CurtPalme.com Home Theater Forum

[tse]

The LEDs are mounted on standard aluminum extrusion heatsinks, air cooled with regular fans.

Scott

[Spanky Ham]

Thanks, Scott. Now I have a better understanding of how they are working. That was the one area I didn't have any knowledge of.

[dturco]

[macgyver655]

Because they are pics of copyrighted material I probably shouldn't. Probably the same reason Scott wouldn't post any. Of course he could probably get permission...

[macgyver655]

I'm still thinking about this group mirror thing, even tho I posted the group reset procedure. I've been doing some random reading but I may have to see if I can acquire the DMD software. However I came across this that has me thinking again.

Quote:

"Error-diffusion artifacts caused by averaging a shade over different pixels, since one pixel cannot render the shade exactly."

End Quote

Now I agree on the 1 mirror per pixel (per se) issue, but I'm still thinking multiple mirrors have to be used to increase a shade or contrast of an area of the image.

I'll continue random reading and post any other stuff of relevance I come across.

Comments are also welcome.

[garyfritz]

I believe I read that the TI DMDs handle 1024 levels of brightness. That means you have 30 bits of color represented in each and every individual pixel. I can't believe you'd need to dither pixels together to handle shading.

[macgyver655]

[macgyver655]

[garyfritz]

So you want to turn the brightness up to 11, eh??

But each mirror is a separate pixel. If you combine adjacent pixels you're not just increasing brightness, you're changing the original picture content. And with normal video content, usually a bright pixel will be surrounded by other bright pixels anyway. You can't increase the light output of the projector that way, unless you're really interested in single isolated pixels being displayed brighter than they would be already -- and "fatter" than they should be. And really they won't be brighter. They'll just be bigger. Each of the individual pixels would still be the same brightness it was before.

Unless I'm really misunderstanding what you're saying, that doesn't seem to make sense.

[VideoGrabber]

garyfritz commented:
> I can't remember how fast the switch them. I remember being surprised that it wasn't any higher -- it was maybe 2-3x faster than current wheels, or less. <

The LEDs are pulsed at 2.9 kHz, as I mentioned last summer. That's 9.6 times a 5x wheel.

[garyfritz]

Hm, OK, thanks Tim. 10x faster sure ought to be fast enough to prevent even hypersensitive types from seeing rainbows. Will be interesting to see one of these projectors when they get more popular and cheaper.

My brother has a rear-projection Samsung LED TV. I assume it pulses at the same speed? I've only seen it once but I can't remember if I could see rainbows with it or not.

[VideoGrabber]

dturco wrote:
> That looks awesome, like air condition coils for cooling. Very cool. <

Yes, they're similar to the Ninja (mini!) heatsink on the CPU in my HTPC.

[VideoGrabber]

Gary,
> My brother has a rear-projection Samsung LED TV. I assume it pulses at the same speed? <

Yes, that's correct. He has the HL67A750, from what I remember. Phlatlight claims a 100,000 hour lifespan on their LED emitters.

[macgyver655]

[Nashou66]

Nope That makes perfect sense Mac, I was wondering about what happens with lower resolutions being up scaled.

And to my Surprise i did not get a head ache !

Nashou

[VideoGrabber]

Mark_A_W commented:
> Unless they are doing a much faster sequence of colours with the leds, but the leds will take some time to turn off, so there will be limits. <

I don't have the exact numbers in front of me, but the LEDs turn off in <100 uS, and there is 0 persistence.

ecrabb wrote:
> As far as I know, for anything recent/current, there's a mirror on the DMD array for each pixel on screen. So, for a 1080p projector, the DMD has an array of 1920x1080 mirrors on the DMD, or exactly 2,073,600 mirrors. <

This is probably true for at least some displays, but others use a technique called wobulation to generate the full 1920x1080 resolution, using a 960x1080 DMD array. This technique also helps to eliminate SDE. Because of this temporal multiplexing, Samsung RPTVs (all the TI-based DLPs?) use the checkerboard method to generate 3D images, at full resolution.

(I was going to link to a TI WhitePaper on the subject, but strangely enough, dlp.com has nuked every link to every document on 3-D display technology they had, just a few days before CES started. Weird. I'll attach a copy of the smaller PDF that I saved locally.)

[garyfritz]

[VideoGrabber]

> Well now I'm really going to throw your a curve and it pertains to my comment on more then 1 mirror per pixel. <

I'll avoid pointing out that there's still only 1 mirror being used to display each pixel on-screen, even though there are more display pixels than source pixels. Whoops.

While you won't find more than one mirror being used to display a single screen pixel, the opposite is certainly possible. Wobulation is already being used to display 2 pixels using 1 mirror, and can be used for up to 4. So in the future, a 1920x1080 DMD device could be used to generate a Quad-Definition 3840x2160 image, at full resolution, with 4 pixels/mirror.

[VideoGrabber]

Dang! Gary beat me to it. Now you know how slow I type.

[ecrabb]