JVC 4K projection discussion. - Page 18

I know right?! This has been a crazy time in Chitown sports... not to mention my Bears stink!

Yes! you are absolutely right because so far there is only 2K imagery projector but Sony and JVC will bring 4K projector. Although i have heard, Sony has brought 4K projector that shows 4K movies and incredible 3D with no triple-flash artifacts. I have reviewed on HD Projector review.

Sorry if I sounded snappy before, but probably because I was in the middle of a program dealing with pixels and you caught me at a frustrated moment

New PJ calculator coming soon, maybe you guys at AVS will find some use for it, put a lot of work into it, and I think everyone is going to like it...

Hopefully I will see an RS-55 eventually at a showroom and get to check this out, many people have said it increases the smoothness of the motion and image, that is very interesting to me.

for what I can see lots of improvements were made in the new JVC projector series. But what about short lamp spare life that plagued the units last year, except the X3. Did Jvc fix the issue? Is it no more a problem? If don't, what was the cause of it, since I read the entire thread and, at least for me, there was not a final word?

Hey Waltie, I would not expect final word on this until June/July

If each individual 1920x1080 frame that create the 4K is at 60htz does that mean having e-shift on limits you to having judder via a 2-3 pull down?

If so this is a step backwards....

The subframes are clocked at 120 Hz, not 60.

Not unless they really messed it up. I'm guessing they're able to switch frames on either shift. There really isn't a reason why you have to have an image displayed the same number of times shifted and not shifted. e.g.


Rather than:

so you are thinking it is doing a 5:5 pull down? If so that is great news

That would be my guess... but take it with a big grain of salt, it is only a guess. Just because it makes sense they would use a 5:5 pulldown doesn't mean they actually do.

My guess is for 24Hz source material the new JVCs with e-shift will display at 96Hz repeating each non-shifted frame 2 times and each shifted frame 2 times (effectively 2:2 pulldown X 2). If frame interpolation is used then the sequence would be the original 4K scaled frame (main + e-shifted) would be displayed one time followed by the interpolated 4K scaled frame (main plus e-shifted displayed one time. The projector's 120 Hz display mode (60Hz main and 60Hz e-shifted) may only be used for 30Hz and 60Hz source video.

This post is probably more likely.

That makes sense, and is consistent with how JVC handles 24fps 3D.

Everything I read implies no frame interpolation in e-shift mode. Probably doesn't have the horsepower for it. The e-shift calculation by itself is quite complicated with it optimizing for the 2-pixel mapping across the entire field. My guess is that next year's models will have FI for both 3D and e-shift.

Don't be so sure that FI doesn't work along with e-Shift. Gary (JVC UK) has confirmed in the AV Forum (again a UK forum) that the FI will work in 2D and not 3D modes and he also has indicated that e-Shift is the normal display mode (others have indicated it can only be turned off with a setting in the service menu). I really don't recall if during the demo/discussion of the RS65 at CEDIA whether the use of FI in combination with e-Shift was specifically demo'ed or mentioned.

Hi Ron. They did say the original 1080p frame would be flashed first, then the pixels created by the scaled frame, then the original 1080p frame, and then the scaled pixels frame Perhaps this was a simplification but I would think whatever two different frames are being flashed, ABAB would better fool the eye into seeing a combined AB frame than AABB being flashed.

I gave an example on the previous page that I still think makes a lot of sense...

According to JVC they first upscale the 1920x1080 frame to 3840x2160.



Obviously they can't project 3840x2160 so this has to be down-converted back to 1920x1080. The algorithm(s) JVC uses to derive the two subframes is proprietary but one way to do it is to discard everything except the ODD Row and Column pixels from the upsampled 3840x2160 image. This results in an 1920x1080 image which we'll call Subframe A. To create Subframe B start with the upsampled 3840x2160 image again but this time discard everything except the EVEN Row and Column pixels.

Notice how the Subframe B pixels fall between the Subframe A pixels. That is the perfect setup for eshifting: By eshifting Subframe B those projected pixels fall between the Subframe A projected pixels just like they do in the upsampled 3840x2160 image. This seems simple yet eloquent.....unless I'm missing something....

@Geof, that would be a pretty bad way to do it. Your idea only makes use of 50% of the 4K pixels. I hope (and I think) JVC has a better method, using some clever resampling.

Anything they do will end up throwing out 50% of the upsampled 3840x2160 pixels. They're only flashing two 1920x1080 subframes per frame (120Hz) so by definition they're throwing out one-half of the 4K upsampled pixels (however with eshift they're also projecting twice as many pixels since they're flashing two 1920x1080 subframes/frame). Their downsampling algorithm may be doing more that just using every other pixel but however they derive those pixels it seems that eshifting is very suitable with the odd/even scheme I outlined above.


EDIT: By saying "throwing out" I've implied they make no use of the pixel data. They may in fact be using info from all 8 Mpixels to derive each subframe -- we just don't know. They are however only projecting 4 Mpixels which are somehow derived from the upsampled 8 Mpixel image.

They clearly have to be discarding one half of the upscaled pixels. There can be no question of that. They haven't said that directly but they were very very clear that only two different 1080p frames are being flashed for each original 1080p frame and that each of those two frames were flashed twice. ABAB with the B frame e shifted. At the Cedia Expo, the presenter said that the A frame was the original 1080p source frame. I question whether that statement was correct. I just don't know but whatever they are doing it certainly does the job presenting a more solid or firm image with greater 2D dimensionality. That's my story and my observations, I am sticking to it (the story) andthem (the observations).

Now it would seem that the discard of half the pixels is really not a total discard of half the pixels. With the overlap method they said they are employing I would think that the discard is 1/2 the total area of the pixels using portions of all the upscaled pixels to creat the two 1080p frames. But to me who really doesn't know, it would seem that frame A would not be the original frame but a creation based on the overlapping.

I really am quite naive technically and non expert on all of this. I yield to the experts, psuedo or otherwise.

Now I'm confused. They clearly say eshift is happening at 120Hz. If the source material is 1080p/60 they cannot be showing ABAB or the movie would end up twice as long..... If the material is 1080p/30 this makes sense. For 1080p/24 they may be doing as Ron described above..

Geof. I really don't know but during the press conference and in printed JVC materials they show an ABAB flash pattern for each original 1080p frame. The graphic showed 4 quadrants with I believe top left being A, top right being B, bottom right being A, and bottom left being B.

Hey Mark,
That sounds weird....is that material still available to look at?

They gave out the press packet as a flash drive. I just checked it and it contains only pictures of the products and the press releases, not the graphics from the press presentation.

To be clear, I do not think they said the frames were placed into four quadrants but there was an implication that four frames of 1920 x 1080 were flashed giving you in total the number of pixels in a 4K 3840 x 2160, which is certainly true but it was equally clear to me that they said there were flashing only two different flames, twice sequentially, ABAB with the B e-shifted. This troubled me intellectually, and several here tried to answer me in this thread. I still don't really get it but I trust that the overlap described by JVC and explained by several posters in this forum takes care of it.

Guys. There is a BIG difference between simply throwing pixels away and between calculating new pixels based on the full upscaled 4K frame. Claiming that they will have to flat out throw away half of the pixels is just plain incorrect. Basically you're claiming that they must use Nearest Neighbor downsampling, which is the worst downsampling algorithm on the planet.

I'm not claiming anything. Rather then get bogged down trying to figure out (and argue about) JVC's downsampling technique I chose a rather straightforward "use every other pixel" example. But, we're getting into semantics here. Given JVC's description of how eshift works only 1/2 of the upscaled pixels can be projected - the other half is not projected. For all we know they may use any or all of them to generate subframe pixels (we do not know) but in the end only 1/2 of them are projected - the other half being discarded (after they are or are not used to generate subframe pixels).

Well I understand the concept of the overlap creating 4 "sub pixels" when both subframes are considered thus it effectively projects 4*(1920x1080) sub pixels. I do not understand how they could flash each source frame 4 times with 1080p/60 material because the panels are being driven at 120Hz. The panels would need to be flashed at 240Hz if each subframe were flashed twice with 1080p/60 material. The ABAB routine works with 1080p/30 material though. Maybe I'm missing something but it seems to me that it's AB for 1080p/60 and ABAB for 1080p/30 and possibly ABAB @ 96Hz for 1080p/24 (as Ron posted).

Gentlemen, the 4K technique will obviously decrease pixel structure in close seating distances, and Mark had noted his impression of the picture having a more solid dimensional look. Did any of the demo material viewed provide any evidence that one of JVCs weaknesses, motion blur, was eliminated or at least equal to the Sony lcos panel?

From Phil Hinton's preview: "Looking at the projected image up close you could hardly see the pixel structure and the X70 handled motion better than any D-ILA projector we've seen before - in fact it reminded us more of DLP".

I agree they are flashing ABAB. If they do support FI in 4K-lite mode then I would assume they would be fashing ABA'B' (where the A' and B' are from the interpolated and scaled 4K frame).