Virtual Reality: 1993 and 2013
Comparison between 1993 Division dVisor and 3D Mouse vs 2013 Oculus Rift and Razer Hydra.
The Division system was around £70,000 new and display a 10 thousand polygon virtual world at 10 fps - the Rift and Hydra together were about £300 and modern systems can display a million polygon VR world at 60 fps and nine times the resolution. What a difference twenty years makes.
The Division system is a standing, walking about system, and designed for a room with about a 10’ radius to move in, there’s nothing to turn your body like a joystick, you just turn your body - the hand controller design reflects this too - it’s a point and grab / shoot thing, not a movement controller with joysticks.
The system was very complex to set up and only ran the software designed for it - VR systems were fully integrated, so no third party software, but an immersive 3D world builder, a drag and drop scenegraph builder with scripting, format converters and a C API.
The development version of the Oculus system is very much a sitting down immersive VR experience, non-positional tracking, short range Hydras that use dual joysticks for move and turn. Short cables throughout - it’s a good system for playing while sat into front of your regular computer, but that’s not as immersive an experience as being able to walk around, physically moving your body over to objects in the virtual world.
It’s easy to set up and is gaining pretty wide software support, with a great buzz surrounding it. It’s directly programmable in C++ but also supports Unity and UDK, which play an equivalent role to dVS/dVise. Consumer software, mainly games, also directly support the Rift, and that market simply didn’t exist in 1993 - most PCs didn’t have 3D graphics and texture mapping was expensive and new on workstations.
Virtual Reality Hardware
Both systems together - the modern one is much smaller, and much simpler and very much cheaper (there’s connectors on the Division kit that cost more than Rift). The older system has a very high end build quality though, and great ergonomics, making it very usable. Both have wide field of view optics and independently tracked hand controllers with several buttons. The Oculus system using a Razer Hydra.
Any VR system has a variety of subsystems, and I’m going to look at those and compare them. These are the Head Mounted Displays (HMDs) themselves, the way the pictures get to the displays and how the peripherals are set up, the tracking system, the hand controllers, the host systems and finally the software used.
Head Mounted Displays
The dVisor is based on the Virtual Research Flight Helmet, a late 1980‘s HMD, also based panoramic LEEP lenses and LCD TV screens. Before this HMDs had to use CRTs and were very bulky - essentially both the Flight Helmet / dVisor and the Oculus Rift made use of technology that had only just become available to create a new kind of VR HMD.
The dVisor has larger optics, using the LEEP lens system and larger, lower resolution screens - very early LCD TVs, taking an NTSC TV signal each, and needing a lot of space in the front of the HMD. The back is heavy to balance it, and the overall fit is pretty comfortable - if heavy. The Rift uses a single, flatter, high resolution display and uses a top strap rather than a counterweight, so the HMD is much smaller.
The dVisor looks huge next to the Rift, but it does include an expensive spatialised sound system, and a single knob to adjust fit. Both can be used with glasses - the Rift’s screen can be moved further away, and the development version has different lenses for near-sightedness.
The red peak at the top of the dVisor holds the tracker for the head - a Polhemus Fastrak tracker.
This shot above shows the optics of the two HMDs, the LEEP optics are large, and fairly complex, the Rift optics are smaller. The LEEP set has a wider field of view (40x120 degree, about 110 degrees horizontal) than the Rift, and both concentrate more pixels towards the centre of the visual field.
The Rift has about 9x the number of pixels, but no pixellation filter - these aren’t really optional at the very low resolutions of the dVisor, the Rift has a noticeable pixel grid.
This shot shows the inside of the dVisor and the adjustment knob,this can make the band pretty tight around your head, so the HMD doesn’t move during use. It also makes for a quite easy and fast don/doff cycle.
The Rift uses elasticated straps and an over-the-head strap to help take the weight of the HMD, and isn’t quite as fast or slick, but is still fine.
More dVisor internals include Sennheiser speakers - not padded as they don’t touch your ears directly, air flow around the head is very good with HMD, it’s not hot or stuffy.
More on the Division system now, this is really illustrating how complex VR systems used to be, and what a cool job the Oculus team have done making it as easy as they have.
Division did sell a computer with all of this inside and high performance graphics, this is the modular version for SGI.
Cabling diagrams for both systems - the Division system uses quite esoteric kit and many types of connector. The Oculus system is much simpler, using a few connections that didn’t exist in 1993.
The dVisor HMD needs to be fed a pair of NTSC connections, so the host computer has to display two screens, with slightly different images (right and left eye). This is done by means of a rare option for SGI Indigio2 workstations that lets them drive up to four displays - the Impact Channel option. It adds an extra (GIO bus) card to the computer and then there’s this wacky breakout box - attached via a 25W3 connector - the first of many odd connects.
The box itself, looks like this - we use the top two, and attach cables with 3 x BNC bayonet connectors at one end, and VGA connectors at the other end. These aren’t easy cables to get if you lose one. The Rift just uses and HDMI cable and one screen with both eyes rendered onto a single screen - this is way easier, and definitely the way forward.
Our video signal makes its way from the SGI, out the ICO breakout box. The cables for use with the ICO are red, green, blue, H.sync and V.sync. We only use the colour lines though - I think it just syncs on green.
From there into the VGA ports on the left of this little box - this a dual VGA to NTSC scan converter. The NTSC signal then heads on down into the Division Integrated Peripheral Unit, coming in as composite NTSC on the back.
The back of the IPU, with left and right video feeds, serial for the tracker, serial for the mouse buttons. I’m not sure if we actually use parallel for anything, or if that was a Division internal thing. Audio is also visible - 3.5 mm audio jack on the back of the computer, splitting to dual RCA connectors on the IPU.
At the front of the IPU the huge round connectors for the HMD and the 3D mouse.
The round connecters are military spec Amphenol connectors, reflecting VR’s customer base at the time. Each connector is about £70, for each side. The connectors by themselves cost more than an entire Oculus Rift.
The Oculus Rift equivalent of this rather heinous setup is much simpler. For the HMD display, the Rift box connects via DVI or HDMI to any PC or Mac and is recognised as a new display (of type “RIFT DK”) of 1280x800 resolution. The tracking data is via a normal USB connection.
The Oculus Rift control box is a bit smaller than the Division IPU and Scan Converter. Its HDMI connecter does tend to take your audio by default, and then not be able to use it, gotcha during set up. Standard DVI or HDMI for video, mini-USB to provide tracking data back to the computer.
The Oculus box has Brightness, Contrast and Power buttons - the Division system doesn’t have any of these, it is what it is, and if it’s connected, it’s on.
The Rift is hardwired to its control box at both ends on a four foot cable, the dVisor connects via those big round connectors at the IPU end of its ten foot cable, and is hardwired at the HMD end.
In terms of what the host computer sees, the dVisor basically requires two screens of 640x480, each one rendering the view for each eye. The Rift offers a single screen of 1280x800 and the software has to display the left and right eye images side by side on this one display.
The Rift’s system is far better, needs half the amount of displays and complexity in the HMD and is easier on the host computer side - completely unavailable to Division in 1993 of course. The computer systems contemporary with the dVisor all ran X11 to talk to the displays.
The Division dual screen system is much more complex from a software point of view, and required the use of their own software (dVS/dVise) to work. If anyone has a copy of this software for Irix, for the love of all that holy please let me know!
No third party software works out of the box for the Division system. I’ve tried to get Maverik and VRPN running with it, but haven’t put a lot of time into it.
At the front of the IPU, the Polhemus Fastrak connections (the Fastrak is inside the IPU) - all DB15 serial at 9600 baud. Male to the magnetic transmitters, and up to four female for the receivers (which are the trackers) - two are used, one for head and one for hand.
Dip switches set the Fastrak up, tell it what speed to use and how many trackers - the Fastrak is a 120 Hz system with a single tracker, 60 Hz with two.DIP switch setting are as per the Polhemus Manual, Division just stuck it in a box along with the other bits and pieces to run the HMD.
The Rift uses IMU tracking - sensor fusion between solid state accelerometers, gyroscopes and megnetometers, to provide 1000 Hz resolution, with three degrees of freedom - the development version of the Rift doesn’t do positional tracking as yet.
OK, the hand controllers and tracking. The Polhemus branded magnetic emitter is on the left, the Razer Hydra magnetic emitter on the right.
The Polhemus unit comes with a five foot pole to mount it on, and has a range of about 15’. The Hydra is a desktop affair, with a range of 4’ or so.Placement is important for both, and magnetic fields interfere with them.
Hand controllers - the Division 3D mouse is heavy - a pistol grip in solid epoxy containing a tracker and five clicking buttons - to on the front, three on top. It’s actually very comfortable and usable, the weight feels really nice.
The Hydra has two controllers, each with a joystick, five buttons and two triggers, but is much lighter and feels like a cross between a pistol grip and a games controller.
Having used both, and with a fully working Division system, in 1994, I have to say the Oculus system feels much more like a consumer version - lighter, cheaper, ergonomics not as good - and this is probably a good thing. It’s cheap enough to create a mass market for VR, which will allow experimentation with HMD configuration, hand controllers, etc. and the market gets better.
1993 in nutshell
110 degrees diagonal field of view, 146k pixels, 2 x VGA inputs, magnetic tracking with six degrees of freedom over RS-232, integrated spatialised sound, 100s of installations of proprietary or high end Silicon Graphics computers.
Systems cost £50,000 to £250,000 and you develop your own worlds and applications for them. Mainly sold for research and military use.
2013 in a nutshell
90 degrees diagonal field of view, 1024k pixels, 1 x DVI/HMDI input,HMD has IMU tracking with three degrees of freedom and Hydras have magnetic tracking with six degree of freedom both over USB, no sound, around 7000 developer installations on any PC or Mac with 3D graphics.
Developer kits cost $300 and the Razer Hydra system cost £50, both support a growing number of games and apps and are supported by the Unity 3D and UDK systems. Mainly for consumer and gaming use.