AVS Forum banner
Status
Not open for further replies.
1 - 20 of 36 Posts

·
Registered
Joined
·
619 Posts
Discussion Starter · #1 ·
I posted the following over on the RPTV Forum, but Rogo thought it might produce a better discussion over here:

I've been reading about DLPs since I saw them written up in an early 90s IEEE Spectrum and I've always thought that they were going to (eventually) replace CRTs. Of course, that was a decade ago and a lot has happened in Display Technology. But, one thing that hasn't changed regarding DLPs is the light source.


There seems to be two ways to produce the full-color image: One DLP Digital Micro-mirror device (DMD) and a spinning color wheel, or three DMDs and a color-splitter to provide RGB for each DMD. Has no one thought to use a single DLP DMD with three separate Primary Color light sources that can be switched at a fast enough rate to eliminate the dreaded "Rainbow" effect?


I know that LEDs have made great strides in the amount of light that can be produced. How many LED stoplights have you seen lately? So, I take it that red and green high-output LEDs can be produced at a cost-effective rate. If blue LEDs are two-four times the cost of red or green, wouldn't it still be possible to produce a bright enough to provide an acceptable light level in an RP DLP set? (This is why I'm asking this question here instead of the Projector forum)
(Okay, now I'm asking it here:))

Instead of a color wheel, the set would cycle thorugh the LEDs at a rate that would be faster than any color wheel (goodbye rainbows!) Better still, instead of R128, R64, R32, R16, R8, R4, R2, R1, pause, G128, G64. G32, G16, G8, G4, G2, G1, pause, B128, B64, B32, B16, B8, B4, B2, B1, pause, start over; you could eliminate the pauses between color changes and interleave the colors (R128, G128, B128, R64, G64, B64, etc.) And, it would be silent since there would be no moving parts. Lastly, LEDs are good for hundreds of thousands of hours, so it should last as long or longer than the MMD.


Another benefit, instead of a fixed light level, you could extend the effective contrast of such a set by reducing the light intensity for low-light scenes, something that can't be done with current light sources. Instead of gray fog, just cut the light level down and increase the number of bits available to render such a scene.


So, will this ever happen?



Now, granted, the lumen requirements for a FP are a lot higher than that needed for RP, but since DLP FP tend to be priced upwards of $5K, you would just use more LEDs. Or, given the higher price point that FPs seem to tolerate, how about red, green, and blue lasers, each tuned for the HDTV color space?
 

·
Registered
Joined
·
3,936 Posts
Foxbat:


this was my question too to Pete Broas DLP TI Marketing manager:

http://www.avsforum.com/avs-vb/showt...0&pagenumber=2


-------------------------------------------------------------------------------------


LEDs in the future?

Hi Pete,

Thanks for your very informative answers.

I'm relatively new in the digital projection arena and I've been quite impressed by DLP technology.


In my short time here at the forum, I encountered numerous comments regarding the "rainbow" phenomenon, headaches etc. etc.

Granted, all these comments come from LCD PJ owners that likely feel threatened by an emerging technology that has more than double the fill ratio of LCD's on top of a few other advantages.


However, all these negative comments can and do affect people especially those with not much experience in this exciting field.


What do you, as TI, plan on doing to put all these rumors/comments to rest by eradicating the "rainbow" phenomenon? By the way, I've heard people claiming to be seeing "rainbows" on projectors with 2x, 3x, 4x and even 5x (!!) wheels. How anybody's brain could discern motion artifacts at 300Hz is beyond me... However ridiculous these comments may be, they're still there and they negatively affect DLP's reputation.


Are there plans to "retire" the color wheel and in its place use R, G and B LED's? I understand that at this point they're incapable of supplying enough luminous intensity for say a 1000 lumen projector but their technology/cost are rapidly approaching the realms of commercialization.


There are MANY advantages in using LED's, some of which are:

(1) No need to use 3-chip DLP's anymore since there is no spinning wheel.

-LOWER COST


(2) Pure, (very close to) single wavelength light for each of the RGB LEDs.

(3) Higher efficiency than short arc lamps. Cooler/cheaper/quieter/smaller/more relaible electronics.

- LOWER COST

(4) No need to replace ultra-expensive lamps after 1-2k hours anymore even if the LEDs are overdriven, MTF ~100khours.

-LOWER COST

(5) instead of 1- 1000lumen lamp, 3- 333 lumen LED's are needed.

-LOWER COST

(6) No more "rainbow" complaints, LCD defenders are defenceless...

(7) Truer, even more vibrant colors, since the spectrum of a short arc lamp is nowhere near flat.

(8) Simpler, cheaper optics

-LOWER COST

(9) No filters used, more light output to the screen.

-LOWER COST


I could go on and on, but I think you get the idea.

I'm not expecting to see LED prjectors in the near future, 5-10 years hopefully, but there are so many advantages tied to them that it would be unwise not to explore the possibility.


Looking forward to read your comments.

Thanks,

Andreas Georgiades


--------------------------------------------------------------------------------------------


This was my firrst question. go to the thread and read the discussion that followed.
 

·
Registered
Joined
·
619 Posts
Discussion Starter · #3 ·
Quote:
This was my first question. go to the thread and read the discussion that followed.
A fascinating thread, I'm sorry I didn't know about it earlier.


I thought the LED idea was too simple not to be figured out by other people, I'm surprised that Pete seemed to think that it was impractical compared to the color wheel. When you consider the heat that the projector lamp generates and how cool high-output LEDs run, it seems that someone out there should have something like this in their lab.


I may have over-estimated the speed of the DLPâ„¢ mirrors, though. I thought they were capable of hundreds of thousands of transitions per second. Since the LEDs can be modulated millions of times per second, perhaps it's overkill since the mirrors can't take advantage of displaying the frame multiple times in 1/30 of a second.


And the comment about not being able to focus all these LEDs to one point? You could use fiber optics or collimating optics to bring the separate LED chips to "one" chromatic light source. And why stop there? What about four points? Then, each projected image would be slightly offset, yielding a higher "perceived" resolution, i.e., your 1280x720 Mustang becomes a 2560x1440 display device! You just need four times the LEDs, and some good software in the drive board.
 

·
Registered
Joined
·
619 Posts
Discussion Starter · #4 ·
A little Google research brought up Lumiledsâ„¢ , a company that has super-bright LEDs (their Luxeonâ„¢ line) capable of up to 120 lumens! They are available in a number of colors and intensities. It looks like heat will be more of a factor than we thought, but still not as much as with a projector bulb.
 

·
Premium Member
Joined
·
24,995 Posts
"And the comment about not being able to focus all these LEDs to one point? You could use fiber optics or collimating optics to bring the separate LED chips to "one" chromatic light source."


Sure you could, but how many parts does it take and how much much would it cost?


"...super-bright LEDs (their Luxeonâ„¢ line) capable of up to 120 lumens!"


So you're probably thinking a handful of each color and you're done. But pj's only get a small fraction of the bulbs output to the screen; a pj with 1000 L output has a bulb putting out 5-10K L.


Now that I think about it, an LED replacement for a 3-chip design might be feasible (I'm sure someone will be happy to shoot the idea down after my wet-blanket comments on others' suggestions), because there isn't the requirement to have all the colors emanating from the same location in order to ultimately hit the screen at the same point.


For the light path for each color, a circular array of LED's could be focused by a reflector to feed each path and combined as usual.
 

·
Registered
Joined
·
578 Posts
I just saw those LumiLED 'Stars' myself in the latest Pop Mechanics, and just looked 'em up on the net...


Seems like on top of their 1W bulb which outputs ~120lumens, that they have (or are working on?) a 4 element bulb which for some reason is 5 watts.


No info on lumens but I guess I'd guess near 500 lumens?


Now imagine a small square of nine and maybe it's a little low, but it's gotta be close to 'enough' light output, and that's right now, not 5-10 years from now.


Certainly more than CRT FTpj's output, and isn't that good enough for a start mated to a high gain screen? Yeah that's less than most digital projectors, but I'm sure some people won't mind that much as long as the picture's great.


Personally for me here in AZ, I just can't have bulbs that are massive space heaters.


I'd take a low-ish lumen LED based pj that had very little heat but had a light that never dimmed over time or had to be replaced ever (probably).


It's what I'm waiting for in a projector.


I'm counting on electrodeless RF bulbs or LEDs to be the answer. Seems like RF bulbs have stalled for some reason?? I don't know why though 'cuz I thought they've already been proven to work and to be bright enough for projector applications?


Anyone know any 'latest news' on that front?


Anyway... those 120lumen LEDs are like $15ea. and I'm guessing way cheaper in bulk.


I think I read the 5W version is $40 (again... each, not in bulk)? So maybe a 4500 lumen array and some element to focus the light wouldn't cost that much.


Certainly not more compared to 10K hours worth of $300-$400 high pressure bulbs, and the hassle of buying new bulbs and replacing old ones, or one popping long before it's rated life.


Would you even have to focus the LEDs much if they were in a square aimed at a similar size DMD chip? Would the optic (or whatever you'd need) really be that complicated and costly?


I'd think if a projector company was going to use LEDs as the light source, they'd specifically design an array that had all the bulb elements mounted on the same board as tightly as possible.


Now, I'm all for getting rid of the color wheel (I've still seen rainbows in fastest designs) but for those who don't have a prob, then you'd still have the no heat and lifetime blub benefit in going to LEDs.


Like Noah said though... LED's seem to be an even better idea for a 3 chip design which due to cost would tend to be LCD or LCoS microdisplays... not DMD's.


And though this ain't the forum section... LEDs should be even closer to 'bright enough' for rear projection microdisplay units right? Maybe Noah or someone knows how much light those bulbs need to put out? Much less right since the screen's so close?
 

·
Registered
Joined
·
111 Posts
The topic comes up often, but here's the problem with LEDs.


The mercury-vapor bulb in the xb31 consumes roughly

160 watts and produces 10,000 lumens, *** but it does

so in a fireball 1.2mm in diameter ***. Because the

fireball is small, the reflector can focus most of the light

into an integrator rod (light homogenizer) that is 4mm x

6mm in size, which produces the small beam of light that

will flow through the rest of the optics system.


As a single fireball grows in size, so does the size of the

point onto which you can focus it using any reflector.


If you have multiple fireballs or light sources, it is difficult

to *efficiently* harness all that light and get into into a

point and then a small beam.


As it stands now, mercury-vapor has about the same

efficiency as bright LEDs (60 lumens/watt), but it also

has a small fireball which can be focused, unlike LEDs.

Tell an optics person they need to harness 150 point

sources of light and get it all into a beam with high

efficiency, and their eyes will cross.


--xb31
 

·
Registered
Joined
·
4,718 Posts
I have several Luxeon Star LED flashlights. My brightest light is a single 5 watt LS flashlight that puts out close to 120 lumens.


The LS bulbs are less efficient than the standard LEDs that you are used to seeing and do put out some heat. The life of these high output LEDs are also shorter, somewhere between 1000 and 10,000 hours instead of the 100,000 hours that the small 1 watt bulbs are capable. All of these factors are improving though, and I expect to see LED's replace all other bulbs in my lifetime :)


Another drawback to the LS bulbs is the board they are attached to is a bit bigger and they require heatsinking to carry away the heat so you can't cram as many of them together as the small LEDS. This is mostly a limitation in flashlights though and a projector has less space restrictions and can use a fan to dissapate heat.


For more information on the latest LED news and flashlights, check out the Candle Power Forum my other home after the AVS Forum :)
 

·
Registered
Joined
·
183 Posts
 Here is a manufacturer's (Luxeon) whitepaper on using LEDs for projection technology. Basically they say they'll be able to get close to 200 lumens out of a DLP unit by the end of 2004 using LEDs as the light source. Based on the roadmap for LEDs, 400 lumens by the end of 2006.


A promising technology, but for the next five years will only have a chance of creaping it's way into smaller display devices, not front projection for home theater.


-jp
 

·
Registered
Joined
·
3,667 Posts
Most often, the output of an LED is/or can be made/ reasonably collimated improving the lumen efficiency.



Also, to my knowledge, well heat sinked LED arrays have never been a high priority for manufacturers. 3 tightly spaced collimated LED arrays and a prism would make a very nice, very efficient light source.


Green LEDs can be very efficient at 500 lumens/watt.

Red LEDs can be found with 150 lumens/watt.

Blue LEDs are the least efficient at maybe 20 lumens/watt.
 

·
Registered
Joined
·
32,172 Posts
OK, so the answer is, as Free notes, LEDs will eventually replace projector bulbs as they will eventually replace almost every bulb in this country.


Apparently, the nation's lighting electricity consumption could fall by 10% using this technique making it that much easier for the U.S. to comply with the Kyoto Accords.


Oh, wait, we didn't join those.


mark
 

·
Registered
Joined
·
6,092 Posts
Quote:
Originally posted by jpschmoe
Here is a manufacturer's (Luxeon) whitepaper on using LEDs for projection technology. Basically they say they'll be able to get close to 200 lumens out of a DLP unit by the end of 2004 using LEDs as the light source. Based on the roadmap for LEDs, 400 lumens by the end of 2006.


A promising technology, but for the next five years will only have a chance of creaping it's way into smaller display devices, not front projection for home theater.


-jp
jp,


Great piece,


This topic has come up in the past on this forum. And in the past I have said something along the lines of "LED's won't be used for quite sometime but 1-chip DLP would be a good candidate because you could duty-cycle the RGB LED's in high output burst mode and get rid of the light-robbing colorwheel."


Well, it looks like I am partially correct. From the report:

Quote:
It is important to note that for the DLP it is assumed that the luminous flux loss by duty cycling the LEDs is compensated for by driving the LEDs at higher current densities.
Which is what I predicted. What I didn't predict is that they could have 120 lumen projector by Dec 2003! I thought it would be much longer than that.


120 lumens doesn't sound like much but for a 46" RPTV that would be about 30ft-L (about 300 nits) when screen gain etc. are factored in.


The best news:


1. No moving parts

2. 50,000 hour light source

3. 100,000 hour DMD

4. No fans

5. No filters

6. No noise

7. No color shift over time

8. 12X+ color sequencing (i.e. No rainbows)

9. Exact color timing (no color wheel transitions for the designers to worry about)

10. NO FLICKER!!


The list goes on. Simply awesome.

Guys, you just leave you TV on all the time. Put up a fish tank screen saver when you aren't watching TV. Whatever you want.


CES2004?


-Mr. Wigggles
 

·
Registered
Joined
·
64 Posts
The Luxeon white paper states that a 175 lumen DLP projector using LEDs is possible by the end of 2004. However, this lumen calculation is based on a 0.7" XGA chip. Since the maximum light output is based on an etendue limit, we can interpolate power outputs for other chips since they will scale by the chip size. Using Figure 5:


DLP with 1280 X 720 pixels: 205 lumens

DLP with 1920 X 1080 pixels: 460 lumens


Note: the above assumes that the mirror pitch remains 13.68 microns for the 1920 X 1080 chip.


SXRD with 1920 X 1080 pixels: 75 lumens


Clearly, DLP will have the advantage when LED based systems are introduced.


MrWigggles,


I would add a couple more advantages to your list:


11. NO temporal or spatial dithering required.


LEDs can have pulse widths much less than a microsecond, where as DLP mirror toggle rates are limited to about 20 usec. If you set aside 6 mirror toggle periods per color and pulse the LEDs for 10 usec, 5 usec, 2,5 usec, ... 0.3 usec during these periods, you can create a 14 bit linear color depth without resorting to temporal or spatial dithering.


12. Extended black level for dark scenes.


In dark scenes there can be intervals when all mirrors are in the off position. During these intervals, the LEDs can be turned off reducing their contribution to black level pollution. In the case of a blank screen, they can be turned off completely resulting in an infinite full on to full off contrast ratio.
 

·
Premium Member
Joined
·
24,995 Posts
Richard,


Re your points 11, I don't see how pulsing the LED's helps the dither problem at all, because it affects illumunation of all of the pixels. Dithering is done to deal with signal levels on a per-pixel basis.


Re your points 12, same thing. The only time you could turn the the LED's off is if the scene is completely black. This would happen seldom enough that it would probably just be a distracting reminder of the less than perfect black levels the rest of the time.
 

·
Registered
Joined
·
64 Posts
Noah,


Re point 11:


DLP devices create gray scale thru pulse width modulation (PWM). The minimum pulse width is set by the maximum toggle rate of the mirrors and is about 20 usecs. This means that the minimum pulse width is about 1/256 of the entire time allocated for a frame. Thus, the DLP device is only capable of 8 bit linear grayscale. To extend the depth of the greyscale, todays DLP projectors dither the image such that the average light level in a given area is correct even if the individual pixels are not.


Lets assume that within one frame and one color you had 261 time slices in which to implement the PWM gray scale. 255 time slices could be used to implement the conventional 8 bit linear scale. During the remaining time slices you pulse the LED with ever shorter pulses as a described in my previous post:


Time Slices 1..255: always on (20 usec)

Time Slice 256: 10 usec

Time Slice 257: 5 usec

Time Slice 258: 2.5 usec

Time Slice 259: 1.25usec

Time Slice 260: 0.625 usec

Time Slice 261: 0.3125 usec


For all those pixels in the image that have an intensity of 1/512 of full scale, the associated mirror is toggled on during the time slice with the 10 usec pulse. Similarly for pixels with 1/1024 intensity, etc. The mirrors can be toggled in combination (as in any binary system) to create the extended bit depth.


Selection the these narrower pulses is on a per pixel basis because they are selected by the mirrors.


Re point 12:


You said: "The only time you could turn the the LED's off is if the scene is completely black" This statement is simply not true. The only time you can turn the LED off is when all the mirrors are in the off position. Take for example a dark scene in which the brightest pixel is one half of full scale. This pixel will be in the off position for one half the time. If the rest of the pixels in the image are also off at the same time the LED can be turned off. Remember that DLP projectors are pulse width modulation systems not amplitude modulation systems.


The extent to which you can take advantage of this feature is dependent on the PWM algorithm. If the time slices used to implement a gray level varies too much from pixel to pixel, there may never be a time when all the mirrors are in the off position at the same time.
 

·
Premium Member
Joined
·
24,995 Posts
Hmm, well perhaps the utility is somewhere between what we're each saying.


Re 11, the brightest pixel will determine the minimum duty cycle of all the LED's.


Re 12, I guess in dark scenes the brightest pixel might well be far below full brightness, so lowering the LED duty cycle could be done to the benefit of black levels.


I wonder what the IRE is of stars in space?
 

·
Registered
Joined
·
6,092 Posts
Richard,


LED's won't help low level dithering. Sorry.


However, at 120 lumens, the image isn't going to be very bright so that you probably won't be able to see them anyway.


Temporal dithering isn't much of a problem if all of the contributing factors are reduced.


1. Keep the gamma in the 2.2 range not 2.5 or some that go all the way to 3.0


2. Keep the image intensity reasonable. Try to keep the image around 25 ft-L. The brighter the image the more temporal dithering is noticeable.


3. Filter the analog inputs well. Dither garbage can happen if the input signal isn't sampled well.


4. Use good source material. Mpeg artifacts can look a lot like dither.


The Sim2 HT300+ does all of these things very well. The Samsung DLP RPTV is the opposite. And with the former I can barely see dither and the latter I see quite often with dark material.


So, LED's won't reduce dither BUT you could be right somewhat with the black level comment. The LED's will still have some transition time on the order of the color wheels transition time but you don't have the problem with the mild light scatter caused by the color wheel. I think LED's all things else being equal could increase contrast ratio by about 20% or so.


-Mr. Wigggles
 

·
Registered
Joined
·
619 Posts
Discussion Starter · #19 ·
Another LED advantage: Near instant "on" times for the FP/RP. On one of the other threads, people are questioning how long their Samsung DLP™ RPs take to turn on. Obviously, the light has to come up to temperature and settle down, so this time is somewhere from 15-60 seconds. An LED set should be able to show an image within a few seconds, depending on the firmware in the set.


Plus, turning the unit off (or, heaven forbid, having the power go out) should also be a yawn: If the heat sink on the LEDs is done right, it should protect the LED array after the current is removed, instead of a projector bulb that needs to be cooled for a couple of minutes after you're done watching TV. And, if you change your mind after turning the set off, you would be able to turn it back on immediately instead of worrying about the bulb going through a full thermal cycle.


On the low-light scenes, it obviously won't help if you're watching Star Wars or some other bright object on a black background, but when you get to a dimly-lit scene that seems so popular these days, the drive controller would look at the brightest pixels in the scene and modulate the LEDs so that 100% DLP mirror cycle would generate the brightest pixel and it would go down from there. Also, if the brightest pixels were mostly red or blue, the other colors could be further reduced, although that would "color" the lower black level to an off-black. It might not be noticeable; that's for the engineers to determine in the lab.
 

·
Registered
Joined
·
245 Posts
[edit: I forgot to include color temperature in the analysis (thanks RichardB!). The power numbers become more reasonable but a 1000 lumen LED projector will still need a large fan for years to come.]


From the Lumileds whitepaper:


Optical efficiency for LCD/DLP is 13%/36%. Achieving 1000 lm output requires 2900 lm from the source for DLP.


From their table:

RED 44 lm/W = 66W

GRN 29 lm/W = 100W

BLU 4 lm/W = 725W


An arc lamp dissipating 200W is easy to keep cool because it can glow white hot (get it? White :D). Keeping 200W (or more) of LEDs cool is going to be very difficult since the LEDs must remain much cooler or they will die. Even a big CPU heatsink/fan won't cut it.


If they improve quantum efficiency by a factor of ten (unlikely in the next decade) the LEDs would still have to dissipate 90 watts to get 1000 lm. This is more power than a 3 GHz Intel Pentium 4 running full blast. LEDs are not going to happen anytime soon without a HUGE fan...
 
1 - 20 of 36 Posts
Status
Not open for further replies.
Top