Posted on the AVS forum, then I was directed here:
The previous DIY projectors have focused primarily on cost - what can you build for <$500 or so. What if that number was higher? Say around $3000?
Another assumption is the relaxation of some constraints. Since I'll probably do rear projection, size, weight and 'beauty' of the projector have little bearing.
Here's my design goals:
1) At least 1920x1080 60p
2) No screen door at <1x screen width viewing (it's all about angle of view subtended for me)
3) Bright, at least 2x a typical movie theater (that's >110 cd/m from a post I've seen here)
4) Excellent CR (this is the really tough one)
Here's the idea:
Display Surface:
I have a laptop with a 13"x8.25" LCD which is 1920x1200 native resolution. At 11" viewing distance and the brightest white background it'll generate, I cannot see any pixel structure. 23" 1920x1200 LCDs with 500:1 CR and 16 ms response times are $1700 and dropping.
Light Source:
Conventional wisdom would take a bright lamp w/ reflector and fresnel lens to illuminate the panel. The non-insurmountable problems with this approach are:
1) Heat
2) Bulb longevity
3) Complexity with alignment
4) Difficulty generating uniform illumination over the entire panel
5) Contrast ratio can only be 500:1 (what you get from the panel)
LEDs however, run cool, are lower power, and last 100,000 hours or more. Additionally, their light output does not decay as rapidly as HID lights.
High-intensity LEDs can now be purchased for <$0.40. The ones I have are 15000 mcd, 3.3 volt with a 15* spread. Using a standard proto board (0.10" center holes), I'm making a couple small arrays for experimentation. A 25 LED grid will fit in a 1.21 in^2. Removing the lip permits 49 LEDs in the same space. With the 25 LEDs/1.21 in^2, that's 2216 LEDs behind the laptop display, or 33.2k candela. At 8% LCD transmission, that's 2659 candela out of a 0.0692 m^2 screen. Blown up to my target 5.1 m^2 screen, that's 521 cd/m^2. I can live with that. Even if the optics drop 50% that's still 2x my target illumination.
As for power, all 2216 LEDs will consume 146 watts (20 mA each). If we derate the LED lifetime by 50%, I could leave the projector on constantly for 5.7 years. That's $0.018/hr bulb cost.
Problems with this solution:
1) Building the LED grid is a pain.
2) How close to D6500 are the LEDs?
3) $900 initial cost for the light, BUT it'll outlast at least 10 regular projector bulbs
4) Fine-grain uniformity over the panel (that's what I'm experimenting with now)
5) Contrast ratio can only be 500:1 - OR CAN IT?
I saw an interesting monitor at SIGGRAPH this year. It's a high dynamic range display. They claim 30x the brightness and a 40,000:1 contrast ratio by using an LED array behind a normal LCD. It was very impressive to see in action. It could produce images that were painful to look at (too bright) and black was BLACK (no light). They filter the video signal to obtain a luminance value for each LED and modulate the LED's output accordingly. Here's the link: http://www.sunnybrooktech.com/ Their HDR paper at SID 2003 under publications has the best info.
Electronics would be a bear for a DIY'er to control 2216 LEDs (tap the panel link receiver's digital RGB output at 200 MHz, filter and drive the LEDs with PWM using a bunch of FPGAs?), but a 1920x1200 display which is >500 cd/m^2 at 5 m^2 and basically room-limited cr is quite a draw.
I'm educating myself on the display optics. Looks like a fresnel lens right after the LCD then a few glass lenses for magnification is the right way to go.
I have a laptop (15" screen, 1600x1200, dead backlight, practically free), a hundred 15000 mcd LEDs and a couple cheap fresnel lenses I'm currently experimenting with as proof of concept. <$100 invested so far. It's quite a learning exercise!
I have a lot of related ideas (flat field uniformity solution, huge color gamut solution, etc.), but this post is long enough already...
The previous DIY projectors have focused primarily on cost - what can you build for <$500 or so. What if that number was higher? Say around $3000?
Another assumption is the relaxation of some constraints. Since I'll probably do rear projection, size, weight and 'beauty' of the projector have little bearing.
Here's my design goals:
1) At least 1920x1080 60p
2) No screen door at <1x screen width viewing (it's all about angle of view subtended for me)
3) Bright, at least 2x a typical movie theater (that's >110 cd/m from a post I've seen here)
4) Excellent CR (this is the really tough one)
Here's the idea:
Display Surface:
I have a laptop with a 13"x8.25" LCD which is 1920x1200 native resolution. At 11" viewing distance and the brightest white background it'll generate, I cannot see any pixel structure. 23" 1920x1200 LCDs with 500:1 CR and 16 ms response times are $1700 and dropping.
Light Source:
Conventional wisdom would take a bright lamp w/ reflector and fresnel lens to illuminate the panel. The non-insurmountable problems with this approach are:
1) Heat
2) Bulb longevity
3) Complexity with alignment
4) Difficulty generating uniform illumination over the entire panel
5) Contrast ratio can only be 500:1 (what you get from the panel)
LEDs however, run cool, are lower power, and last 100,000 hours or more. Additionally, their light output does not decay as rapidly as HID lights.
High-intensity LEDs can now be purchased for <$0.40. The ones I have are 15000 mcd, 3.3 volt with a 15* spread. Using a standard proto board (0.10" center holes), I'm making a couple small arrays for experimentation. A 25 LED grid will fit in a 1.21 in^2. Removing the lip permits 49 LEDs in the same space. With the 25 LEDs/1.21 in^2, that's 2216 LEDs behind the laptop display, or 33.2k candela. At 8% LCD transmission, that's 2659 candela out of a 0.0692 m^2 screen. Blown up to my target 5.1 m^2 screen, that's 521 cd/m^2. I can live with that. Even if the optics drop 50% that's still 2x my target illumination.
As for power, all 2216 LEDs will consume 146 watts (20 mA each). If we derate the LED lifetime by 50%, I could leave the projector on constantly for 5.7 years. That's $0.018/hr bulb cost.
Problems with this solution:
1) Building the LED grid is a pain.
2) How close to D6500 are the LEDs?
3) $900 initial cost for the light, BUT it'll outlast at least 10 regular projector bulbs
4) Fine-grain uniformity over the panel (that's what I'm experimenting with now)
5) Contrast ratio can only be 500:1 - OR CAN IT?
I saw an interesting monitor at SIGGRAPH this year. It's a high dynamic range display. They claim 30x the brightness and a 40,000:1 contrast ratio by using an LED array behind a normal LCD. It was very impressive to see in action. It could produce images that were painful to look at (too bright) and black was BLACK (no light). They filter the video signal to obtain a luminance value for each LED and modulate the LED's output accordingly. Here's the link: http://www.sunnybrooktech.com/ Their HDR paper at SID 2003 under publications has the best info.
Electronics would be a bear for a DIY'er to control 2216 LEDs (tap the panel link receiver's digital RGB output at 200 MHz, filter and drive the LEDs with PWM using a bunch of FPGAs?), but a 1920x1200 display which is >500 cd/m^2 at 5 m^2 and basically room-limited cr is quite a draw.
I'm educating myself on the display optics. Looks like a fresnel lens right after the LCD then a few glass lenses for magnification is the right way to go.
I have a laptop (15" screen, 1600x1200, dead backlight, practically free), a hundred 15000 mcd LEDs and a couple cheap fresnel lenses I'm currently experimenting with as proof of concept. <$100 invested so far. It's quite a learning exercise!
I have a lot of related ideas (flat field uniformity solution, huge color gamut solution, etc.), but this post is long enough already...
Hi MrHDTV,
Your thread title is "The ultimate DIY projector"? Would a 3-chip + prism DLP using the latest TI 3HD prototype chips and higer power white LEDs be a "more ultimate" projector design than LCD? From what I recently read, TI will soon yield 1920 x 1080 DLPs for next year volume production.
I have been looking at 50" LCD flat screen panels, and just started reading about projection and learning about improvements in non-organic dyes for long life LCDs, and DLP nano technology improvements with 3-chip DLPs. Nirvana seems within our reach.
Your thread title is "The ultimate DIY projector"? Would a 3-chip + prism DLP using the latest TI 3HD prototype chips and higer power white LEDs be a "more ultimate" projector design than LCD? From what I recently read, TI will soon yield 1920 x 1080 DLPs for next year volume production.
I have been looking at 50" LCD flat screen panels, and just started reading about projection and learning about improvements in non-organic dyes for long life LCDs, and DLP nano technology improvements with 3-chip DLPs. Nirvana seems within our reach.
The key to this approach is 2-fold:
1) The ability to modulate the light source for individual pixel regions to achieve CRT-like contrast ratios (>10k:1, which is really room limited)
2) Zero screen door at <1x screen width viewing distance
Plus 15" or 23" LCD panels are a bit more obtainable then a trio of 1920x1080 DMD devices for the one-off DIY'er.
<$3000 for a bright, zero screen door 1920x1200 projector with a piddily 500:1 CR. If I can figure out a way to drive the LEDs with PWM in a reasonable fashion, then >10k:1 CR to boot. The's the holy grail for me.
1) The ability to modulate the light source for individual pixel regions to achieve CRT-like contrast ratios (>10k:1, which is really room limited)
2) Zero screen door at <1x screen width viewing distance
Plus 15" or 23" LCD panels are a bit more obtainable then a trio of 1920x1080 DMD devices for the one-off DIY'er.
<$3000 for a bright, zero screen door 1920x1200 projector with a piddily 500:1 CR. If I can figure out a way to drive the LEDs with PWM in a reasonable fashion, then >10k:1 CR to boot. The's the holy grail for me.
There are some mistakes in your math. You can not convert directly from Candela's out of the LED to candelas off your screen. 15,000 mCd at 15 degrees is about 0.8 lumens. 2200 of them gives you about 1,760 lumens. You used 8% transmission which gives you 140 lumens out of the LCD. With optical losses, you will likely get down to about 100 lumens.
Your output will be lumens / (area*sr)
- sr is the spherical angles of "reflection" from the screen. If you have a perfect 180 degree reflection, it would be pi steradians. However, will will give you the benefit of some gain and say it is 1/2 pi steradians. Area if 5.1 meters.
That gives you about 12 candela/m^2.
Hence you are going to need 10 times more LEDS to get what you want. I think you will find Luxeons much more efficient in terms of light output than 5mm LEDS.
You had some other questions on the LEDS. I think you will find if you look at the LED spec sheet that the brightness can vary on the order of 2x, and the color temperature can vary from 5000K to 15,000K.
Alvaius
Your output will be lumens / (area*sr)
- sr is the spherical angles of "reflection" from the screen. If you have a perfect 180 degree reflection, it would be pi steradians. However, will will give you the benefit of some gain and say it is 1/2 pi steradians. Area if 5.1 meters.
That gives you about 12 candela/m^2.
Hence you are going to need 10 times more LEDS to get what you want. I think you will find Luxeons much more efficient in terms of light output than 5mm LEDS.
You had some other questions on the LEDS. I think you will find if you look at the LED spec sheet that the brightness can vary on the order of 2x, and the color temperature can vary from 5000K to 15,000K.
Alvaius
I have been thinking about the "Ultimate" DIY projector design, and this is what I've come up with:
Lamp: 1200W or 2500W HMI film light. These produce 110,000 and 240,000 lumens respectively. The great thing is that the arc lengths are just 1cm and 1.4 cm respectively, so a reflector/condenser could conceivably capture 50% or more of the light produced. These lamps are relatively cheap ($100 on eBay for the 1200W) and last 750 to 1000 hours. The ballasts are expensive, but DIY comes to the rescue.
DIYLabs 15" Fresnel and standard or 135mm triplet, depending on throw distance requirements/LCD size.
LCD: Best available for a reasonable price, with a lexan "box" siliconed to each side for direct contact liquid cooling, which will be essential with 500+ watts of light being absorbed by the panel if using a 2500W lamp. A refrigeration system may be necessary to chill the water to keep the LCD cool enough. It remains to be seen if high intensity light damages the LCD over time. A cheap test LCD would be worthwhile before spending big bucks on a good one.
With the 2500W light and 50% reflector/condenser efficiency, assuming a 6% transmittance from Fresnel to objective lens, the light output would be 7200 lumens! With the more practical (no 240V required for power if using an electronic ballast) 1200W lamp, output would be about 3300 lumens with the same optical system efficiency.
Total cost would be about $1500 with a good LCD.
David
Lamp: 1200W or 2500W HMI film light. These produce 110,000 and 240,000 lumens respectively. The great thing is that the arc lengths are just 1cm and 1.4 cm respectively, so a reflector/condenser could conceivably capture 50% or more of the light produced. These lamps are relatively cheap ($100 on eBay for the 1200W) and last 750 to 1000 hours. The ballasts are expensive, but DIY comes to the rescue.
DIYLabs 15" Fresnel and standard or 135mm triplet, depending on throw distance requirements/LCD size.
LCD: Best available for a reasonable price, with a lexan "box" siliconed to each side for direct contact liquid cooling, which will be essential with 500+ watts of light being absorbed by the panel if using a 2500W lamp. A refrigeration system may be necessary to chill the water to keep the LCD cool enough. It remains to be seen if high intensity light damages the LCD over time. A cheap test LCD would be worthwhile before spending big bucks on a good one.
With the 2500W light and 50% reflector/condenser efficiency, assuming a 6% transmittance from Fresnel to objective lens, the light output would be 7200 lumens! With the more practical (no 240V required for power if using an electronic ballast) 1200W lamp, output would be about 3300 lumens with the same optical system efficiency.
Total cost would be about $1500 with a good LCD.
David
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.