TestHMD - FOV, SDE, Res, Super Sampling - The Rift(s) against everything else! :-)

Couldn't find any thread about this great app that Sebastian (MRTV - https://www.youtube.com/channel/UC2mgZjuHRDW02mx_ok4wfPw ) lately used to measure FOV on HP Reverb G2 and many more hmds. Here's an intro to TestHMD:

According to the dev, current HMDs have these specs, especially when it comes to FOV (Field of View):

Or zoomed in on the results:

Note that when you're inside the app, the above results are seen on a wall, like this:



Comparing my own Valve Index and CV1, and spending some time double and triple checking my results, I got the exact same results as the above - that is for the Rift CV1 I got 88 degrees horizontal fov and 116 degrees vertical fov. And even using glasses - but Norm from Tested also could dial in the lenses all the way using glasses - I got 108 degrees horizontal fov and 132 degrees vertical for the Valve Index. It does take some practice to do these measurements correctly. I've defined the upper limit as where you no longer can see the stimulus material, for example when going from 114 to 116 using CV1, you can sense the last movement, but none when going from 116 to 118 or above. This takes some practice - and for horizontal fov you may have to look left to best see what's moving to the right, might seem counter-intuitive, but has to do with how our eyes are constructed. 

At least I can confirm the results shown inside TestHMD for CV1 and Index, and my IPD is 63.94 mm (right: 31.23 mm, left 32.71). I don't know how much IPD can or will affect these results. Interestingly MRTV (Sebastian) got quite different results for some HMDs, if he was too fast and forgot to triple check his results or if his IPD caused different results, I have no idea. 

Measured as a square, Valve Index (108 x 132) would have 40% larger fov the Rift CV1 (88 x116), which closely fits my subjective experience. 

Fov isn't the only interesting subject that can be examined and measured using TestHMD - I found the reading and vision tests to be just as important:

This test was kind of an eye-opener to me - because it's a great way to test super sampling and compare different HMDs - like CV1 and Index. How much better is CV1 using 1.0 super sampling (ss) compared to Index using SteamVR resolution 100 %? Or CV1 ss 2.0 vs. Index res 200 %? In the reading test you just focus on the text and move backwards until you just barely can read the text. On the floor you can see how many meters you can stand from the text and still be able to read it. My results were somewhat surprising:

Rift CV1:
Ss 1.0 = 4 meters
Ss 2.0 = 6 meters

Valve Index:
Res 100 % = 4.5 meters
Res 200 % = 6.5 meters

Note that the ruler on the floor only goes to 5 meters, but I'm confident I'm not very mistaken for the results beyond 5 meters. 

I've spent a lot of time checking these results, but they are what I observed. Index res 200% looks a lot better, but I can't zoom out much more than CV1 ss 2.0. Also it helps moving your head slightly using CV1, because it removes/reduces the SDE, while you don't have to do that using Index. Index res 200 % is a total of 18 mill pixels per image, it's quite massive, so very surprising I can't zoom out a lot more... Also note that the TestHMD worked much better through SteamVR, where I could easily use Index res 200 %, but forcing CV1 ss 2.0 was so incredibly gpu demanding that the app once crashed - might be some driver bug (I used the regular version of TestHMD). 

Another interesting vision test is this one, where I went for the red line, namely normal 20/20 vision:

 

In this test you measure the  ability to identify each letter, especially separating the F from the P - and I got:

Rift CV1:
Ss 1.0 = 2 meters
Ss 2.0 = 3.5 meters

Valve Index:
Res 100 % = 2.25 meters
Res 200 % = 3.25 meters

I just wrote down the numbers after each test, but strange that I got a better number for the Rift ss 2.0 here than Index res 200%. Important factors could be better contrast due to oled making black text easier to read, and also I've got the Index lenses all dialed-in, so the pixel sizes may not be very different. That said, the Index looks a lot better due to no SDE and much larger fov. Again CV1 profited from moving my head slightly to remove the SDE, while you don't need to do that with the Index. 

I'm slightly baffled by these results - I would have thought that I could zoom out much more with the Index, but that's not the results I got. I was quite amazed with the CV1's image quality, especially for the readability. Comparing ss 1.0 with 2.0 it's easy to see the large impact of super sampling. This goes for both hmds. Also SteamVR res 100 % may correspond more the "ss 1.4", so it's not fully apples to apples, but I do consider CV1 ss 2.0 and Index res 200% for close to the best image quality you can get with these hmds - further increasing ss does not result in large gains. 

Now it could be really interesting to test Reverb G1 and/or G2 for visual acuity! Or if the above results can be confirmed. Note that I used the SteamVR default settings for antialiasing and everything else - to change res I only used the SteamVR res slider. For the CV1, I used Oculus Tray Tool to change between ss 1.0 and 2.0. Index used 90 Hz and lenses all dialed-in.