1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Displays The VR thread

Discussion in 'Hardware' started by Parge, 10 Apr 2013.

  1. edzieba

    edzieba Virtual Realist

    Joined:
    14 Jan 2009
    Posts:
    3,179
    Likes Received:
    284
    Got it in one.

    Important things we don't yet know, and are very important to know BEFORE making a decision:
    - Pricing of both systems
    - Required specifications for the Vive (if you don't meet the minimum requirements, then don't bother getting the HMD. Just don't. You'll have a terrible time. Upgrades come before HMD)
    - Launch lineups for both

    This simply isn't true. The Rift uses the same sort of outside-in optical marker tracking as commercial MOCAP systems. Those effortlessly scale up to large tracking volumes with multiple cameras (every time you see actors in a huge set wearing mocap suits? Probably Vivon, OptiTrack or similar). Constellation specifically uses the same pulsed marker sequence the DK2 does, and as long as you tell each camera what point in the sequence you are in the current frame (i.e. run the sync cable between cameras) you can scale from the current two cameras up to as many as you want for the desired tracking volume and occlusion robustness.
    It remains to be seen which system is cheaper in volume production. Constellation uses a relatively common sensor, some OK optics, and some degree of onboard processing (i.e. a low-end SoC to do some basic thresholding and compression). Lighthouse is electronically simpler on the emitter side, but has multiple high-speed precision mechanical elements, some actually-decent lasers, and needs robust safety interlocks to achieve consumer-grade laser safety classification. On the 'marker' end, the two systems are very equivalent in terms of components: Constellation has a timer driving pulses of multiple photodiodes (LEDs) per tracked object, Lighthouse has multiple photosensors driving one or more timers per tracked object.
    Electromechanical devices have invariably proved to be more expensive to manufacture than purely electronic devices, but Lighthouse's emitters can potentially cover a larger volume per-emitter than a lensed camera. On the other hand, Lighthouse currently uses sequential operation of the basestations (one will active both axes in sequence then blast a global sync pulse, the the next base station will use that as its cue to start its scans, etc), which would drop tracking update rates as more basestations are added. Constellations cameras are globally synchronised so would scale with the number of objects tracked rather than the number of cameras. Adding more objects could be accomplished by increasing pulse code length (or reusing pulse codes) at the expense of initial constellation acquisition time, and just hope that nothing gets occluded from ALL cameras for long enough for the IMU based position tracking to drift too far (1 second is enough for a few metres of drift, so we're talking about sub-second total occlusions here) to prompt reacquisition from first principles (needs to wait for a whole code sequence).
     
  2. d_stilgar

    d_stilgar Old School Modder

    Joined:
    11 Feb 2010
    Posts:
    926
    Likes Received:
    47
    The pricing issue is obvious. If there's a big enough difference in price without a big difference in user experience then one system will come out on top. But most people asking "which should I buy" aren't asking based on price, or else they'd be asking what the prices are.

    Required specs to run the Vive is a strawman issue at best. "Hey, will my graphics card run this Dell 1080p 60hz display as well as this NEC 1080p 60hz display?" The specs of the two HMDs are close enough that hardware requirements are going to be nearly identical.

    Launch lineups will be the same barring exclusives, which I mentioned in my post. The only other difference will be accidental, temporary exclusives based on the Rift CV1 not shipping its motion controllers until H2 2016.

    On the camera tech not having more overhead you're wrong. Comparing someone's home PC to a $50k professional mo-cap studio is silly at best. Your computer may be fine running 1-3 cameras, but at some point you just get a bandwidth issue if not a CPU processing limitation. You're asking the machine to 1) Have the inputs necessary to accept multiple cameras. 2) Have the bandwidth to actually push a video signal through all these inputs. 3) The CPU capacity to process each of these signals. 4) Have enough leftover bandwidth to still run your game and then push out 1080x1200 per eye at 90hz.

    This isn't solved with lower resolution cameras since the ability to differentiate between tracked IR LEDs is directly affected by the camera's resolution (low res=short distance from camera, high res=further distance). So you either have fewer high resolution cameras or more low res to track the same area, but neither case really solves the bandwidth problem.

    You could solve the problem by removing the bandwidth and processing load from your computer and onto the cameras themselves so that all that is passed back to the computer is the telemetry data, but then you have to add hardware to each camera which will bump up costs greatly.

    These HMD systems are going to push 233 million pixels per second. The specs we have now are in a sweet spot that modern GPUs can handle while giving a good experience to users. I'm saying to accentuate the fact that much higher res displays already exist, but our computers can't handle 90hz at higher resolutions yet.

    Valve's lighthouse system flashes 60x a second. I don't know the frame rate or resolution of the Rift cameras, but let's replace the two lighthouse base stations with two 1080p cameras capturing 60 fps to compare to the Vive. That's now 250 million pixels per second that need to be processed to determine your location. Let's try and scale up the room with a couple more cameras. Now you have 500 million pixels per second that need to be processed to track your location. Things could be optimized slightly so that certain cameras aren't processed until you're nearby, but that eliminates the possibility of multiple users being tracked in that larger space at once.

    So at a minimum with the Rift you need to not only output 233 million pixels per second, you also need to input and process 250 million pixels per second.

    So keep pretending there isn't a difference in the scalability of these systems.


    While there will be a difference in scaling between the two systems, it will be very slight and we're not sure which is better worse at this point. Obviously analyzing large amounts of pixels is much easier than creating rendered 3D output (I never thought they were equal but didn't elaborate for the sake of keeping a long post from growing more).
     
    Last edited: 20 Oct 2015
  3. Parge

    Parge the worst Super Moderator

    Joined:
    16 Jul 2010
    Posts:
    12,866
    Likes Received:
    552
    Ha ha! No. I don't know if its misunderstanding or fanboisim but you are wildly misrepresenting the load created by the Oculus Camera.

    When talking about in game rendered pixels, and camera captured pixels you can't just add the two up and go 'here is the load' - it’s not just like comparing apples to oranges, it’s like comparing apples to dogs – they aren’t even slightly the same, in any way.

    Palmer addresses this exact issue here

    "“Even in the multi camera demos,” Palmer says, “we are well under 1% CPU power, it’s just insignificant to do this kind of math.” Even when adding “more cameras and more objects,” we are guessing something like of four cameras, two headsets, and two sets of controllers, “it is only eating up 5% of one core.”

    So, the Rifts cameras will scale, just like the Vive's with no appreciable difference or load on the host computer.
     
    Last edited: 19 Oct 2015
  4. edzieba

    edzieba Virtual Realist

    Joined:
    14 Jan 2009
    Posts:
    3,179
    Likes Received:
    284
    Both Lighthouse and Constellation need to take several 2D matrices of tracked points (with IDed points) and perform a Model Fit with aid from IMU orientation data and past position. This is the main fixed tracking load, and is identical for both systems. Adding more matrices has minimal impact, and may even make things faster by reducing uncertainty by reducing the problem-space for the intial setup model fit.

    In terms of camera processing overhead for Constellation, literally all that needs to be done is to:
    - Threshold the image to cut down on background illuminants (done in the camera by the ISP). This also dramatically cuts down on the amount of data that passes over the USB bus
    - Locate the centroids of the blobs (coughOpenCVacceleratesthiswithnoeffortcough)
    - Decide if the blob is bright or dark, and compare this to previous blob IDs to keep track of the pulse sequence

    All of these could conceivably be done by the camera itself with a very low end ARM chip, it's just not worth bothering with for the relatively low performance cameras currently being used. All that would then go over the USB bus would be a string of co-ords. An architecture like this could even take advantages of contrast-sensors like IniLabs' DVS cameras (though that would require a huge increase in LED marker switching rate, this is trivial to implement).
     
  5. d_stilgar

    d_stilgar Old School Modder

    Joined:
    11 Feb 2010
    Posts:
    926
    Likes Received:
    47
    I spent some time this morning looking into it. I was wrong and you guys are right, which is great. I'll go through after work tonight and add the strikethrough to all false info so that you can still mock me (if you like) but I won't be spreading false information.

    From what I was reading the max CPU load people were getting from the tracking was ~5%. It's still more than zero to scale, but not something anyone doing any reasonable setup needs to worry about. We'll have to wait to see which system scales better. (For those who still buy a sound card for the minuscule performance boost, it might matter).

    It doesn't remove the trouble of more wires. Lighthouse base stations can just be powered and sit somewhere. They don't transmit any data back to the PC. That's a plus in terms of scaling, but for other reasons, not CPU load.

    EDIT: And I'll add that my statement about possible future hand tracking via cameras remains true, which is a potential future benefit the Rift has over the Vive, although in later generations. I'm not trying to poop on any one system right now.
     
    Last edited: 19 Oct 2015
    edzieba likes this.
  6. phuzz

    phuzz This is a title

    Joined:
    28 May 2004
    Posts:
    1,700
    Likes Received:
    24
    Just to confuse things further, the cameras usually used for tracking are normally IR, so they're effectively recording a grey-scale image, whereas the display is obviously colour.
    Also the refresh rate can be different for the camera, and it's the amount of calculations per second that we're interested in here.
    There's an interesting look at the DK2 camera and LED system in this post.
     
  7. rainbowbridge

    rainbowbridge Well-Known Member

    Joined:
    26 Apr 2009
    Posts:
    3,171
    Likes Received:
    69
    just home looking up any kind of SMS alert service for when the HTC vive will become available? just thought it would be a great idea if people ganged together to create VR game cafes in major citys.


    You don't have enough money to get the full rig? but you do have £10 and want to game for 2 hours? come x studio or warhorse and get to game with up to 10x or 20 others in a modern VR mmo game, elite, star citizen, etc.


    there is going to be a window of 6 to 12 months which if you set up a comfortable safe space for people to explore VR it would be successful, there are far too many cool nerds that want to meet up and experience high level gaming experiences. like the pictures from the 60's but for today.

    would be uber cool to get to a huge space converted to allow fokes to take on serious competition VR environments, like star citizen for example, 5, 10 , 25 crew ships could do with a meet up first, maybe one two bears then into major operation.


    I think VR is going to be a really cool thing man, i think some of the stuff is going to be base but there are going to be tittles that reach high brow and beyond into life changing games.



    google image, internet café for further reading
     
  8. rainbowbridge

    rainbowbridge Well-Known Member

    Joined:
    26 Apr 2009
    Posts:
    3,171
    Likes Received:
    69
  9. Ficky Pucker

    Ficky Pucker I

    Joined:
    9 Jul 2009
    Posts:
    1,591
    Likes Received:
    94
    i think i know what to ask santa for christmas :D
     
  10. rainbowbridge

    rainbowbridge Well-Known Member

    Joined:
    26 Apr 2009
    Posts:
    3,171
    Likes Received:
    69
    Wounder what the proce is going to be? if its over £500 then thats stupid.

    It could be any thing and they would sell but the idea is to be accessable to everyone asap, it shouldnt be a toy for the big budget only.

    Its the elite dangerious patch to enable vive support which is
    Going to create interest.


    Htc has hopefull created enough supply to put at lesst 10k units in each region ?

    Why dont they just open the order request page now and a sense of unit Requirements, this would surely help in correct manufacture amounts.

    A proper large scale manufacting factory should be able to knock out 1k units a day surely?
     
  11. rainbowbridge

    rainbowbridge Well-Known Member

    Joined:
    26 Apr 2009
    Posts:
    3,171
    Likes Received:
    69
  12. edzieba

    edzieba Virtual Realist

    Joined:
    14 Jan 2009
    Posts:
    3,179
    Likes Received:
    284
    It's not outside the realm of possibility. A more mechanically complex (it's harder to cheaply build something with high speed rotating parts than solid-state optoelectronics) tracking system, plus bundling of two short-production-run controllers (I bet most of the price of the steam Controller is in those custom capacitive pads, shared by the Vive wands) rather than a single controller already produced in megaunit quantities, could push up the price, plus HTC likely really want to sell it at a profit rather than at-cost.
    The initial wave of VR isn't going to be cheap. The computer alone will set you back a few hundred quid unless you already have splashed out a lot on a high-end rig (in which case you can afford £300-£500 for a HMD, which is comparable to many good quality desktop monitors).

    Magic Leap; I'll repost what I'vce written elsewhere on that video:
    - The images appear opaque, but they also appear to be using the same trick Microsoft did with the in-person Hololens demos: use an exceptionally dark environment and display only very bright objects.
    - The focus switchs mean that the displayed image is not in the same plane as the background (as is the case with the majority of AR headsets anyway), but it does not demonstrate that a lightfield is being generated
    - That appears to be a very narrow field of view. The actual visible field of view could be larger than the cameras field of view, but I'm not inclined to be generous given the poor demonstrations from Microsoft (artificial view with no frustum culling) and others (e.g. the initial FoV diagram from Oculus' Kickstarter)
    - The tracking is atrocious, but this can be improved easily using existing systems, so is not a black mark

    Magic Leap have released vast quantities of total guff, and even their patent filings have thrown as many HMD designs as possible against the wall to see what sticks (TIR diffraction grating optics, actuated fibres, mobile zone-plates, etc). They're throwing up all the classic flags of a technobabble cash-in scam (similar to Euclideon's 'unlimited detail' farce) that I'm holding them to a higher bar in terms of actually demonstrating they have something close to what they claim.
     
  13. wolfticket

    wolfticket Downwind from the bloodhounds

    Joined:
    19 Apr 2008
    Posts:
    2,887
    Likes Received:
    208
    £499 wouldn't surprise me :)

    At the moment you need a pretty expensive system to connect it to anyway, which may imply early adopters will have the sort of funds and inclination to spend that sort of money. I can't see them making a huge mass adoption-loss leader push as is common with new consoles as the market isn't proven yet and supply will still probably outweigh demand even at a fairly high pricepoint.

    Spending £4-600 on a high end monitor is hardly considered ridiculous and this is all that tech and more compressed down into something you can strap on you head. As phones compared with tablets sometimes show, smaller≠cheaper.
     
  14. rainbowbridge

    rainbowbridge Well-Known Member

    Joined:
    26 Apr 2009
    Posts:
    3,171
    Likes Received:
    69
    fresh look in filtered games supporting VR, date released shows new game vector 36,




    reviews,

    This is the game VR headsets are made for. Try it with steam VR or oculus if you get a chance, it's INSANE. Especially when you're going 1000kph, 10 feet off the ground, through clouds of flaming debris, and just barely clipping your wings(?) on the cliffs.

    I found myself spending more and more time customizing my skimmers once I realized I could make them look and fly pretty much however I want. For higher speed races, it becomes a whole challenge in itself to basically make the fastest manned rocket possible that can still be steered. The interplay of engineering and skill required is really unique, and even though it's only single player right now, the bots are surprisingly competitive. It definitely takes some mastering.

    Overall it's showing tons of promise... especially for a one-man development team. It's still in the early stages so hopefully we'll see more and more variety down the road. I can't wait for multiplayer, to see (and demolish) other people's hard work!


    --
     
  15. rainbowbridge

    rainbowbridge Well-Known Member

    Joined:
    26 Apr 2009
    Posts:
    3,171
    Likes Received:
    69
    fellow bit techers with dk2.

    please sign up here.

    https://account.evevalkyrie.com/signup/

    put your dk2 s/n in..

    apparently its pretty good.. hopefully we get the email about the acces in the coming weeks and help the alpha!

    :thumb:
     
  16. Griffter

    Griffter New Member

    Joined:
    1 Jun 2012
    Posts:
    414
    Likes Received:
    1

    awesome, thx!
     
  17. edzieba

    edzieba Virtual Realist

    Joined:
    14 Jan 2009
    Posts:
    3,179
    Likes Received:
    284
    A heads-up on EVE: Valkyrie: you will need an Xbox 360 gamepadpad (or other compliant Xinput device). No other controller is currently compatible, and there is no remapping.
     
  18. rainbowbridge

    rainbowbridge Well-Known Member

    Joined:
    26 Apr 2009
    Posts:
    3,171
    Likes Received:
    69
    edzieba you know every thing about VR,

    what are you most excited about regarding VR in the coming 12months?
     
  19. edzieba

    edzieba Virtual Realist

    Joined:
    14 Jan 2009
    Posts:
    3,179
    Likes Received:
    284
    12 months? Realistically, there's not going to be any hardware crazyness in that timeframe (I've got a Cyberith on order, but I'm under no illusions it's going to garner much support that isnlt hacked in), so I'm most exited about the first consumer-facing generation of decent VR drumming up development. With the Vive and CV1 combined, I'd make a ballpark estimate of about 1 million HMDs sold by the end of 2016. That's enough of a market that it will attract quite a few developers, particularly as anyone who owns one will de-facto have a significant amount of spare cash (you own a ~£700 PC and a £300-£400 HMD, of course you have cash) and initially not a whole shedload of content to run on it. By the end of those 12 months, the profit margins will have the larger studios actually starting to pay attention and attempting to jump on the bandwagon at the last moment, so with any luck we'll roll into 2017 with companies with large marketing budgets advertising VR and pushing further adoption.
    The pessimistic view though would be if initial content availability is too low, and interest fizzles out by the end of the year. The might of Nintendo couldn't push the Wii U and that had existing AAA partnerships, multiple first party studios, established franchises, a large and mostly devout fanbase, and a large marketing push. But not enough attractive games at launch, so momentum fizzled fast. We know Oculus are entirely funding several external studios to create games for the CV1, and Valve is probably doing something similar, but who knows if it will be enough?
     
  20. rainbowbridge

    rainbowbridge Well-Known Member

    Joined:
    26 Apr 2009
    Posts:
    3,171
    Likes Received:
    69

Share This Page