Demostrating Pimax 8k has real RGB subpixels with an 4k panel resolution. And explaining the source of its image problems


#1

From SweViver new through the lens 8k vs 5k+ video some people have come to the conclusion that the Pimax 8k does not have real RGB 4k panels. But is not the case. The reality is that we have full 4k resolution with all RGB subpixels BUT in a diamond grid.
The followoing analysis is referred to the figures numbers in the image below because as a new user I can only post one image per post.

CONSIDERING:

  1. The diamond pixel array as shown in FIGURE 1 does indeed have all the RGB colors but they are not spaced equally. Green takes more space between colums and we have red an blue sub-pixels sharing the space between rows. This asimetrical pattern led to some people to believe that 8k did not had RGB panels. But we can demostrate it by superposing a zoomed image from Sweviver video to the diamond pixel array in FIGURE 2.

  2. Columns have more distance separation than rows. This have an adverse effect: you can’t map exactly a square “normal” pixels of an image that have the same separation between columns and rows to diamond pixels. Also this led to believe to some people that 8k did not had 4k panels when measuring pixels linearly (not by area).

  3. To demonstrate resolution: For a 4k resolution panel we should be having 3840 columns by 2160 rows in a normal square grid. In a diamond pixel array like the pimax 8k will have instead 5430 colums by 1527 rows. [that is 3840*sqrt(2) colums by 2160/sqrt(2) rows]. The same amount of RGB pixles (8.3MegaPixels) is equal to any real 4k panel, but distributed in a different grid and different quantity of rows and columns.

THERFORE:
This explain almost all the image problems we are seeing with the 8k. When we try to re-sample (not even upscale) an square source image to a diamond array at the same resolution we have to remap the grid but they have different colum and row locations!
Let’s see a simplified example of the problem in FIGURE 3. If you want to draw just a couple of black pixels (“A” and “B”) from the green “normal squere grid” and draw them in the red “diamond grid”. You either draw it to only one of the closest diamond pixel to have a clear picture but create sever jaggies. Or you can apply a more complete resample with a grey scale to all adjacent diamond pixles and loose clarity (sounds familiar?).
Here is another example is FIGURE 4. Just exactly the problem we see with text in the Sweviver images.

CONCLUSIONS:
The only way for the diamond pixels panel to look as sharp when you try to display an image composed by square pixles is is feeding a much Higher resolution image (higher than 4k), that way, when down sample not much info is lost in the resampling of the pixels. It is the same principle as used by Super Sampling Anti-Aliasing at higher resolutions. Because it is exactly aliasing the consequence of resampling to diamond grid as shown in the FIGURE 5.
With the 8k we have an even worst scenario where the source image even at a lower 1440p resolution. And no upscaler can make up for the missed information to feed those diamond pixels.
In theory, pixels in a diamond shape can give a better “perceived” resolution (more likely a marketing spin). But that is only true if the source image resolution is also higher than the panel to account for the re-sample to different pixel positions.

I anticipate that a Software solution to the 8k lack of clarity would not be possible. The problem is inherent to its hardware characteristic because what is causing the lack or “clarity” in the 8k is the diamond pixels grid.
But please take notice that the examples here are “worst” case (letters and patterns perfectly aligned to a square grid). In practice, especially for moving pictures this should not be as bad.
SOURCE
https://www.kguttag.com/2012/02/09/ti-dlp-diamond-pixel/

EDIT: Got mixed up Horizontal/Vertical Resolution. As per the Image I have recalculated the “real” resolution of the 8k to 5430x1527 in point 3 above.


New through-the-lens 8K vs 5K+ video :)
[Personal opinion] The 5K+ is not better than the 8K. The problem is that the 8K is below the expectations and I think Pimax can correct its blurryness by software
#2

WOW great post, i learn a lot with this, thanks you so much.


#3

It seems like the upscaling causes more issues than it fixes. I think they should keep the upscaler, but add a pass through to bypass the upscaler. That way, once foveated rendering rolls around we might be able to make 4k at 90hz happen. Even if it doesn’t, maybe they can use brainwarp in combination with a lower hz 4k native signal.


#4

Or simply add add another cable.


#5

Diamond pixel grids are a technique for getting around projector limitations afaik they aren’t used in any LCDs as there would be no point.

I’m pretty sure this isn’t correct as it no matter how you rotate the images sweviver captured you won’t get enough sub pixels to match true RGB.

The resampling issue described however does apply to the awkward sub pixels we observed for a slightly different reason.


#6

The problem is also that the 8K has far less pixels than it should have had. And this seems to be because the panel isn’t fully used. So we are now comparing close amounts of pixels just in a different configuration. Where the higher pixel count of the 8K should have won regardless.


#7

Plese zoom in Figure 2 in this post. the super imposed image is from SweViver video. It has ALL RGB sub-pixels in the correct place corresponding to the diamond array.


#8

Don’t think it’s this either. The 8k panel works out about ten percent larger from the visible horizontal resolution. It should still have more resolution in both dimensions but doesn’t.


#9

No, Resolution of the diamond grid should be around 2715 x 3055 (see the post for the math). Similar resolution to 2560 x 1440 in one dimension but more than double in the other dimension.
That couse to some people counting pixels linearly to say 8k and 5k+ had similar resolution.


#10

Which frame is that from the sweviver video? (Where in his video did it come from?)


#11

I’m not fully convinced about what your claiming. The count of subpixels for the same areas just do not match. Try drawing an image that shows what you think one pixel of the 8K panel looks like showing the respective areas of the subpixels.

For me the pattern looks much more like this:


#12

Pretty much what I thought. With the caveat that that is showing three rows rather than 1.5. you get a lovely RGB pattern if you are happy to call it a 3840*1080 screen though.


#13

I fully agree with your picture. You made rectangular pixels instead of the diamond pixels. But your picture is fully RGB because: It has distinctive sub pixels for each color and all three RGB subpixels are present in each pixel. Just like diamond pixels your rectangular pixels the the Space occupied by each subpixel varies and green pixel has more space. Also just like diamond pixels your rectangular pixels are asymmetrical. But at the end if you count them there are the same quantity of red, green and blue in your picture (six each) that is full RGB.


#14


#15

Yeah I think by pulling from the edge of the image the focus is going and making you see additional blue sub pixels that don’t exist. Grab from the middle or one of the more zoomed images and you can see it’s more rectangular.


#16

If you define one pixel as a rectangular area spanning over three subpixel either horizontally or vertically than each pixel might have 3 subpixel, but the count of pixels is too low either horizontally or vertically.

Following picture from the other thread shows this:

So in this case one pixel has one green subpixel + a red and blue subpixel below. The horizontal resolution is now as expected. However the vertical resolution is half of what I would expect in such a case.


#17

It would be if there were enough vertical sub pixels to make up the expected resolution but if you count like that you get 14 vertical pixels on the 8k to 18 on the 5k. The ratio between the horizontal pixels is correct though so it’s not a screen utilisation difference. The only way the 8k screen could have every sub pixel it should is if it’s a weird stretched display with a 16:18 aspect ratio but 3840*2160 stretched over that.

Which would be silly.

Edit: ninjaed


#18

If the 8k had a proper RGB stripe then all those red and green dots would be a beautiful white entire pixel. Imagine how great that would look. Sigh.


#20

i dont know if it makes any sense but i like your graphs.


#21

To add another cable you need another bridge chip.