Originally Posted by stanger89
I suspect they're both "right", perhaps one in a more technical way and one in a more philosophical way. What I mean is by everything I've been able to gather, the input to the e-shift system is a "4k" (2160p) signal. For the older JVC models, this was scaled inside the projector, with the new ones, it sounds like it can be input directly. e-shift then uses some fancy algorithms to extract two
1920x1080 subframes from the 2160p input and feeds it to the hardware which flashes them each, offset by a half a pixel in each direction.
That's the trade off, the system does create 3840x2160 unique pixels, as in it can render a full 3840x2160 areas of different color on screen. However it cannot directly, uniquely address each of those areas individually. So you won't (I don't think) be able to put up a 1x1 checkerboard grid like you could with a native 4k system, but depending on the algorithm they use to extract the two 1080p subframes they should (and apparently can) produce something significantly better than 1080p with real world content.
This is exactly my understanding as well. If this is how e-shift works then it's 100% incorrect to say that 4K input would be downscaled first, then afterwards converted to e-shift, which is what kraine said. There's no philisophical leeway here.
Originally Posted by stanger89
The thing to remember is that it's likely (almost certain) that neither 1080p subframe is merely a subsampled/downscaled version of the input 2160p signal. Likely the algorithm evaluates each pixel in the 2160p input, along with those around it to figure out what each pixel of each subframe needs to be to create the best output when combined.
Originally Posted by fisher191
I would have thought that kraine is correct - JVC have to downscale because the panels are '2K'. Sure, they can shift them and interlace (if that is the right word), but each frame is going to be downscaled to 1920x1080 because that is the panel's pixel size.
If kraine is right then the 4K input would be downscaled to 1080p first, then afterwards converted to e-shift. This would mean that there wouldn't be *any* benefit to feed 4K into the projector at all. You could just as well let the 4K source downscale to 1080p and feed that to the projector, the end result would be the same.
This is not likely what is happening (although it is possible, we can't know for sure right now). The more likely technical implementation is that the conversion to e-shift works by taking a 4K video frame, then running some complicated mathematic algorithms on the 4K frame to extract (not downscale!) two separate 1080p subframes which when displayed with a half pixel offset will result in a 4K projected image which "somewhat" approaches the original 4K video frame. Obviously JVC can't control every 4K pixel separately. If they change one 1080p pixel in one of the two extracted 1080p subframes, this will affect four 4K pixels in the final blended 4K image on screen. So the algorithm which extracts the two 1080p subframes will have to take this into account. The final projected 4K image will only be an "approximation" of the original 4K image. But still, it should be possible to achieve some improvement over a simple 1080p image.
So, it is my understanding that the current JVC e-shift projectors are doing this:
(1) a 1080p frame arrives via HDMI
(2) the projector internally upconverts it to 4K
(3) the e-shift algorithm extracts two separate 1080p subframes from the upscaled 4K video frame
Now the new JVC projectors, when fed with a 4K source via HDMI will probably simply skip step (2). So they will probably do this:
(1) a 4K frame arrives via HDMI
(2) the e-shift algorithm extracts two separate 1080p subframes from the 4K video frame
This means that sending the new JVC projectors a native 4K source should look better than sending them a 1080p source.
Of course this is all guess-work, so it might be incorrect. It is possible that kraine is correct, but he's just guessing, too - without saying so!! And that is what I have been critizising. If you guess, you should say so, and not just state your guess as fact.