Originally Posted by Dave Harper
Yes, but doesn't it address 8.3 million of those 16.6 individually, making it comply with CTA standards? I am seriously curious as I don't know and am still in the basic confusion stage of exactly how this tech works. Could it be that each mirror isn't exactly a "pixel" of info, rather just a reflection surface to project the image and fill in the space, and the actual "pixels" are within this and spread amongst and between a couple of the mirrors and wobulation/panel shift? Damn, I confused myself just saying that, haha!
First, lets look at the most technical info we have available, that being Barco's whitepaper on the subject:
Based on that whitepaper, "XPR" DLP works (conceptually at least) exactly the same way e-Shift works. There is a chip with some native resolution, and an optical actuator that can shift that chip's pixels by a half pixel diagonally. Each frame on screen is then made by the sum of the two overlapping fields. The second subframe effectively splits each pixel in the first frame into four pixels.
Of course, since there's overlap, you can't control each of those resulting pixels independently, but with clever processing you can produce something that's significantly better in detail than you would with the chip's native resolution alone. But at the same time, it means you can't reproduce artificial patterns correctly.
As for CTA, all I have to go on is what's in the linked press releases: "Display Resolution – Has at least eight million active pixels, with at least 3840 horizontally and at least 2160 vertically."
Which brings us to the issue of the ambiguity in that "specification". Considering CEA/CTA is basically a marketing organization, I think we have to assume a rather loose definition, one that is free enough that pretty much any manufacturer can claim to meet it so long as they can make a reasonable explanation. TI's marketing department was very clever in this regard, they found they had a ~4Mpix DMD on the shelf, and there are already optical actuators available, they could combine the two and then have a very easy, unique marketing
solution for 4K, regardless of the technical or engineering realities.
Now when it comes to the actual technical reality, I don't think the number of pixels used to create the image is necessarily that important in it's own right, it's the effects, benefits, and side effects of that implementation that are important. For example the fact that single chip DLP only uses one chip to show all three colors is not really and important distinction vs 3-chip machines, you wouldn't say it's not color because it's not showing all colors at the same time. Likewise you wouldn't say the same as a 3-chip, since there are technical limitations to the choice (RBE namely).
So when it comes to pixel shifting, how many pixels are used in the chip, is not really important, it's what's on screen that matters. Consider for a moment, the Picobit, it shows one pixel at a time, so how many active pixels does it have?
I guess what I'm saying is, the CTA definition of 4K, and linked in this thread (which is all I have to go on, if there are more details elsewhere I'd like to read them), is loose enough that both JVC/Epson's solution based on a 2K chip and TI's solution based on an approximately 2.7k chip, meet the definition because they place more than 8 million pixels on screen, and have a resolution of at least 3840x2160.
On the flip side, neither solution produces 3840x2160 independently addressable pixels, and both solutions have all the limitations and side effects that have been so well advertised of the JVC/Epson solution.
Originally Posted by Ruined
Actually 2716x1528 * 2 is *exactly* what is displayed. For each frame, it first displays 2716x1528 pixels (4,150,048 individually addressable pixels), then follows it with another 2716x1528 pixels a second time shifted a diagonal half pixel (another 4,150,058 pixels). This is a total of 8,300,096 individually addressable pixels per frame.
By this logic single chip DLP isn't color, since it shows Red, Green, and Blue sequentially. You have to remember the screen is part of the system and that is not what's displayed (ie seen) on the screen. What you see is the integration of those two frames which is 16.6 million pixels that are not independently addressable.
It's not 16.6 million individually addressable pixels with XPR. XPR displays 4,150,058 pixels two times (shifted), for a total of 8+mil individually addressable pixels per frame. The 16.6 million composite of pixels created is not individually addressable. You cannot pick out one of the 16 million and turn it on and off, because there are not physically 16 million pixels. There are only physically 8.3 million pixels per frame, and you can individually turn on/off any of the 8million pixels in the 2716x1528 x 2 shift.
Except you can't turn off just one of the pixels on screen, if you turn off one of the DMD mirrors in one subframe, it affects 4 pixels on screen, thus, not individually addressible.
This is why JVC/Epson fail to meet the CTA benchmark, as well. They display 1920x1080 x2, or 4,147,200 individually addressable pixels. There are only 4 million physical pixels that can be individually addressed per frame, not 8 million.
Where does CTA say it has to be individually addressable? The links you provided, and the releases they reference just say "pixels".