New CAVE Design

Hept-Cave Design Wiki

This page serves as a repository of the design issues, knowledge, and decisions for the new Brown Cave funded through an NSF grant from 2009-2013.

Design Goals

• 20/30 resolution (== 100dpi desktop monitor?); 20/20 would be a real coup, if possible -- "retina cave"
• 360° FOV (vertical sufaces plus ceiling and floor)
• seamless; this likely implies overlapped rear projection and rounded corners
• stereo
• contrast and brightness comparable to today's conventional desktop computer
• design must accomodate 3D tracking (ergonomic, "versatile", full-volume, high-accuracy, low-latency, head+hand+props+fingers+feet)
• equipment (hardware plus construction) budget of $1M
• all parts available for order by 9/30/10
• avoid technology unnecessary for primary goals (unless it can change the primary goals)
• easy to program
• supports commercial software solutions
• reliable, i.e., high uptime

Some questions that should probably migrate to the "Design Issues to Resolve" section:

• Q: is it 8' cube again? what are pros/cons of larger/smaller display?
• Q: Is the software that will drive the display a constraint on complexity of the hardware? how many machines will drive display? 50? does VRG3D and/or other software scale trivially to expected # of machines that will drive it?

Charge for Dec 31 2010 final implementation plan

The charge is to have screen material ordered, projector supports defined and possibly ordered, blending and projectors, all engineering drawings done, floor decided and ordered. Do we need any software dealt with during this time? AMIRA Aviso? VRG3D?

For Nov 16 2010 meeting

• sgf: new projector plan
• dhl: facilities contact and plan for space committee
• dhl: get 2 stereo projectors

For Nov 16 2010 meeting

• asf,brad: vrg3d

For Nov 9 2010 meeting

• jvr may attend via skype/phone
• new topic: projector latency [jvc now says 4-5 frames latency (!); "frame" not defined; stereo may be only 24FPS?; half-res maybe; sharp XVZ7000?-full 1080p stereo; mono projectors self-switching to black?]
• all: jvr role in projector holders
• jnh: pro/cons for 40 vs. 50 decision (query Rod via e-mail for any needs) [no ethernet on 40, only RS232, but maybe that's a feature]
• jnh: make contact with citigraphics (4th try)
• jnh: scope renovations
• sgf: facilities input re room possibilities (and costs?) [cliff resnick top dog; bobbie delegated]
• sgf: splint experiment of 2mm from engineering
• jnh: prep to talk to integrators at supercomputing
• sgf: approach for faster iteration
• sgf: bigger piece of 1.2 matte screen material
• sgf: color management conf call results to help decide whether to pay for internal color management (projector choice) [prob no CMS]
• sgf+jnh: lay out the arguments for the two cases of color management; need scalable cost estimate, as well (find out about controls for both conditions)
• sgf: find out from Rod what controls and interface on the two projectors we're considering
• dhl: is it time to get demos from Tom, despite our concerns?
• asf: VRG3D build proposal for "make" installation on linux
• brad: one-command VRG3D download/build/install that asf blesses

For Nov 2 2010 meeting

• brad: one-command VRG3D download/build/install that asf blesses
• all: jvr role in projector holders
• jnh: pro/cons for 40 vs. 50 decision (query Rod via e-mail for any needs) [no ethernet on 40, only RS232]
• jnh: make contact with citigraphics (3rd try) [several e-mails out]
• jnh: scope renovations [haven't started, by end of week]
• jnh: prep to talk to integrators at supercomputing
• sgf: everything from last meeting
• new topic: splitting up work on mounts and alignment platforms (could we skip alignment?)
• jvr may attend via skype/phone

For Oct 19 2010 meeting

• dhl: get distortion measurements onto Sam or John's list
  • john already did it!
• sgf: splint experiment of 2mm from engineering
• sgf: approach for faster iteration
• sgf: bigger piece of 1.2 matte screen material
• jnh: make contact with citigraphics (2nd try)
  • e-mail just sent...
• sgf: color management conf call results to help decide whether to pay for internal color management (projector choice)
• sgf+jnh: lay out the arguments for the two cases of color management; need scalable cost estimate, as well (find out about controls for both conditions)
• jnh: more info about projector, including step size for various motorized controls
  • min step size for button shift is ~1pixel
  • min zoom moves edge ~1.5 pixel
  • keystoning is digital and should not be used
  • acutance seems a little fuzzy on a whiteboard
  • barrel/pincushion distortion < 1 pixel
• sgf: find out from Rod what controls and interface on the two projectors we're considering
• asf: VRG3D build proposal for "make" installation on linux
  • see e-mail
• sgf: facilities input re room possibilities (and costs?)
• sgf+jnh: decide what sw issues need to be resolved this fall (tracking? color? reliability/testing? cluster management?)
  • resolving issues of color blending and distortion correction
  • get VRG3D acceptable to dfk and others

For Oct 13 2010 meeting

• jnh: chinese screen-maker progress [citigraphics contacted, but no response yet]
• sgf: results from seam experiments and recommendations [jvr experiment too weak; likely to need some additional physical reinforcement; something like a thick piece of clear rigid tape to splint the joint along the seam -- in process by next week;]
• jnh: set up projector, probably Tuesday [looks pretty! noisy auto lens cover; eco mode quiet; one side exhaust; lens shift and zoom seems maybe fine-grain enough to use; lots of color correction menu options]
• sgf: facilities input re room possibilities (and costs?) [not yet]
• dhl: touch bases with Jan re process for facilities work and budget
• asf: VRG3D testing and feedback from dfk [high order bit from dfk: best way to go forward is to not use any of the new build stuff but use the source forge version with their own build process; still using older G3D; dfk is missing a build process where you can just type "make";]
• sgf+jnh: time commitment to accomplish this through 12/31 [sgf 20%, jnh 25%?]
• sgf+jnh: decide what sw issues need to be resolved this fall

Charge for Sep 30 2010 final design

The charge is to have design doc done enough to order from. Includes screen, geometry, projector, mechanical, cooling, believable budget.

For Sep 28 meeting

• sgf: doc and budget on 9/21
• sgf+jnh: final doc basically done
  • budget
  • well defined pieces for all the component/hardware areas: projectors, screens, floor, mounting/mirrors, mechanical
  • thoughts on backup projectors: better contrast 720p active stereo
• dhl: tom stites re projectors (chassis specs, pricing, delivery schedule)
  • price, specs, order time, loaners of mono ones (poke with NSF-deadline)

For Sep 21 meeting

• sgf: put design doc in google docs
• sgf/jnh: concrete suggestions for scope of work for outside contractor (or us or jvr (projector mount)
  • rear-projection floor
  • projector/mirror mounts
• sgf: put budget numbers in your doc
• sgf: nail down screen material
  • 50/50 chance that clarex can't be bonded
  • citi graphics (sp?) screen material
• jnh: chinese screen material from Raj
• sgf: ping JVC on price, specs, order time, loaners of mono ones (poke with NSF-deadline)

For Sep 7 meeting

• sgf: scope of work for outside contractor (or us!)
  • refined "methods" document with decisions made and motivations removed
  • sgf: heat and cooling budget
  • ???: refined budget and questions impacting it
• sgf: rest of info from Astra about Clarex blue ocean on sizes, thicknesses, cost, thermal expansion (2mm may work, call scheduled for 8/31)
• sgf: more info next week on facilities -- dhl thinks that grant and cost-sharing funding can't pay for renovations
• sgf: ping JVC on price, specs, order time, loaners of mono ones (e-mail pinged him again Monday
• brad: wrap of VRG3D stuff

Charge for Sep 30 2010 final design

For Aug 31 meeting

• all: comments, change suggestions, "todo" list from jvr proposal
• sgf: info from Astra about Clarex blue ocean on sizes, thicknesses, cost (we have 0451 blend 2mm piece -- pretty rigid
• sgf: more info next week on facilities -- dhl thinks that grant and cost-sharing funding can't pay for renovations
• sgf: ping JVC on price, specs, order time, loaners of mono ones (e-mail pinged him again Monday)
• jnh: check into G3D video hardware/driver knowledge -- can we leverage it? (not really, but testing some ATI cards)
• brad: vrg3d for OSX (checked in, broke everywhere else)

Charge to jvr, sgf, jh for Aug 2010 design-a-thon

The charge for the first two weeks of August is based on the two deadlines that our NSF program officer has explicitly asked for updates on:

1. 8/15 draft revised Cave design with feasible, existing hardware
2. 9/30 final revised Cave design
   • wiki design document
     • including mechanical issues
   • seams
     • nippura discussion
     • machining/welding (but probably not next week)
   • de-cornerizing explorations -- a chance to distinguish the design
     • vertical edges probably more important than horizontal
   • talk to Rajeev
     • curved projection materials
     • can they support the JVC?
     • can they do corners? curved corners?
   • ponder de-top-edging (aka ceiling) -- about human perception
   • ponder wider "front" wall
   • can outside viewers be able to see more (wider opening)

John, you are the boss for the purposes of creating a document in response to this charge. For 8/6 (next Friday) I would like to see at least one and not more than four sketches of designs with proposed specific technologies. Any proposed technology must have at least one example of a product on the market that can be used to instantiate it. For each of these tentative designs, I would like an estimate of the various factors that define the design goals: accuity (e.g., 20/30 and dpi), contrast, brightness, FOV, cost, implementation complexity, programming complexity, ease-of-commercial-software use, reliability. For some of these, a letter grade may be fine. For others, reasonable units would be best. For this stage, back-of-the-envelope estimates are fine -- there are only 4 days to do it, after all.

On Friday, 8/6, I would like us to have a shootout of the designs, with selection of a (proper) subset (or perhaps hybrid?) to refine.

The effort should culminate the following Friday, 8/13/10, in a draft design that satisfies NSF's deadline: specific display hardware and estimates of the qualities of the result.

David L will be at 401-354-2819

Example design sketch for 1998 Cave

• four 800x800 XXX projectors
• 8' cube
• 3 rear-projected 8'x8' walls
• front-projected floor

Eval:

• accuity: 8.3 dpi, 8.6 arc min/pixel at 4'
• brightness: not enough to make out many colors, esp dark ones
• contrast: ~10 jnd's
• fov: 270 horizontal at center, ~30 degrees above horizon
• tracking: easily supports inertial, magnetic, sound, camera
• seamless: walls fine, edges and corners provide noticeable issues
• cost: $200K? (today)
• programming: same as any cave
• commercial sw: ditto
• reliability: collective projector MTBF 12 mos, driver/os problems (gfx esp) every 6-12 months, disk/network access failures 1x/month, etc.
• simplicity: can't think of an example of a bleeding-edge element to trim...

Weekly Cave action items

Below is a list of action items for each Cave design meeting.

• Action Items for Sept. 15th, 2009
• Action Items for Sept. 22nd, 2009
• Action Items for Sept. 29th, 2009
• Action Items for Oct. 27th, 2009

Initial Design Issues

University facility

• Problem: What are the requirements for University researchers in general?
• Input needed
  • Interview existing Cave-users and potential Cave-users to consider needs/requirements
• Experiments needed
  • (None?)
• Decision: ?

Display Surface Angles

Assigned to DHL. Updates are due by 12:00 noon on 9/14.

• Problem: What are the consequences of display surface angles and/or curvature (horizontal and vertical angles may be different)?
• Input needed
  • Expert feedback from Acevedo re KAUST design choices
• Experiments needed
  • Set up two side-by-side and top/bottom displays and evaluate anecdotally
• Decision: ?

Blending/Butting at Display Boundaries

Assigned to SGF. Updates are due on 9/14 by 12:00 noon

• Problem: What are the consequences of tiling displays together?
• Input needed
  • Expert feedback from sites with tiled displays like the ones we'll use (Calit2, Disney, KAUST)
• Experiments needed
  • Setup 2 displays and report on artifacts at edge between them (color, geometry)
• Decision: ?

Will commodity DLP projection technology work

• Problem: Are commodity DLP projection technologies high-enough quality & reliable for a next-gen Cave?
• Input needed
  • An understanding of how much effort Mark Mine et al. at Disney have put in
  • What is Mark's estimate on how much more time/effort it will take to have a high-quality display
  • standard desktop text does not look good on Mitsubishi HDTVs. Is that a "show stopper"?
  • Why is a projector-based system worse than HDTV approach?
• Experiments needed
  • ?
• Decision: ?

Text with Checkerboard Stereo

Assigned to JH. Updates are due by 12:00 noon on Monday, 9/14.

• Problem: What text resolution will be usable with checkerboard stereo and will that level be sufficient to do the experiments and applications that we want to do?
• Input needed
  • Expert feedback from sites already experimenting with the displays we'll use (Calit2, Disney, KAUST)
• Experiments needed
  • Compare visual angle of smallest comfortable text on desktop display versus on checkerboard stereo display.
• Decision: ?

=== Floor Design === Assigned to SGF. Updates are due by 12:00 noon on Monday, 9/14.

Treadmill Options

( not Omni-directional) Assigned to SGF. Updates are due by 12:00 noon on Monday, 9/14.

Fixed wall design, or can it unfold into a powerwall?

Assigned to SGF. Updates are due by 12:00 noon on Monday, 9/14.

Ceiling Value, and Design Options

Assigned to SGF. Updates are due by 12:00 noon on Monday, 9/14.

Contracting to an Integrator

Assigned to SGF. Updates are due by 12:00 noon on Monday, 9/14.

• Problem: Would contracting to an integrator to build based on an RFP be effective?
• Input:
  • Phone feedback from some of the potential integrators
• Decision:

3D Tracking options

Assigned to JH. Updates are due by 12:00 noon on Monday, 9/14.

• Problem: Would camera-based tracking be a value or time-sink for the Cave?
• Input:
  • Talk with people using camera-based tracking (Calit2, Disney, KAUST, Lit3D, Iowa State, UNC?, ..)
• Experiments needed:
  • Try first hand most promising camera-based tracking and document:
    • the degree to which it is: wireless, accurate, low-latency, unobtrusive
    • the degree to which it provides: head-tracking, wand-tracking, tracking of fingers/feet/body, tracking of "phicons"/props
• Decision: ?

Hardware supporting usability studies

Assigned to ASF. Updates are due by 12:00 noon on Monday, 9/14.

• Problem: is special hardware needed to support evaluation of Cave-like systems (e.g., devices that measure galvanic skin response (GSR))?
• Input needed
  • Advice from Mel Slater (How does Mel decide when GSR, etc. is important? How expensive is the system?)
  • Mel says, "[we use] a device called Nexus 4, which has skin conductance, ECG, respiration, and can also be used for EMG and EEG. I think that it is of the order of 10K or less depending on how many channels are used.".
  • VRPN supports some GSR device -- Russ Taylor says to check with Mary Whitton to learn more
  • Consider whether results from future (or past) studies would be significantly improved with such devices?
    • asf thoughts (may be wrong!): GSR helps detects changes in a person's "comfort"-- this might reflect the level of frustration or happiness while evaluating some user interaction or visualization method that the person doesn't verbalize; that may be useful additional information for formal studies. PTSD studies would benefit from the information, I should think. However, if the device requires stringing wires all around, it might cause subjects to behave differently which would be bad.
• Experiments needed
  • none
• Decision: ?

Software: commercial quality

Assigned to JH. Updates are due by 12:00 noon on Monday, 9/14.

• Problem: What do we need to do to make sure that commercial software will be usable in our new facility?
• Input:
  • Feedback from some of the software players (Avizo/Amira, Ensight, OpenSceneGraph, ParaView)
• Experiments:
  • try visualizing Leopold Grinberg's arterial branch dataset (we don't have easy tool for it right now, but want one!)
  • << who else on campus has wanted to visualize data but we didn't have a tool to do it-- use as case study >>
• Decision:

Software: researcher/developer tools

• Cave Software Overview
• VRG3D Requirements

• Vis 2009 discussion dfk/dhl/asf

Schedule, Timeline, and Deliverables

Assigned to SGF. Updates are due by 12:00 noon on Monday, 9/14.

The following describes our schedule, timeline, and deliverables using absolute months into the project (months 0-11 is year 1, months 12-23 is year 2, etc.).

Upcoming Trip to Californina

• Agenda and action items for Disney Imagineering
• Agenda and action items for CALLIT2

Year 1 (0.50 hw-person FTE +consulting)

We will finalize hardware decisions which will involve evaluating changes in availability since proposal submission; demonstrations of competing manufacturer’s equipment; consultation regarding details of acquiring, machining, and integrating elements; and ordering the central display equipment (projectors, screen material, computers, trackers). The new equipment will be received late in the year. Through a process that minimizes system downtime, the 1998 Cave will be taken down and removed.

• Acquire test projection system(s) (month 1)
• Acquire test graphics system(s) (month 1)
• Acquire test tracking system(s) (month 1)
• multi-site visit, design discussions, consultent visits (month 1 - month 3)
• finalize refined design (due month 3, 1.5 hw-person mos)
• order additional wall/test equipment (month 2)
• Begin build on test tiled display wall (month 3 - month 6, 1.0 hw-person mo)
• order Cave equipment (due month 5, 1.0 hw-person mos) screens; projectors;computer parts, racks; computer interconnect; computer-display connections; display-display interconnections; display calibration hardware/software; tracking hardware.
• receive equipment (due month 8, 0.5 hw-person mos)
• computer cluster assembled (due month 11, 1 hw-person mo) machine/construct/assemble computer supports; assemble computer hw with high-speed interconnect; install and test OS on cluster
• raised floor and other site preparations complete (due month 11, 1.0 hw-person mo)
• Install replacement tile wall, temporary system (month 11)
• remove 1998 Cave (due at least disruptive time in year 1 TBD, 0.5 hw-person mo)

Year 2: (0.67 hw-person FTE, 0.58 swe-person FTE)

In year 2 we will complete construction of the display, bring computer systems online, install TGS’s Avizo and our own VRG3D software, and begin testing individual and linked systems.

• display hardware (due month 20, 4.5 hw-person mos, +machining/fab), (will start in year 1) detailed screen support specifications; detailed monitor support specifications; machine and construct screen supports; machine and construct projector platforms; mount screens and projectors; rough-calibrate projectors
• display-computer interconnect (due month 20, 0.5 hw-person mo)
• tracker system(s) (due month 20, 2-4 hw-person mos) install and test; calibrate; create hooks for monitoring and testing
• VRG3D display software library (due month 23, 6 swe-person mos) source-control setup; ability to distribute software configurable for arbitrary multi-display devices; implemented for new display configuration; multi-display optimizations incorporated; tracking incorporated; hooks for automated testing incorporated
• install Avizo (due month 23, 1 swe-person mos)

Year 3 (0.63 hw-person FTE, 0.63 swe-person FTE)

Monitoring and testing reliability software will be designed, developed, and deployed. Application users and developers will be able to begin using the system.

• VRG3D-based example application (due month 26, 2 swe-person mos) identify additional problems with library
• distribution version of VRG3D available (due month 30, 1 swe-person mos)
• application users can now begin creating new applications
• monitoring/testing reliability software (due month 30, 2 hw-person mos, 4.5 swe-person mos) daily builds of software library and application; daily tests of machine accessibility (tracking, compute cluster functionality/performance, display device availability); touchscreen display of component status/availability; touchscreen access to troubleshooting options; easily extensible touchscreen interface for startup of (experiments, applications, demos, homework assignments) item displayfacet alignment system (due month 35, 5.5 hw-person mos)

Year 4 (0.17 hw-person FTE, 0.33 swe-person FTE)

In year 4 we will finalize the software infrastructure, complete the auto-calibration system.

• software library changes (due month 36, 2 hw-person mos, 3 swe-person mos)

(note that these are not maintenance, but part of the development process); feature request from specific application areas; application users will change apps to use them

• final software distributions to SourceForge (due month 47, 1 swe-person mos)