[RicardoD]

I finally received the SCR Colorwheel paper presented by Texas Instruments at the SID (society for information display) conference. I would like to thank D. Scott Dewald, the lead author of the paper, for generously providing this to us. I let him know we appreciate his support and applaud his efforts here at the AVSforum.

Sequential Color Recapture and Dynamic Filtering: A Method of Scrolling Color by D. Scott Dewald, Steven M. Penn, and Michael Davis


Many of you already know that this paper addresses a new type of color wheel/integrator for DLPs, which allows one chip devices to have the same light output as a 3 chip DLP.


I hope this helps everyone understand this new technology better.


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

All the best,

Ricardo
Sony KP-xxHS10 Zone


[This message has been edited by RicardoD (edited 06-30-2001).]

 

[Mark Hunter]

Great work, Ricardo! Fascinating stuff.


I still haven't put together the "SCR Recycling Process" though.


I also didn't realize that the spiral can be so flared! It's almost like a 100-section color wheel instead of the more "spiral" pattern that we were looking at before.


Thanks for posting.

 

[RicardoD]

A couple of interesting points I noticed:

• paper mentions reduction of "rainbow effect", called "color separation artifacts" by the author
• authors believe 10 lumens/watt (lamp) possible for single chip DLP projector / >13 lumens/watt for projectors with clear segment color wheels-- ANSI brightness measurement

anyone have typical lamp wattages of current DLP projectors?


Ricardo



[This message has been edited by RicardoD (edited 06-30-2001).]

 

[MrWigggles]

Ricardo,


What a Score!


Pete Broas just told me two days ago that the paper wasn't avaiable.


When he says 10 lumens/ Watt I would gather that he is refering to UHP bulbs.


The biggest UHP bulb right now is about 200 Watts and it puts out the same amount of lumens as the 270 Watt SHP. (The SHP having a little better colors).


This would me that a true 2000 lumen projector would be possible without any clear section. That is quite nice and is generally more than enough for most people.


I haven't read the paper yet, but am looking forward to it.


-Mr. Wigggles


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

The Mothership is now boarding.

 

[DanHouck]

Wow, one stop shopping for getting serious brightness from a DLP AND reduced/eliminated rainbow!


Anybody want to guess how long it will take this to make it into HT products? If less than a year, probably well worth waiting for.


Dan

 

[mflaster]

Am I way off base here, or does this system seem to really only boost brightness of whites?


For example, compare a RGB color wheel with this system. For white, leaving the mirrors on all the time, the RGB color wheel only lets through 1/3 of the input light, as 2/3 of the light is blocked at each color wheel section. For the SCR system, you get 1/3 + 1/3*2/3 + 1/3*2/3*2/3, or something like that, I guess they say for an 80% increase in efficiency.


But let's say you want to do pure red. For the RGB system, your lamp is on 1/3 of the time, and you get 1/3 of the light through, so you have 1/9 the input on the screen. For the SCR system, you have to wait for an all-red section to be in front of the "integrator" (right?). For an all-red section, you still only let 1/3 of the light through - you can bounce the light back and forth forever, but only the red light is getting through. But the problem is, what percentage of the time is 'all red' in front of the integrator? It can't be 33%, because a lot of the time you have these mixed color segments. So your pure colors will be *worse* than on an RGB wheel. In other words, if the RGB color wheel puts out a white at 1 unit, and a red or blue or green at 1/3, I think the SCR will put out the white at (theoretically) 1.8, but the red or blue or green at *less* than 1/3!


Similar analysis could be done for blended colors, assuming we knew what percentage of time a "useable" section of the SCR spiral would be in front of the integrator. For example, if you just want a mixture of R and B, you can never turn the mirrors on when you have any G in the color spiral in front of the integrator. That immediately eliminates ___ % of the device....


But I could be barking up the wrong tree here...


Mike

 

[dcottle]

Mike


The filter has all 3 colors present simultaneously in different layers. Each color of the filter is exposed to white light, passes one color, and reflects the other 2 colors. All 3 colors are simultaneously transmitted to the DMD and reflected back into the optical system, but this occurs in different sections of the filter for each color. There is never a solid red section as you suggest.


With this system, 2/3 of the light incident on the filter (assuming no white section) is reflected back, and since this is all 3 colors, the reflected light is white, on average. The integrator takes the reflected light and light coming from the bulb and emits light with uniform intensity that again passes or is reflected by the filter. Since each color is passed by only one layer, the light emitted from that layer is enhanced in brightness. So, no the system does not only enhance whites.


Dave

 

[MrWigggles]

Mike,


At any given moment, all three color sections are present at the end of the tube. The light bounces back and forth until it finds a section it can pass through.


What will be a little tricky is not losing to much color "focus" in the process (i.e. making distinct scrolling color bands of RGB and not just a scrolling progression of colors where, for instance, inbetween R and G you have an "unusable" band of yellow. To make it usable, the electronics would have to know how to use it.


I think it is very promissing technology but I wouldn't look forward to having it readilly available until at least the start of 2003 or maybe the Christamas season 2002 if there was enormous demand and a manufacturer really went after it for HT.


What this technology really unlocks is the potential for TI to make 3 times as many high end projectors than they are now. Currently the "black" chip has had some low yields due to manufacturing complexities. This would also put an end to convergence and the expensive optics involved.


DLP, in my mind has the ability to have an unbelievable upside. Brightness is the only thing holding it back and it appears to be solved.


My fearless prediction: a palm-sized 720p DLP unit that does 2000 lumens, has a true 1500:1 contrast ratio and SMPTE C colors by the end of 2003.


I might be a dreamer, but I just pinched myself so I am not currently dreaming.


-Mr. Wigggles


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

The Mothership is now boarding.

 

[mflaster]

I still don't understand.


If you want to project an all red screen, it seems clear to me that recycling light can't help - the G & B must be discarded, regardless.


Now I'm guessing that there are bands of colors moving across the DMD. There's a red, green, and blue band present on parts of the mirrors. As far as I can tell, the widths of these bands must be many pixels, but less than the size of the whole mirror assembly. So, this means that some pixels will be all red, some pixels on the border between bands will have some red and some green, etc. To show an all red screen, only the mirrors that are being bathed in all red light are turned on. Whereas with a color wheel, this is either all the mirrors or none, in the SCR, this is going to be a subset of mirrors in a pattern that matches the band from the spiral.


If the above is true, first of all, is a mirror ever turned on when it's *not* bathed in a pure color? IOW, are they ever turned on during the border section, where they have multiple colors? I could easily imagine that mirrors are only turned on during *non*-border sections, i.e. pure colors.


It still seems to me (based on my above assumptions) that a pure red screen will be at best 1/9, but really will be a lot less (as there are many more borders, compared to just an RGB wheel, so more unuseable space).


OK, next round of corrections please. http://www.avsforum.com/ubb/smile.gif


Mike

 

[Federico]

Mike, I think I understand your point.

Let's take a 3 dmd projector, when projecting a full red screen we will have only one dmd on and all the red component part of light from the lamp (1/3 white)will reach the screen. With the spiral only 1/3 of the white light would reach the screen but only during a time proportional to the size of area of red on the spiral wheel, if this is

13 of the area then we get 1/9 of the white light on the screen. Is that right? http://www.avsforum.com/ubb/rolleyes.gif


Federico

 

[mflaster]

Quote:

Originally posted by Federico: With the spiral only 1/3 of the white light would reach the screen but only during a time proportional to the size of area of red on the spiral wheel, if this is 1/3 of the area then we get 1/9 of the white light on the screen.

And that's assuming that *all* of the red in the spiral is used to push light through. I can definitely imagine scenarios where some of the red is unuseable, either for any color (e.g. the borders as I postulated previously), or specifically for a solid color.


Mike

 

[dcottle]

This seems to be difficult to explain, but once you get it, it seems obvious. The filter is 1/3 red, and 2/3 other colors in a pattern which cycles through all the colors with time. If the spiral filter is a pass-through filter as in a standard color wheel, 1/3 of the light passing the filter will be red, and it will reflect off of 1/3 of the DMD mirrors to the screen. However, the dichroic filter is a pass-through/reflective filter. Red will be reflected by the non-red portions of the filter, bounce around in the optics, and ultimately some of it will ultimately pass through the red section of the filter to the same 1/3 of the DMD mirrors. Whatever light that has been recovered by reflection from the non-red portions of the filter adds to the brightness of the red on the screen.


The red would be 1/9 of the white light with a pass-through filter, but because unused red from the non-red sections is recovered (partially), the system more closely approaches the 1/3 of the white light which would be the best that could be achieved.


Dave

 

[mflaster]

Aha, an explanation I think I understand!


Let's pretend that there are exactly 3 bands hitting the mirrors at a certain time (one in the green "shadow" of the color spiral, one in the red, and one in the blue), and let's also pretend for explanation simplicity that these bands are horizontal (e.g. red is on the top rows of pixels, green is in the middle, and blue is on the bottom). You're trying to show an all red screen. At this point in time, the top 1/3 of mirrors are on, and all others are off. The red on that top row is intensified with the recycled light from the lower 2/3 of the pixels, from the blue and the green rows. If you had 100% recycling efficiency, all of the red light that left the lamp and would have hit anywhere on the DMD is reflected around and eventually pushed out onto the top 1/3 of the pixels. So the red is 3 times as intense (with an impossibly perfect recycling chamber) as compared to just using a red filter without any recycling.


As time goes on, the "shadows" from the spiral marches (say) downwards. There are obviously lots of intermediate steps where there are 4 bands, which is slightly harder to describe, so I'll just skip ahead to when there's 3 bands again, but now red is in the middle, green is on the bottom, and blue has wrapped around to the top. Now, just like in the previous step, all the red light is used, except now it's hitting the middle rows.


The upshot is that in an RGB color wheel, if you could watch in ultra-slow motion, you would see the entire screen the same color, first red, then black, then black, then red, then black, then black, etc...


For SCR, 1/3 of the screen would be red at any point in time, and this red section would scroll down, eventually wrapping around, and coming back out from the top. This red is actually extra-bright, it has all the red (if recycling was perfect) from the full screen color wheel example squished down into a smaller area, and you don't have the screen black 2/3 of the time. Hence red is theoretically 3 times brighter.


Hopefully I'm not the only thick one that benefited from this discussion. And if I'm still confused, please correct me.


Mike

 

[Federico]

Godd description dcottle, now i can see where the light is.
http://www.avsforum.com/ubb/wink.gif

Federico

 

[PhilB]

I just looked at the paper and noticed that the picture of the SCR prototype on page 4 appears to show red, green, and blue on the DMD at the same time. I don't know if this is an optical illusion of some sort or if this is the SCR in action.


-phil

 

[KFung]

That's the SCR colour wheel in action, using dichroic filters to reflect back unused light and showing all three primary colours on the screen at the same time is what allows the SCR colour wheel to work.


Mike's got the right idea now (of course the recycling is not perfect http://www.avsforum.com/ubb/smile.gif ).


Regards,


Kam Fung