The first element is a dichroic mirror that reflects green light, but passes blue light. (Made with a thin-film coating of a particular thickness.) The second element is another dichroic mirror that reflects red light but passes blue and green light. These mirrors are not very expensive, especially considering how small they must be.
The third and fourth are probably a pair of selective polarity mirrors to get all the polarization pointing in a single direction. (A beam retarder.) Then the fourth element would be two prisms glued together with another selective polarity mirror in between them. That would reflect all the incoming light, but then pass all the light bouncing off the DLP.
The lenses all just colimate the rays into a nice even converging cone that illuminates the entire surface of the DLP, and then comes to a point in the center of the projection lens. That gets all the available light through the projection lens, while also minimizing the required lens diameter and aberrations.
I bet you could make a projector almost as good without most of that stuff: Just place the three LEDs as close together as possible right next to the projection lens. Then bounce the light off the DLP and into a big projection lens like one of those 5" diameter 18" fl opaque projector lenses. There would be a slight difference in the size of the opposite edges of the projected image, but you could compensate for that by changing the projector-to-screen angle.
The third and fourth are probably a pair of selective polarity mirrors to get all the polarization pointing in a single direction. (A beam retarder.) Then the fourth element would be two prisms glued together with another selective polarity mirror in between them. That would reflect all the incoming light, but then pass all the light bouncing off the DLP.
The lenses all just colimate the rays into a nice even converging cone that illuminates the entire surface of the DLP, and then comes to a point in the center of the projection lens. That gets all the available light through the projection lens, while also minimizing the required lens diameter and aberrations.
I bet you could make a projector almost as good without most of that stuff: Just place the three LEDs as close together as possible right next to the projection lens. Then bounce the light off the DLP and into a big projection lens like one of those 5" diameter 18" fl opaque projector lenses. There would be a slight difference in the size of the opposite edges of the projected image, but you could compensate for that by changing the projector-to-screen angle.
Guy Grotke:
I bet you could make a projector almost as good without most of that stuff: Just place the three LEDs as close together as possible right next to the projection lens. Then bounce the light off the DLP and into a big projection lens like one of those 5" diameter 18" fl opaque projector lenses. There would be a slight difference in the size of the opposite edges of the projected image, but you could compensate for that by changing the projector-to-screen angle.
I actually tried to mix 3 x 3Watt Green Led + 3 x 3 Watt Red Led + 1 x 3 Watt Blue Led by placing them as close together as possible in a cycle (the single Blue Led in the middle of the cycle). Then placed a piece of light diffusing sheet (which is from the lcd back light unit) in front of the Led array. The power consumption of the led array was about 10 Watt, because I lighted all the leds at half or lower than half of its typical power. The led array was glued to the surface of a CPU cooler. The picture was not bad especially the color was very good, the brightness was not very high but it was ok when watching in the dark. The details of the very dark area of the projected picture was not lost.
I bet you could make a projector almost as good without most of that stuff: Just place the three LEDs as close together as possible right next to the projection lens. Then bounce the light off the DLP and into a big projection lens like one of those 5" diameter 18" fl opaque projector lenses. There would be a slight difference in the size of the opposite edges of the projected image, but you could compensate for that by changing the projector-to-screen angle.
I actually tried to mix 3 x 3Watt Green Led + 3 x 3 Watt Red Led + 1 x 3 Watt Blue Led by placing them as close together as possible in a cycle (the single Blue Led in the middle of the cycle). Then placed a piece of light diffusing sheet (which is from the lcd back light unit) in front of the Led array. The power consumption of the led array was about 10 Watt, because I lighted all the leds at half or lower than half of its typical power. The led array was glued to the surface of a CPU cooler. The picture was not bad especially the color was very good, the brightness was not very high but it was ok when watching in the dark. The details of the very dark area of the projected picture was not lost.
You can use semi-mirrored surfaces to do that, but they tend to lose a lot of the "passed" light to reflection off the back side. I suppose you could do something with mirrors that reflect light polarized in one direction and pass light polarized in the other, but you would need one such mirror and a beam retarder for each additional LED: Crazy expensive!
You might also be able to take advantage of "critical angle", to make a prism-shaped glass object that passes light from air to glass, but light hitting that surface from the inside would be totally reflected back into the glass. But you would still need one of these beam combiners for each additional LED.
A much cheaper way to do this would be to build a hollow spherical array of LEDs, with them all pointing at a negative lens (ie. a plano-convex) of the right strength to refract all those converging rays into a parallel beam. You could build almost any size array you wanted, without needing a mirror & lens for each LED.
You will still have a problem with "conservation of etendue", which means that you can never make a wide light source (or an array of LEDs) into a perfectly parallel beam. Because of this, the rays at each pixel of the DLP will come from N different angles (where N is the number of LEDs). As long as most of the rays get through the projection lens, you should be ok, since the projection lens will focus all the rays from each DLP pixel to a single screen pixel.
If anybody wants to try this, check out www.surplusshed.com. They have lots of different simple lenses and even big ones (ie. 4" diameter) go for under $6 each! Get anti-reflective coated lenses if you can.
You might also be able to take advantage of "critical angle", to make a prism-shaped glass object that passes light from air to glass, but light hitting that surface from the inside would be totally reflected back into the glass. But you would still need one of these beam combiners for each additional LED.
A much cheaper way to do this would be to build a hollow spherical array of LEDs, with them all pointing at a negative lens (ie. a plano-convex) of the right strength to refract all those converging rays into a parallel beam. You could build almost any size array you wanted, without needing a mirror & lens for each LED.
You will still have a problem with "conservation of etendue", which means that you can never make a wide light source (or an array of LEDs) into a perfectly parallel beam. Because of this, the rays at each pixel of the DLP will come from N different angles (where N is the number of LEDs). As long as most of the rays get through the projection lens, you should be ok, since the projection lens will focus all the rays from each DLP pixel to a single screen pixel.
If anybody wants to try this, check out www.surplusshed.com. They have lots of different simple lenses and even big ones (ie. 4" diameter) go for under $6 each! Get anti-reflective coated lenses if you can.
Why not? It is pretty much just a rod of high refractive index glass, with each end ground and polished.
You could just get some thick glass rod from a lab supply house and experiment with it. Normal lab glass might work fine for your application. If you want total internal reflection, then most of the glass should be surrounded by air. If you have it in contact with something of higher refractive index (IE. glass or water) then much of the light would escape out the sides.
You can also buy "light pipe" from surplus or scientific supply houses. I think Edmund Scientific may have it. (Don't buy "image pipe": That would not homogenize the intensity.)
BTW: This is the function of those funny sheets of plastic you have to remove from the back of your LCD to make a projector. They spread out the light from just one or two fluorescent tubes along the edge, and make a uniform backlight.
You could just get some thick glass rod from a lab supply house and experiment with it. Normal lab glass might work fine for your application. If you want total internal reflection, then most of the glass should be surrounded by air. If you have it in contact with something of higher refractive index (IE. glass or water) then much of the light would escape out the sides.
You can also buy "light pipe" from surplus or scientific supply houses. I think Edmund Scientific may have it. (Don't buy "image pipe": That would not homogenize the intensity.)
BTW: This is the function of those funny sheets of plastic you have to remove from the back of your LCD to make a projector. They spread out the light from just one or two fluorescent tubes along the edge, and make a uniform backlight.
Possibly. It depends on the spectral characteristics of the "white" LEDs and the color wheel. If the LEDs have narrow blue and yellow bands, then they will not work in an RGB color wheel system. But I think there are now white LEDs with broader spectra.
But the whole issue is moot: There is little reason to use white LEDs and a color wheel. The color wheel would always be blocking 2/3 of your light bandwidth. Why not just use red, green, and blue LEDs and pulse them sequencially to simulate a color wheel?
A color wheel DLP system only makes sense if you are using a single white light source like a MH lamp.
But the whole issue is moot: There is little reason to use white LEDs and a color wheel. The color wheel would always be blocking 2/3 of your light bandwidth. Why not just use red, green, and blue LEDs and pulse them sequencially to simulate a color wheel?
A color wheel DLP system only makes sense if you are using a single white light source like a MH lamp.
In general, yes: That is the "conservation of etendue" problem I mentioned a few posts back. You can read about etendue on wikipedia (very technical) or any number of other web sites.
If your light source is a single point, then you can reflect or refract light from it into a perfectly parallel beam of any diameter you like. If the source is larger, then you can't. And conservation of etendue means you can't make it "look" like a point source.
But for some projection applications it doesn't matter: If you really had a point-source light, then you would not need a projection lens since every LCD pixel or DLP pixel would result in a single ray that would illuminate a single screen pixel!
Since nobody really has a point-source lamp, we use a lens to refract all of each LCD pixel or DLP pixel's rays to a single screen pixel.
If your light source is a single point, then you can reflect or refract light from it into a perfectly parallel beam of any diameter you like. If the source is larger, then you can't. And conservation of etendue means you can't make it "look" like a point source.
But for some projection applications it doesn't matter: If you really had a point-source light, then you would not need a projection lens since every LCD pixel or DLP pixel would result in a single ray that would illuminate a single screen pixel!
Since nobody really has a point-source lamp, we use a lens to refract all of each LCD pixel or DLP pixel's rays to a single screen pixel.
This might be possible right now (although probably not practical).
RGB LED > fiber optic > collimator > DLP chip
The LED can provide the exact color needed for each pixel of the image.
post 250
The diagram of the LED DLP RPTV suggests that it would be possible to superimpose 2 sources like post 250 suggests. We could have a similar beam coming from the lens side and polarized in the other plane, and the two beams would be superimposed were the DLP chip is. The polarization and prism passing should lose less than half of each beam.
Yet, where does this break the etendue conservation?
The same for those focusing devices.
If we placed a point source behind an LCD wouldn't difraction at each LCD pixel be an issue? The same difraction can help create one first.
I wonder if colour LCDs with 50% light transmission are available by now, like this 1998 article describes: http://www.research.ibm.com/journal/rd/423/tanase.html
Now that we know how pipelight backlights are, won't our Chinese friends mold us some with the back dimples arranged so that we have a light cone already, so we could skip the fresnels and bulky bulbs? A PC monitor with two modes: regular computer use and projection mode.
The diagram of the LED DLP RPTV suggests that it would be possible to superimpose 2 sources like post 250 suggests. We could have a similar beam coming from the lens side and polarized in the other plane, and the two beams would be superimposed were the DLP chip is. The polarization and prism passing should lose less than half of each beam.
Yet, where does this break the etendue conservation?
The same for those focusing devices.
If we placed a point source behind an LCD wouldn't difraction at each LCD pixel be an issue? The same difraction can help create one first.
I wonder if colour LCDs with 50% light transmission are available by now, like this 1998 article describes: http://www.research.ibm.com/journal/rd/423/tanase.html
Now that we know how pipelight backlights are, won't our Chinese friends mold us some with the back dimples arranged so that we have a light cone already, so we could skip the fresnels and bulky bulbs? A PC monitor with two modes: regular computer use and projection mode.
Re: post 250
Probably the maxium light transmission a tri-color LCD panel can reach is 33%, even if the polarizing light pipe is used in the backlight.
This company is selling 10 Watt, 30 Watt, 50 Watt, 100 Watt LED. http://www.jnsaver.com/LED/led1.htm
A sunboy is testing with the 50 Watt LED and the result is not bad.
http://diy2005.aa.topzj.com/viewthread.php?tid=373241&extra=page=1
zzonbi said:
Probably the maxium light transmission a tri-color LCD panel can reach is 33%, even if the polarizing light pipe is used in the backlight.
This company is selling 10 Watt, 30 Watt, 50 Watt, 100 Watt LED. http://www.jnsaver.com/LED/led1.htm
A sunboy is testing with the 50 Watt LED and the result is not bad.
http://diy2005.aa.topzj.com/viewthread.php?tid=373241&extra=page=1
That is true: With the three color filters, the most you could get through a color LCD is 33%. But the most you can get reflected off a DLP also has the same kind of limitation (in time instead of space).
If you really want to get most of your light on the screen, then I would go with a system that uses dichroic mirrors to split white MH light into red, green, and blue beams. Then use a beam retarder to rotate each beam to a single polarity. Then send each beam through its own mono-LCD. Then recombine the beams with another set of dichroic mirrors before it goes through the projection lens.
Of course, that is a description of most commercial LCD-based projectors!
There are opportunities for DIY along those lines: If we could find a cheap source for small fast high-res mono LCDs, then we could build "triple gun" projectors. These would work like three seperate projectors (ie. red, green, or blue MH lamp + LCD + fresnels & projection len), and the three color images would be converged on the screen. The electronics would be very easy using a VGA interface, since the three colors are already seperated.
If you really want to get most of your light on the screen, then I would go with a system that uses dichroic mirrors to split white MH light into red, green, and blue beams. Then use a beam retarder to rotate each beam to a single polarity. Then send each beam through its own mono-LCD. Then recombine the beams with another set of dichroic mirrors before it goes through the projection lens.
Of course, that is a description of most commercial LCD-based projectors!
There are opportunities for DIY along those lines: If we could find a cheap source for small fast high-res mono LCDs, then we could build "triple gun" projectors. These would work like three seperate projectors (ie. red, green, or blue MH lamp + LCD + fresnels & projection len), and the three color images would be converged on the screen. The electronics would be very easy using a VGA interface, since the three colors are already seperated.
Guy Grotke said:
It is true that the most light flux you can get reflected off a DLP also has the same kind of limitation but in time, but the human eyes probably do not percept that short light-off time interval. However, I am not sure the human eyes percept the continuous light and the flashing light same or not, if the intensity of the two types of light source are same. We probably need to study the visual physiology of the rods and cones in retina.
I was of course talking about LCDs with dichroic filters/mirrors. The absorbtion losses pose indeed a severe 1/3 limit. Some LCD TVs use a fourth white pixel for this reason, just like some DLP colour wheels.
Guy was probably talking about the losses in colour wheels, which happen sequentially, in time. The end result is the same as for the LCD "parallel" colour filters. A DLP has less disadvantage with 1/3 pulsing leds though, while a 3DLP should be superior to 3LCD.
I am not worried about using 3 LCDs now, I'd be content to interface a single small LCD (the PSP one, sold for cheap separately). There are plenty other such TFTs in various MP4 players or playstations, but they come without a VGA interface more likely than not. Some have AV IN lines, but I suspect they're only for recording and not live monitoring.
I read something about conductive ink to directly trace fine circuits with inkjet printers.
Those leds are hot, but get us some good small panels first.
Guy was probably talking about the losses in colour wheels, which happen sequentially, in time. The end result is the same as for the LCD "parallel" colour filters. A DLP has less disadvantage with 1/3 pulsing leds though, while a 3DLP should be superior to 3LCD.
I am not worried about using 3 LCDs now, I'd be content to interface a single small LCD (the PSP one, sold for cheap separately). There are plenty other such TFTs in various MP4 players or playstations, but they come without a VGA interface more likely than not. Some have AV IN lines, but I suspect they're only for recording and not live monitoring.
I read something about conductive ink to directly trace fine circuits with inkjet printers.
Those leds are hot, but get us some good small panels first.
Human Visual Response
That is already well known: The human color visual mechanism integrates the light level over a time a bit slower than 1/16th of a second. (This is why movies work!)
If you show somebody a certain light level at 100% duty cycle, they perceive it the same as 3 times that level at 33% on / 67% off. (As long as the overall frequency is over 15 Hz.) There is no advantage to gain by pulsing the light. On the other hand, human visual perception is not linear and the nervous system does a huge amount of "pre-processing" and extrapolation (EG. we do not perceive objects moving around when we move our heads).
The only reason to pulse the light source is so you can use the same light modulator time-sequencially with the three different colors. You could do that with a DLP or with a mono-LCD. And you can make the sequences with different color LEDs or a white source with a spinning color filter wheel.
That is already well known: The human color visual mechanism integrates the light level over a time a bit slower than 1/16th of a second. (This is why movies work!)
If you show somebody a certain light level at 100% duty cycle, they perceive it the same as 3 times that level at 33% on / 67% off. (As long as the overall frequency is over 15 Hz.) There is no advantage to gain by pulsing the light. On the other hand, human visual perception is not linear and the nervous system does a huge amount of "pre-processing" and extrapolation (EG. we do not perceive objects moving around when we move our heads).
The only reason to pulse the light source is so you can use the same light modulator time-sequencially with the three different colors. You could do that with a DLP or with a mono-LCD. And you can make the sequences with different color LEDs or a white source with a spinning color filter wheel.
flicker
I perceive very easily 60Hz flicker on my old CRT monitor. Its phosphorus has rather low persistence, but 75Hz is OK. 50Hz TV is OK. 50Hz neon lamps weren't. Seeing fluid motion on the other hand is sufficed by much lower frequencies, hence the cinema only needs 24 frames each second. Yet some use antiflicker techniques, like showing each frame twice.
A common trick with leds is to pulse them to make them look brighter to the eye. The pulse can use higher currents than continuous regime. Thus a DLP LED is a little helped vs a classic lamp and colour wheel.
Many discharge lamps use AC too, if not most.
I perceive very easily 60Hz flicker on my old CRT monitor. Its phosphorus has rather low persistence, but 75Hz is OK. 50Hz TV is OK. 50Hz neon lamps weren't. Seeing fluid motion on the other hand is sufficed by much lower frequencies, hence the cinema only needs 24 frames each second. Yet some use antiflicker techniques, like showing each frame twice.
A common trick with leds is to pulse them to make them look brighter to the eye. The pulse can use higher currents than continuous regime. Thus a DLP LED is a little helped vs a classic lamp and colour wheel.
Many discharge lamps use AC too, if not most.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.