Tech Talk #6 : The Key Elements of VR HMD Clarity

General discussion
23

The point of a properly constructed eye exam is that your mind can’t just fill in the gaps with accuracy. If you can subvert the test like that, then it’s just not a good and reliable eye exam. Which admittedly not all of them are.

Actually, part of my point is to include the software and settings into the comparison as well. Because even if the hardware itself is potentially capable of better, the software part of the product may not be reaching that potential. There are a lot of ways that can and does happen. And I think it’s meaningful to also include how software algorithms and settings impact those results as part of the total package end to end.

For instance, I found in testing the 8KX against an eye chart that I could keep increasing its resolving power by increasing super sampling well beyond resolutions that could reasonably be run in a game on even the fastest modern GPUs. Whereas with an Index, its resolving power would max out and see no further gains at resolutions my 3080 Ti could easily keep up with.

I used this kind of approach to tune my settings on the 8KX objectively to get the most resolving power out of limited GPU capacity.

An upshot of this is that the faster your GPU is, the higher clarity you can have on your 8KX. Which does mean that people running 8KXs on slow GPUs and/or on systems which are not well tuned are not experiencing the same product that people with fast GPUs and good tuning are. And I believe this is one of the reasons why there is such large disparity in reviews of the 8KX.

If you looked at the 8KX hardware alone, you’d get misleading results in terms of the clarity the customer will actually experience with the product. Pimax’s software also matters. As does software (such as nVidia drivers and SteamVR) which Pimax has no control over.

Yeah. I don’t believe that claim either.

I am highly skeptical. If you were able to read the 20/20 line with an 8KX, I question whether the construction of your test was able to produce valid results.

I searched for and found your post on the subject. I don’t see anything obviously wrong in what you’ve posted about the methodology that I am able to identify, but I think the method is fraught with danger of getting some aspect of it wrong either in implementation or misunderstanding how the chart needs to be used. And since the result you got appears to be bogus, I think that’s what has happened.

I’ve never seen anyone else claim 20/20 vision on an 8KX or any other VR headset besides maybe the Varjo headsets (which might just be marketing claims). I’d want to see your test results corroborated by credible 3rd parties.

It’s obvious that the 8KX degrades sharpness significantly relative to normal vision outside of VR. So how could you be getting a valid 20/20 result?

3 Likes
24

Cool. Indeed as you suggest, the testing methodology was complicated enough that it was fraught with risk of misinterpretation. However, I am absolutely confident in my results. It was not something I did in just a few minutes, but IIRC, took many days to arrange, and checking methods and results multiple ways for extra certainty of plausibility. I found no inconsistencies, and I 100% stand by my claim of 20/20 readability.

I think it’s worth pointing out that exhaustive due diligence is basically what I have been doing for several years now, so I can do this kind of stuff.

Where I think there is room for interpretation is just how barely readable it was. Technically I could read that line without errors. But it was marginal, and anyone with that quality vision IRL would definitely be using corrective lenses most of the time. I myself have glasses to correct from my almost 20/15 vision to nearly 20/10 .

25

Because 20/20 is very average vision, and 20/20 readability is not the same as ‘without artifacts’. Just because it is technically possible to read the same line, does not suggest there is not a very substantial difference in quality.

IIRC, I remember looking at the Snellen chart and thinking someone with something like 20/35 vision would not notice the difference of the Pimax Vision 8kX vs IRL. IIRC I think I mentioned something like that.

But the Oculus CV1, Vive, and such, were definitely only something like 20/40 readability at all, absolutely so terrible it would have been unusable for such things as Virtual Desktop if any worse, and a severe eyestrain to use for any amount of time.

26

According to wiki (Snellen chart - Wikipedia) the 20/20 (or 6/6) acuity means a person can read (distinguish) a character of the size of 5 arc minutes (height x width), where the individual lines are 1 arc minute wide.

Considering 8k-X and its PPD ~25 (estimated), 5 arc minutes will constitue 25 / 60 * 5 = 2 pixels per character width (or height) or ~4 pixels per character. I would argue that it would be difficult to read.

Using a term acuity in regard with VR headset may be misleading too. The acuity defines an ability to focus at a certain distance, and it is known that the ability to focus depends on the distance. However in VR, each headset is designed with different focal distance and it is usually far less than 6 m (20 ft), which means that the eye, which would in real test show worse than “normal” acuity, can still read the image in an HMD perfectly fine.

In other words, people with different real world acuity may “measure” the same acuity in simulated Snellen test in VR scene, and vice versa.

4 Likes
27

With VR headsets, the combined effects of supersampling, dithering from any slight head motion, and stereoscopic vision, arguably might improve observed resolution slightly above native panel resolution. But difficult to read, yes it was.

By the way, your expertise is always appreciated @risa2000 .

2 Likes
28

Yes, this the only possible explanation.

1 Like
29

I think this may be the key factor. If you have above normal human vision, then it may make sense that the VR headset is degrading your vision to merely human norm (20/10 → 20/20). Whereas for someone who has normal 20/20 vision in the real world, the VR headset would degrade their results to something less than that such as 20/20 → 20/40, for example.

I’m pretty sure that the nature of this is that everything in the optical path (which in the VR case includes part of the pipeline in software before it is even physically displayed on a panel) is cumulatively decreasing acuity. At best, any part of it can only not make acuity measurably worse.

So in your case, the total of adding up those cumulative degradations including your own eyes is able to read the 20/20 line on the chart.

This seems to really be getting deeper into optometry, and I am not an expert. I suspect neither of us are? I think what may be happening here is non-experts in the field trying to interpret what being able to read a line labeled as 20/20 means in this context, and our interpretations especially in such a novel case are probably dubious.

To be sure, when I’m talking about using eye charts in VR, I mean using them to objectively determine relative differences between VR headsets along with the effects of configuration changes and software updates within those products. It seems rather like measuring horsepower with dynos. Even experts will argue endlessly over the absolute interpretation of dyno readings. But they are considered reliable in terms of measuring relative power increases (and decreases) when comparing readings back to back on the same dyno.

1 Like
30

I was able to read the 20/20 line in these experiments without the corrective lenses (ie. glasses/contacts) that would have given me substantially better than average vision. Though the slightly nearsighted focus of the headset does match my eyes rather well, there is still a noticeable difference in comfort and clarity if I do use those glasses.

So no, my better than average vision was not much of a difference for this, not more so than for something like the 40th percentile for human vision, or most anyone else using their normal glasses/contacts.

31

Well as he did say it was not easy to read it. This is where our brains adaptive interpetation comes in. If you can recognize the basics of letters your brain can pretty accurately read something not completely clear.

His test is good but to quantify the results it needs to be conducted on a range of different individuals using similar hardware and software settings.

32

No, I was not relying on that. I could clearly discern the features of each character. I remember the chunkiness made it take longer to read than normal. As I stated, readable, but with severe artifacts.

33

As I said without a group to test with there is potential bias. Many ppl discern things without needing to be consciously aware as there doing fairly instantly.

One study I have read said you need about 6000 pixels wide and a bit over 3000 pixels in height to have around human clarity.

Reality 12k or even the Crystal with the right set of lenses should indeed come close.

As Joshua said in the presentation full is around 60ppd the Crystal having 42(75%) and the 8kX with other simularly spec’d headsets around 33%.

1 Like
34

Proper eye test charts use only specific letters that are difficult to distinguish from specific other letters by using such clues when you can’t discern the fine details. For example “P” and “F”. Or “O” and “C”. Some eye test charts don’t use letters at all, but rather symbols in order to avoid any possibly of interpretation filling in what can’t actually be distinguished by vision.

2 Likes
35

Apart of mentioned points clarity is affected by ability of a hmd to resist fog, it’s very annoying to constantly clear lenses. Plus you wear it’s surface and they become more blurry over time. And as other mentioned If image is clear only in static like 8kx , it’s a severe flow, as during dynamic gameplay you won’t have any clear picture. Lenses also suffer from good rays, chromatic aberration and other issues which make image less pleasant and clear. Wispy should be clear in both eyes at once. And modern hmds often have awful sweet spot which makes even a good panel to look like a garbage one.

1 Like
36

Now, chromatic abberations, specifically (along with overall distortion), is something that can be software compensated for much better if one have a higher resolution headset than a lower res one, even when using low resolution source frames…

2 Likes
37

Technically, more PPD = better detail in the final image if you also have good lenses, where good means they can focus on the image (pixel) at least in some (central) part of FOV, and also have good eye box. With a poor eye box a slight shift of the eye may dramatically change the optical properties of the lens, chromatic aberration included then no matter how good is the warping algo the image becomes distorted.

3 Likes
38

I was happily plodding along enjoying my 1080ti and then had cataract surgery. Now i see so well i had to upgrade my gpu :grinning:

3 Likes
39

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.