What is the real resolution to go for with the Varjo Aero

[darkman222]

The Aero is 2880x2720 per eye on the specs page of Varjo.

The Aero is 1200x1200 / 2192x1880 if I trust Varjo Base Software.

And the Aero is 4148x3556 per eye if I trust Steam VR.

(screenshots)

When I use mbucchia OpenXR toolkit, I can read out the same resolution as steam vr, and when I reduce the resolution in steamVR the openXR toolkit will show the same resolution.

So I conclude that the "real resolution" the Aero runs is 4148*3556 per eye which is much higher than advertised. Of course you'd need to render a higher resolution of the image but this is way higher than the physical resolution the Aero seems to run.

As the real render resolution it runs is the exact multiplier value of 1.89 horizontal and vertical resolution, I wonder where does it come from that Varjo considers this to be 100 % resolution.

This is an exact multiplier of 1.89 compared to maybe 1.2 I would use with the Pimax to make a nice image.

Has someone an idea where these values come from?

Edited March 20, 2023 by darkman222

[Supmua]

There’s a link from mbucchia’s Varjo Foveated site to Varjo’s doc on resolutions, and it’s complicated because Varjo HMDs offer stereo as well as Quad view modes. For VR in general you will see much better visuals by upsampling over the physical screen res, so I’m not sure if there is any true 100% res for any headset. 
I still lack full understanding of how the quad view works so can’t help you there, but 39 PPD is only achievable in this mode with helps of DFR. 
The current Highest setting in stereo view mode has 4.1K pix but you can also go even higher via upsampling. From my experiences the point of diminishing return in visuals starts there but you can still see better sharpness up until around 4.5K.  Of course running at that res is ridiculously taxing and probably doesn’t make sense right now. 

[darkman222]

Okay thanks for the hints. I have to look into that quad mode too. I just forgot about it as I used the openvr_api.dll that basically disables quad view.

But my point is, how can Varjo talk about PPD aka pixels per degree if they start upsampling or you may call it oversampling an image? I mean, the headset has a physical given amount of pixels it can display.

If you choose a setting thats below this physical resolution it will lower the PPD. But if you go higher than the physical resolution of the headset you dont increase the PPD. Only the image gets clearer by rendering at a higher resolution, smoothing pixels out.

Found this in another thread:

null

Compared to the technical specs somewhere between medium and high is where the real resolution of the Aero sits.

Talking about a higher PPD is only done by rendering a higher resolution that smoothes out the image from there on if I understand correctly.

Very high and high already are higher than the Aero can display.

 

But if the aero cant display more than the high setting, why was the the decision that highest is the best achievable quality. this could go on forever I mean. Just to throw stupid numbers, but why is not there an "ultra" setting with 70 PPD at 8296x7112

Of course no GPU can process that, but with that definition in mind there is no upper limit.

Edited March 21, 2023 by darkman222

[Supmua]

PPD is somewhat of a marketing term but Varjo do give out the correlating numbers at specific pixel settings as seen in those charts, and it gets more complicated after you add DFR into the mix because now the resolution is not uniformed across the entire FOV. 
This is why when someone said headset A is much sharper than headset B even though the physical screen has lower res, you have to also question the pixel renderings that were used. Even headset B can still appear much sharper than the same headset B via supersampling.

Edited March 21, 2023 by Supmua

[dburne]

This type of stuff is why VR in it's current form will never reach mass adoption.

[void68]

Each screen has to be fed with approx 1,4 times (depends on the model) oversampling of the native resolution to neutralize barrel / pin cushion effect due to the lenses' distortion.

Edited March 21, 2023 by void68

[darkman222]

1 hour ago, void68 said:

Each screen has to be fed with approx 1,4 times

Yes, thats where I am pointing my question into. Looks like the Aero does 1.89 times in the highest preset

[Supmua]

19 hours ago, void68 said:

Each screen has to be fed with approx 1,4 times (depends on the model) oversampling of the native resolution to neutralize barrel / pin cushion effect due to the lenses' distortion.

 

This may be true with fresnel lens but not sure with the current pancake or aspheric lenses because they all have different properties. For example, the current pancakes while provide great optical quality suffer from low FOV and significant light reduction compared to other lens types.  Aspherics on the other hand tend to have more prominent geometric distortion and chromatic aberration. 

[darkman222]

Okay, but to sum it up and for my understanding. The upsampling of 1.89 times higher than the Aero physically can go is an imaginary value some of the technicans at Varjo found to be the best value to get the sharpest image out of the Aero while keeping the resolution at a reasonable level. Could it be put that way?

[Supmua]

4 hours ago, darkman222 said:

Okay, but to sum it up and for my understanding. The upsampling of 1.89 times higher than the Aero physically can go is an imaginary value some of the technicans at Varjo found to be the best value to get the sharpest image out of the Aero while keeping the resolution at a reasonable level. Could it be put that way?

Yes, although some users on Varjo Discord are now requesting resolution even higher than what’s currently available because there’s so much performance gain via DFR.  This might just be what the Aero needs to stay competitive with the Pimax Crystal.

[dburne]

One thing is for sure - Dynamic Foveated Rendering with Eye Tracking in the Aero with DCS = pure bliss.

[darkman222]

Thanks for pointing that out @dburne. I was about to miss DFR while experimenting with Fixed Foveated Rendering.