SDE and PPI - Pimax vs Valve Index vs Others

SDE and PPI - Pimax vs Valve Index vs Others
0

images%20(25)
images%20(26)

Did you wonder about these images and why Pimax headsets offer comparable resolution to the Vive Pro and wonder, why does the XTAL have so much less SDE as the Pimax if the screen resolutions are the same? Pimax 5K+ is RGB LCD and so is XTAL, so why so much more SDE on Pimax?

Well, Pimax magnifies it’s screens and only renders to three quarters of each screen, the lenses move to adjust IPD and a digitally rendered image moves across the screen in correspondence with the lens positions (see video here).

You might also be wondering why Valve Index chose to use such low resolution displays, I was too until I saw this thread. It’s clear now, it’s an RGB 1600x1400 (3K combined) set of screens, not a Pentile 1600x1400 set which the vive pro offers.
The fact is the Valve Index will likely have one of the least amount of SDE on any headset; here’s why:

Due to the fact Pimax headsets only render to 75-77% of the panel, I made the following graph with PPI measurements instead of PPD.
The rendering pipeline does not lessen pixels used in the useable area of the screen, and as each HMD has a different FOV we can get a rough idea of the density of pixels and SDE in the upcoming headset by looking at the PPI of each headset relative to its diagonal screen size.
(Not factoring in scaling as this discussion is talking about physical screen res)

Rendered subpixel amounts affect the visibility of SDE. In general, the higher the PPI the lower the visibility of SDE, noting that rendered subpixels are spread over different sized screens and FOVS.

PCVR HMDs ranked most to least SDE by PPI:


9. Pimax 5K XR

Pixel density - Pimax 5K XR

  • 5K XR 2560 × 1440 / 5.5in (diagonal of screen) = 534 PPI (in the rendered part of the screen)
  • (Effective ppi of the 5K XR) is 5K XR PPI /3 x 2 because pentile subpixel arrangement, 356 PPI).
  • Due to the fact the screen only renders 3/4 of the images, they must be magnified by 1/4, 356 is divided by 100 and multiplied by 75 to make a final result of 267PPI for the 5K XR.

Rendered subpixels - Pimax 5K XR

  • 5K XR (2560 x 1440) x 2 (OLED Pentile) = 7,372,800 subpixels per eye.
  • Same as the 5k+, divide the result by 100 and multiply by 75 or 80 to take into account for panel utilization
  • = 5,529,600 - 5,898,240 visible subpixels per eye on the 5K XR.

8. Vive Pro

Pixel Density - Vive Pro

  • 1600×1440 / 3.5in (diagonal of screen) = 607 PPI
  • (Actual ppi is /3 x 2 because pentile subpixel arrangement for a final value of 404 PPI).

Rendered subpixels - Vive Pro

  • (1600x1440) x2) (OLED Pentile)
    4,608,000 subpixels per eye.

7. Samsung Oddessey+ (Anti SDE)

Pixel Density - Samsung Oddessey+

  • 1600×1440 / 3.5in (diagonal of screen) = 607 PPI
  • (Actual ppi is /3 x 2 because pentile subpixel arrangement for a final value of 404 PPI).

Rendered subpixels - Samsung Oddessey+

  • (1600x1440) x2) (OLED Pentile)
    4,608,000 subpixels per eye.

6. Pimax 5K+

Pixel density - Pimax 5k+

  • Pimax 5k+ 2560 × 1440 / 5.5in (diagonal of screen) = 534 PPI (in the rendered part of the screen, which is a magnified 534 PPI.)
  • Due to the fact the screen only renders 3/4 of the images, they must be magnified by 1/4, 534 is divided by 100 and multiplied by 75 to make a final result of approx. 400PPI. This is why the Pimax has the same unmagnified PPI as XTAL, but more SDE.

Rendered subpixels - Pimax 5k+

  • RGB LCD has three discrete subpixels per pixel, or ((2560x1440) x3) per eye, or 11,059,200 subpixels per eye.
  • However ; due to the lack of complete panel utilization (seen in the above video and teardown) the final result 11,059,200 needs to be ÷100 x 75 or 80 to accomodate for the fact that only 3/4 of the panel is rendered to on pimax headsets, the lenses magnify the image and only the lenses move during IPD correction and the screens remain fixed in Pimax headsets, the software only renders to half the available screen and keeps. the rendered image in line with the lens position.
  • Final 5k+ rendered subpixel = approx
    8,294,400 - 8,847,360 visible subpixels per eye

5. 8KX

Pixel Density - Pimax 8K/8KX

  • Pimax 8K / 8K X (3,840×2,160 / 5.5 = 801 PPI )
  • (Effective ppi of the 8K and 8KX (if the 8KX uses the same screen as the 8K) is 801 /3 x 2 because pentile subpixel arrangement = 534 PPI.
  • 534 PPI Magnified… /100 ×75 Also approx. 400 PPI.

Rendered subpixels - Pimax 8K/ 8KX

  • 8K (3840×2160) ×2) (Pentile LCD ) = 16,588,800 subpixels per eye
  • Same as the 5k+, divide the result by 100 x 75 or 80 to take into account for panel utilization
  • 12,441,600 - 13,271,040 possible visible subpixels per eye with native 4k input on 8K X using 8K screens.

4. StarVR One

Pixel Density - StarVR One

  • (1,830 × 1,464 / 5.5 diag = 426 PPI)

Rendered Subpixels - StarVR One

  • AMOLED RGB has three discrete subpixels per pixel, or ((1830x1464) x3) per eye, or
    8,037,360 subpixels per eye.

3. XTAL

Pixel Density - XTAL

  • (2560x1440 ) / 5.5in diag )= 534 PPI.

Rendered Subpixels - XTAL

  • OLED RGB has three discrete subpixels per pixel, or ((2560x1440) x3) per eye, or
    11,059,200 subpixels per eye.

2. Valve Index

Pixel density - Valve Index

  • Valve Index - 1600×1440 / 3.6in (diagonal of screen) = 598 PPI

Rendered subpixels - Valve Index

  • RGB has three discrete subpixels per pixel, or ((1600x1440) x3) per eye, or
    6,912,000 subpixels per eye.

1. HP REVERB & Acer ConceptD OJO

Pixel density - Reverb & ConceptD

  • (2160 x 2160) / 2.89in )= 1057 PPI.

Rendered Subpixels - Reverb & ConceptD

  • LCD RGB has three discrete subpixels per pixel, or ((2160x2160) x3) per eye, or
    13,996,800 subpixels per eye.

@anon74848233 @Sean.Huang @Pimaxusa @Sweviver Please consider incorporating this mod into 8KX’s design to make moving screens and lenses a possibility, to get extra panel utilization to increase binocular overlap, and minimum IPD down to 55mm/65mm from 60/70 and screen usage to 100%, as StarVR and XTAL have shown it is possible to have moving screens and lenses in HMDs with cantered displays. (Maybe reducing the cantering from 15 degrees to 5 degrees is required)

12 Likes

What’s interesting is the pimax could get way better results by just licensing the xtal lenses and they wouldn’t even have to change the panels if they just licensed the lenses and have them produced to the specs they need to fit the panels already in the 5k plus and you’d get near xtal results. They could probably release a 5k X model with the better lenses that use all of the panel and magnify less or users could replace the lenses themselves via a kit they buy. They wouldn’t even need to touch the panels to keep the cost down. Since the screens are basically the same size, you wouldn’t even have to have the lenses adjusted that much in production. Of course then xtal might not do such a deal as then nobody would buy their super expensive headset. Having said that, I don’t think pimax only uses half of their panels, it’s gotta be more than that!

2 Likes

Something that costs 5x the price is better, what a surprise really…

1 Like

Umm, i see way better the IL2 dials in my pimax 5k that in that photo

and i had a samsung oddisey, that every know, has the vive pro scren and pimax 5k is very, very, very far in resolution and sharp image

anyway, great post! thanks!

5 Likes

I don’t think it’s a truthful representation of the difference between Pimax 5K XR, 5K+ and 8K. It looks 5K XR is clearer than 5K+, and 5K+ is clearer than 8K, but my test is the opposite. 5K+ is clearer than 8K: 5K+ or 8K? My test of M2 through pictures of Aerofly FS2, Project Cars, and Virtual Desktop (Update with pics of extreme setting on floor 28)

3 Likes

I didnt include Pimax 8K because the 8K has scaling problems due to it being upscaled from a resolution which isnt in a good ratio for upscaling to 4k, as it’s not a 1:1 upscale (1080p would be)

In terms of SDE
5K + > 8K > 5K XR

5K+ has PPI of 534, same as 8K, but 8K would have 801 PPI if it was not pentile.
After pentile adjustment 8K is also 534 PPI the same as the 5k+(801/3x2), which is why 5k and 8k look so similar.
But both the 5k+ and 8K are magnified so the true PPI value of both the 5K+ and 8K is between 400-411PPI, depending on whether you go by either 534/100 ÷ 75 or 77 to take into account the magnification and the used screen real estate (refer to the link in the op)
Thats why they also look so similar to Vive Pro’s SDE.
5K XR has 356 PPI, or 267PPI after taking into account magnification, so it’s worse than the

Original Vive and Oculus Rift both had 430 PPI at 1080×1200 / 2.5 inch panels, but only 2,592,000 million sub pixels spread over a small space, which obviously as the sub pixel amount increases the PPI increases, which makes SDE less visible, but a larger screen size/FOV/resolution makes the PPI less.
It’s a higher value PPI in this case compared to the Vive Pro or the 5KXR, but it has more SDE because the resolution was lower and the screens used were smaller.
All the headsets above have similar resolutions and screen sizes, so PPI is a good reference point in addition to total subpixels per frame, per eye.
Determining SDE from subpixel count alone can be unreliable as well though, as Vive pro has less SDE than 5KXR, but has less subpixels, even after factoring in the renderable part of the screen in the XR.
In cases where screens used are if similar or the same size (Vive pro and Index… or Pimax, Xtal, StarVRone) PPI is a useful metric.

Are you on crack? I tried the 5kxr at a backer meetup and the sde was close to the regular vive. It was in no way clearer then the 5k+

3 Likes

Would love to see this image BUT with the reverb, index and rift s also compared!

1 Like

Think you need to lay off the LSD. He said the pics the OP posted make the XR look better than the 5k+.

Which is true the pics are poor representations.

5 Likes

why is the 5k+ not as sharp as before compared to the 8k? doesnt look the same

1 Like

@Mantidtings I believe that (with all the respect I have for @SweViver) the video is confusing two different things. What is rendered by the app and what is displayed on the panel.

What is rendered by the app is rendered at the render target resolution and there are no black margins on sides as the video suggests. The only masked area (hidden area mask) is in the corners of the rectangle. What is displayed on the panel is another story as there are other aspects playing the role in case of Pimax (FOV config, IPD accommodation), which are not present in Xtal.

If you want to see what is the difference between the rendered and displayed image check this post for an example and an explanation (How the render target resolutions and the supersampling factors of Pimax and SteamVR work together).

The reason why you do not see the real image displayed on the panel in the video is that they are not so easy to get. They are however not what SteamVR or OBS capture shows you.

Now coming to your reasoning about the magnification:

The magnification (and the optics in general) is set in a way that it can deliver the required FOV. In particular in Pimax, the percentage of the panel utilization changes dramatically with the selected FOV. If you apply your reasoning to the different FOV configuration you would have to conclude that the magnification of the Pimax changes for different FOV. Which is impossible, because the optics does not change.

If you want to make any meaningful comparison of the perceived visual quality, you need to consider the PPD (pixel per degree), which means to know exactly the FOV and the resolution at the same time. Calculating (measuring) of PPD is a bit more difficult though (Pixel per degree (PPD) of Pimax 5k+).

If you would want to make the comparison by the screenshots you need to guarantee the same conditions for all, which basically boils down to the very simple requirement - the object displayed (in this case the altimeter gauge) has to be observed at the same angular dimensions in all pictures taken.

So while the pics you posted are authentic I would simply question if the condition was met as there is no way to tell from just looking at them.

5 Likes

If we want those result to have more meaning a specialized application must be built for this purpose only. A few degrees in the angle of view or a few centimeters of distance vs the target can the blur the result between 2 ( or 3 :wink: ) similar headsets and I didn’t event talked about the numerous setting and possible combinations.

From the presented pictures we can only assert with that the Xtal must better than the others

It’s difficult to take good (representative) pictures “through the lens”, so variation is expected. Remember too, these are zoomed shots to see the pixels. In-game, it will probably look a little better (because eyes are very different from cameras and respond differently to light).

1 Like

There’s the visible part of the panels, which @anon33864247 found out in his teardown being 75%-77% of the panels.

The subpixel size stays constant no matter what FOV is used, the magnification also is a constant amount and does not vary.
The subpixel amount varies depending on FOV (FOV settings change the amount of rendered subpixels, they’re still visible through the lenses), but that’s irrelevant to the figures I presented, as the first PPI figure takes into account the whole screen size, and then alters the value depending on whether it’s pentile/rgb, and then once again to factor in magnification.

Hidden area mask and invisible portions of the screen are irrelevant when we’re talking just about the visible subpixels on the screen (which do not change in size)

Yes the FOV cuts off some rendering in the visible range, but we assume the whole screen is the PPI value and rendered subpixel value, and then factor in visible screen amount to determine Magnification amount (and perceived loss in pixel density in the last PPI value for Pimax, and the visible rendered subpixel amount is also extrapolated from the visible percentage of panel utilization (75-77%)
Magnification affects the perceived pixel density, which is why the PPI value is adjusted once again, with respect to sjef’s findings on visible screen space.

In other words, the magnification amount is constant and the visible subpixels (77% of the screen you can see in the lenses) are the same size as the whole screen in fact; but perceived pixel density is less because the magnification occurs, so the PPI value after magnification decreases the PPI by roughly 1/4, to take into account the decrease of perceived pixel density as a result of magnification.

With regards to the discrepancy in viewing distances to take photos, some discrepancy may exist but I wouldn’t discount any findings solely on that basis as those discrepancies are likely small enough to be attributed to human error and they would not greatly affect the result
(in statistical analysis in journals there is always a few margins of error allowed)

I agree that PPD is a useful metric (but difficult as you need to know panel utilization for all headsets and viewing distances between screens and lenses for each. which is different for all HMDs).
As vertical PPI is the same for pentile and RGB screens of the same size and resolution; Vertical PPI may be useful to compare wide fov and small fov headsets, but I don’t see much point in comparing pentile to rgb if it’s the same measurement of pixels vertically (This is why I think diagonal PPI is useful for comparing pentile and rgb screens with the same diagonal measurement (5.5 inches or 3.5); and for comparing headsets with similar resolutions (in total subpixel amount, … > 5 million - 10 million)
It’s not really useful when comparing a 5.5 to a 2.5 inch screen.

2 Likes

While we know the sizes of the 5k+ (5.5" & 8k 5.7")

The ppi can be calculated based on knowing panel size. Sj says approximately 77% to 79% panel utilization. We don’t I believe know Xtal’s panel utilization or if it’s using any tricks to help eliminate sde(ie film).

Aspherical lenses do however have less distortion than Fresnel lenses and can in that respect have more uniformed magnification to have better clarity due to uniformed focus.

We don’t know the physical screen size is so ppi is not known.

Here is updated specs.

79% was before factoring in vertical usage. 77% after vertical, but less considering the corners aren’t used. We can go with 77 or 75, I chose 75 to calculate PPI of the Pimax in the OP to take into account the corners that aren’t seen, which is 400-411 diagonal ppi on the 5K+ (75-77% of 534) and 400-411 on the 8K and 8KX if it uses the 8K’s pentile screen.

Approximately is the key word. 77% can still be approximately + or - approx 3%

1 Like

There are many imprecision in those “through lens” images. They in no way accurately represent what the eye really sees.

I have WAY better results on my 8K, and I don’t have a killer GFX card, just a 980ti. I was a bit afraid after choosing the 8K over the 5K back then, scared of having horribly blurred text… and I was wrong.

You just need to get the settings right. My take is to get the input resolution as close as possible to the native input resolution of the headset.
I have absolutely no problem in reading the dials in games like Elite Dangerous or Assetto Corsa. They are as clear as if they were on a monitor.

3 Likes

Gauges in my 5K+ and 8K both look like the XTAL in the far right upper METpol gauge HOWEVER I have to assume these thru lens shots are very much zoomed in

2 Likes

Are you blind? Could you finish with the latter part of my sentence: “…but my test is the opposite.”

2 Likes