-12 degrees). This translates into a rather small needed etendue.
It is not the shape offending, but the fact that the source isn't a point. Points do not exist in reality. They are just a geometry concept.
Etendue of a light beam is its section area times its solid angle. Solid angle roughly averages the ordinary angle between the most diverging component rays along the section periphery. So it isn't constant as the light beam travels, but it cannot decrease vs its value in the origin "point", the source.
I guess a single led has smaller etendue than discharge lamps. Too bad one isn't enough. But there's a catch: we can control each led aiming individually, unlike the ordinary emitting sources in classical theory. I would try use an uncut lens for glasses, the largest dioptry power and diameter you can get. Then cluster the leds you want to use so that their focused image on the other side of the lens is about the size of the dlp chip, and in plane. Aim each to not miss the lens. To combat light uniformity problems use the old defocus trick (at the glass lens, not projection objective). With the jumbos you should be easily able to collect light from 50 leds.
So, you have two projection systems: one projects the cluster to the dlp surface, the other the dlp to the wall.
The etendues need not match. It's just that if you get too high one from the source than the max of the rest of the system, you'll lose some light. DLPs have extra danger: the off mirror light ending to the wall as well (the flip mirror angle is small, like 8-12 degrees). This translates into a rather small needed etendue.
light channeling
Hey, ozone, how's your dlp attempt going?
I wanted to let you know I spotted some new leds on e-bay. They are quotted directly in lumens (13), not candelas. Considering the much higher current they need vs the jumbos, I guess those have really only 2 lumens at 20mA. These look good because they're even more compact, 8mm.
Some Lithuanian guy on the lumen lab forums cut the leds edges to pack them in a more compact cluster. Looks like it worked great. The light has better uniformity. I was thinking of something that may help your needs: couldn't you couple each led to an optical fiber, then tighten all the fibers in a rope with the section of the dlp chip, and beam it directly over it? The coupling is done routinely with the use of small plastic lenses, for communication equipment. The fibers have extremely low light losses.
http://cgi.ebay.ca/POWERED-LEDs-20P...2260125QQihZ012QQcategoryZ66952QQcmdZViewItem
On second thought, for this to work you'd need leds which can flash different colours, controlled. They exist.
Hey, ozone, how's your dlp attempt going?
I wanted to let you know I spotted some new leds on e-bay. They are quotted directly in lumens (13), not candelas. Considering the much higher current they need vs the jumbos, I guess those have really only 2 lumens at 20mA. These look good because they're even more compact, 8mm.
Some Lithuanian guy on the lumen lab forums cut the leds edges to pack them in a more compact cluster. Looks like it worked great. The light has better uniformity. I was thinking of something that may help your needs: couldn't you couple each led to an optical fiber, then tighten all the fibers in a rope with the section of the dlp chip, and beam it directly over it? The coupling is done routinely with the use of small plastic lenses, for communication equipment. The fibers have extremely low light losses.
http://cgi.ebay.ca/POWERED-LEDs-20P...2260125QQihZ012QQcategoryZ66952QQcmdZViewItem
On second thought, for this to work you'd need leds which can flash different colours, controlled. They exist.
Hi, zzonbi!
Those new eBay LEDs do look quite interesting. They should be much better than the jumbos, as the new ones have a narrower beam output (55 degrees). The rated lumens seems to be much better too.
However, although our experiments with the jumbos were quite promising, I think we'd still need a far brighter LED solution to make the image watchable.
Your fibre-optics idea is good! I used to work for a telecoms company aligning optical multiplexors and splicing fibres etc., so this is quite an interesting idea. The use of a fibre bundle would also need optics at the output end because fibres still don't generally collimate the beam much (if at all).
I can't give away too much about the CSB project at the moment, except to say that it now has a few more "added features" which are quite unique, and hopefully many people will be very interested in it once we've worked everything out.
Of course, we're also still experimenting with the LED "lamp" stuff! 😀
OzOnE.
Those new eBay LEDs do look quite interesting. They should be much better than the jumbos, as the new ones have a narrower beam output (55 degrees). The rated lumens seems to be much better too.
However, although our experiments with the jumbos were quite promising, I think we'd still need a far brighter LED solution to make the image watchable.
Your fibre-optics idea is good! I used to work for a telecoms company aligning optical multiplexors and splicing fibres etc., so this is quite an interesting idea. The use of a fibre bundle would also need optics at the output end because fibres still don't generally collimate the beam much (if at all).
I can't give away too much about the CSB project at the moment, except to say that it now has a few more "added features" which are quite unique, and hopefully many people will be very interested in it once we've worked everything out.
Of course, we're also still experimenting with the LED "lamp" stuff! 😀
OzOnE.
http://www.cree.com/press/press_detail.asp?i=1143496856921
What about those things?
And... did I read it correctly?
57 lumens of typical light output?
From an led?
What about those things?
And... did I read it correctly?
57 lumens of typical light output?
From an led?
There are leds putting out more than 57lm (check lumileds), but 47lm/W is probably the best for a commercial product. The usual values are around 25.
So where can we buy those, and at what prices?
So where can we buy those, and at what prices?
Umm...
I dunno where to get them...
I was just thinking someone was getting off at 15 lm... and 57 sounded a whole lot better...
So I posted the link.
I'm kind of scattered right now...
I'm really interested in getting a portable outdoor speaker system going...
I don't even remember how I got to this thread...
I've been browsing the diyaudio forums (mainly referencing the t-amp) for about a week... and didn't even know this was a part of it till I registered...
I dunno where to get them...
I was just thinking someone was getting off at 15 lm... and 57 sounded a whole lot better...
So I posted the link.
I'm kind of scattered right now...
I'm really interested in getting a portable outdoor speaker system going...
I don't even remember how I got to this thread...
I've been browsing the diyaudio forums (mainly referencing the t-amp) for about a week... and didn't even know this was a part of it till I registered...
CREE 7090 suppliers
I see them at:
http://www.cutter.com.au/products.php?cat=12
for a base price of about $10 AU ($7.58 US) for white. If you want colored ones (ie. for simulating a DLP color wheel by switching them on and off), the reds are cheaper and the greens and blues are more expensive. They have a little price calculator for quantity price breaks. They also have the collimator lenses, which would be essential for building a projector. Those are about $9 AU each! Ouch!
Here is a US supplier (of many):
http://ledlightingsupply.com/ledlightingsupply/products.htm
Here is one in the EU:
http://www.vs-optoelectronic.com/eng/produkte/cree/index.php
These are all just from googling [CREE 7090 LED price]. There are lots more to check out.
I see them at:
http://www.cutter.com.au/products.php?cat=12
for a base price of about $10 AU ($7.58 US) for white. If you want colored ones (ie. for simulating a DLP color wheel by switching them on and off), the reds are cheaper and the greens and blues are more expensive. They have a little price calculator for quantity price breaks. They also have the collimator lenses, which would be essential for building a projector. Those are about $9 AU each! Ouch!
Here is a US supplier (of many):
http://ledlightingsupply.com/ledlightingsupply/products.htm
Here is one in the EU:
http://www.vs-optoelectronic.com/eng/produkte/cree/index.php
These are all just from googling [CREE 7090 LED price]. There are lots more to check out.
Only the first links to an actual shop.
The efficient ones (XR7090) are 9.1$ in small quantities. So besides 47lm/W we have 6.3lm/$.
The less efficient ones are cheaper per lumen though.
They even have cool hexagonal lenses. The 7 cell zoom cluster is interesting, as it varies the beam collimation from 12 to 90 degrees.
The efficient ones (XR7090) are 9.1$ in small quantities. So besides 47lm/W we have 6.3lm/$.
The less efficient ones are cheaper per lumen though.
They even have cool hexagonal lenses. The 7 cell zoom cluster is interesting, as it varies the beam collimation from 12 to 90 degrees.
FYI
Check these sites out. LED DLP's coming soon. They have made an LED light engine with 1500 lumens! Seems interesting.
http://www.nuvision.com/ledlp/
http://www.luminus.com/phlatlight/
http://www.dlp.com/home_entertainment/led_hdtvs.asp?bhcp=1
Check these sites out. LED DLP's coming soon. They have made an LED light engine with 1500 lumens! Seems interesting.
http://www.nuvision.com/ledlp/
http://www.luminus.com/phlatlight/
http://www.dlp.com/home_entertainment/led_hdtvs.asp?bhcp=1
Check out Boxlight Bumblebee model: DLP led front projector, up to 150lm, palm size, 19W, 39dB, 800$.
Perhaps the end of the short lived bulb commercial units.
Still too expensive to not consider diy led projectors.
Until now they only had 25lm.
Still unable to find a decent small tft, 2"-4".
Perhaps the end of the short lived bulb commercial units.
Still too expensive to not consider diy led projectors.
Until now they only had 25lm.
Still unable to find a decent small tft, 2"-4".
There is a video of these LCD premade kits from China under
ambient light conditions, not bad. They are using 8.9" HDTV display at least that's I understand from the context..
http://diy2005.aa.topzj.com/viewthread.php?tid=180260&extra=page=1&page=7
http://tv.mofile.com/cn/xplayer.swf?v=LU15U4O1
ambient light conditions, not bad. They are using 8.9" HDTV display at least that's I understand from the context..
http://diy2005.aa.topzj.com/viewthread.php?tid=180260&extra=page=1&page=7
http://tv.mofile.com/cn/xplayer.swf?v=LU15U4O1
An externally hosted image should be here but it was not working when we last tested it.
This is the chinese setup detail, linked above..
btw. the french guys claim 3400lumens at light source with 900 x
(10mm 130.000mcd angle 12°) LEDs ? The previous less powerfull version of his setup had above 60lumes at the screen..
http://forum.aspect-geek.net/aspect...0w-LED-Allinbox-projet-HDLED-sujet-8342-7.htm
- Status
- Not open for further replies.