Meant to be Seen

A good read and an answer for those that are concerned about motion blur on the Rift.

http://www.roadtovr.com/2013/03/31/gdc- ... lides-4415

The short answer

[youtube]http://www.youtube.com/watch?v=PvkzFhowItM[/youtube]

Thanks for the link. I really should keep my eye on that site more often.

This is a great presentation, it's something everyone with the slightest interest in VR owes it to themselves to read. Abrash is clearly several light years above my head in his understanding of all this stuff, and yet this was the first discussion of these topics that never completely got away from me. One of the biggest challenges for VR right now is just getting everyone who wants to participate in such discussions talking the same language, educating people to the point where they avoid proposing the same obvious but unworkable solutions, arriving at a place where real work can begin. More than any other single source I've seen (MTBS being disqualified as a vast collection of sources), Abrash has here laid out a common framework for everyone to build upon.

jesus now we need 1000hrz display lol or some type of laser projector

GUNGRAVE wrote:jesus now we need 1000hrz display lol or some type of laser projector

Forth Dimension Displays seem to be on the right track, for a price...

It's extremely interesting how big of a difference there is between having an LCD monitor sitting at your desk, and having one strapped to your face, which isn't intuitively obvious. I can't help but to wonder that while our mind might be able to blur the image, if the underlying mechanism that enables it also contributes somewhat to nausea or discomfort? Michael's comments also makes it obvious that LCDs are not a long-term solution as they'll likely never reach the required performance levels. Another thing that's interesting is that even at a super-high resolution, if the refresh rate stays at 60Hz, then so will the amount of perceived blurriness - only slow movements can remain sharp. 120Hz displays can only theoretically cut that down by half.

Going further, lets just say we actually had an ideal display as well as robust and high-performance 6DOF tracking. The limiting factor now is software performance. There has been talk elsewhere on this forum of a sort of rendered frame post processing (interpolation, stretch, skew, transform etc) which, while not ideal, may be one of the few options available to keep the display "fed" if the actual renderer is too slow to keep up. But in reality, an ideal display is ways off, as this brings up the issue of bandwidth. Even if we could send frames down current display links (DP, DVI) at say 120 or even 240Hz and have post-processing done on-board the HMD, there's still the link between the on-board processor and display to contend with. Since an ultra-high bandwidth bus needs to be devised anyway, maybe it's best to have the software and hardware based around a data-intensive (relatively) low processing-intensive paradigm, rather than the current method. This probably sounds crazy as modern games already take up tens of gigabytes, but I'm thinking of graphics that are largely pre-rendered in the form of spacial light fields that would likely require TERABYTES for an entire game. At any rate, I fully agree with Michael that very high-quality VR will take numerous companies and large investments to make happen on every front. Commodity hardware from the mobile industry will only take it so far.

Has anyone tried pushing their HMD's LCD at 75hz and strobing the backlight matching the refresh rate?

Inscothen wrote:Has anyone tried pushing their HMD's LCD at 75hz and strobing the backlight matching the refresh rate?

If I understand Michael's presentation correctly, then strobing the display at such a low frequency might lead to doubling of the image instead of blurring, which is likely worse.

So illuminating each frame and blacking out while the pixel switches at 75+hz is worse than say the rift at 60hz?

Inscothen wrote:So illuminating each frame and blacking out while the pixel switches at 75+hz is worse than say the rift at 60hz?

I have no personal experience with this, so I can't really say with any certainty. I just based my response off my understanding of that presentation, particularly when he talked about displays with low persistence.

Regardless, increasing the refresh rate would be of benefit whether you strobed the backlight or not.

Thanks MSat, Michael said pretty much the same as you

Here's what Abrash said:


MAbrash says:
April 3, 2013 at 7:57 pm
Excellent questions!

The higher the refresh rate, the better.

The shorter the persistence, the better for whatever your eye is tracking – less blurring – but the worse for whatever is moving relative to the eye – more strobing.

–Michael

I have no problem with motion blur on my HMZ, whats the difference there?

Just curious not saying Rift is a lesser HMD (Quite the opposite)

CyberVillain wrote:I have no problem with motion blur on my HMZ, whats the difference there?

TFT vs OLED

EDIT:
TFT vs OLED??

Did you add a head tracker to your HMZ? On my DIY rift clone when turning head quickly there is a kind of smearing. It's probably because you have OLED screens in your HMZ like PatimPatam said.

If you don't have head tracking, that's part of the reason. The other reason I think would be the lower FOV of the HMZ compounded by the game being rendered at a larger FOV so your eye rotates less to track moving objects.

I get a smearing effect on my Z800's which use OLED. That said, most Z800's actually only run at 30hz (signal is buffered) so I always put it down to that...

Michael Abrash showed a slide saying that "effects of raster-scan (rolling) displays" is one of the problems.
I assume that a laser projector (with moving mirror) is such a type of display.

What exactly are the problems for this kind of displays?

mechamania wrote:Michael Abrash showed a slide saying that "effects of raster-scan (rolling) displays" is one of the problems.
I assume that a laser projector (with moving mirror) is such a type of display.

What exactly are the problems for this kind of displays?

Rolling shutter displays allow "racing the beam", which is a good thing... The problem is more software complexity.

http://blogs.valvesoftware.com/abrash/r ... s-the-eye/

Do you mean that the graphics engine should update for each rendered line (or actually for each pixel)?
So, no (frame) buffering?

Doesn't this result in tearing?

mechamania wrote:Do you mean that the graphics engine should update for each rendered line (or actually for each pixel)?
So, no (frame) buffering?

Doesn't this result in tearing?

The screen does not tear because each line is drawn exactly where your brain expects it to be on the moving display while rotating your head (using head tracker data to position the line). However, a screen capture may look like vertical lines are slanted, but only because that screen capture is not viewed while moving your head the same way.

The effect of this is a drastic reduction in perceived motion blur that results from dragging an image with you while it persists on the display until the next screen refresh. Or at least that is how I understand it, without having actually experienced it myself.

In essence, "racing the beam" attempts to duplicate the effect of analog video cameras and analog CRTs, where the visible beam is a real-time copy of what the camera is seeing at that place at that linear time (i.e. raster beam scanning).

Ohh wait,

I guess it is the opposite. Because of the screen is build pixel by pixel and not persistent you can make the ideal motion artifact free
display by updating the contents per pixel. Wright?

So the ideal system is some ray-tracing engine combined with laser projectors.
Unfortunately you can not doe massive parallel ray-tracing, because you have to update the graphic contents for each pixel!

Wright?

Yes, laser projectors do act like CRTs in that they do raster scanning, which can always show the current view and the current beam position, even when the head is moving. This can eliminate perceptual motion ghosts, just as Michael Abrash described.

When thinking longer about it, I guess there are still motion artifacts because of the scanning.
I think objects moving from left to right will look longer when scanning goes from left to right and shorter when the object
moves from right to left.
And when the motion speed is equal to the scan speed (and in same direction), the object may be infinitely long or disappear completely!
That may cause serious problems!
Same kind of problems occur for vertical movements I guess.