Re: Highest resolution fulldome projection
I failed to explain my point sufficiently, I'll have another go.
The space based real time systems and data sets that are available are
very good and provide a compelling way to explore the universe. What I
was trying to say is that not all shows are about space or to be more
specific stars, planets and other cool space stuff.
Our latest show Astronaut featured a section explaining the effects of
microgravity on the human body. We go inside the inner ear, observe bone
decaying etc, Another scene features advanced character animation
containing cloth simulation, particle effects, soft and hard bodied
dynamics simulations. To my knowledge it is not possible to implement
all the rendering effects and density of geometry in any current
Fulldome real-time system. Modern day graphics cards can do some amazing
things but to perform indirect lighting, displacement mapping,
raytracing etc at the kind of resolutions we require and are looking at
for the future is just not available as we speak.
Real-time offers many, many advantages over pre-rendered opening up a
new paradigm for story telling, but you can not achieve the look and
feel that people have come to expect from their experiences with the
world of cinema... yet.
My personal passion lies in trying to recreate a cinematic experience
within the unique environment of the dome. I think there is room for
both pre-rendered and real-time applications for some time to come but
my initial point was that you can not achieve what I want to put up on
the dome with current real-time systems. When the real-time systems can
do what I want and better (which is inevitable... maybe) I will be more
than happy to transition over!
- Clearly there will always be the need for pre-rendered no matter how good real-time gets. Some astrophysical simulations take CPU years to run. Artist Scott Draves uses 30,000 CPU's to run his particle animations. And live-action is rendered on an image capture system requiring linear playback.
I think the most popular use of real-time is to replace (and exceed) planetarium functionality in a digital system. Planetarians have traditionally told stories interactively, reaching down to turn a knob to produce diurnal, latitude, or heading star motion, advance through the annual seasons, display grids, etc. The digital planetarium adds a “z” axis allowing liftoff from the planet and travel through the known universe as well. This adds a tremendous educational dimension - knowledge of our place in the universe.
Regarding Ken's comments on HYBRID, I greatly respect what GOTO has done here. Many planetarians DO place emphasis on naked eye astronomy, and optomechanical does produce a superior sky simulation. A dedicated “star theater” tasked with naked-eye simulation of the night sky may require an optical projector for the best possible simulation.
A facility with broader educational goals may not see the value in such an investment, since naked eye astronomy is a narrow educational goal within the field of astronomy and astrophysics, which is a narrow branch of physics, which is only one of the sciences, which in turn is only one possible curricula at an educational institution which might include art, drama and other schools of study that would take an interest in a digital theater resource (planetarians may have to learn to share). Still, starballs are kewl and I do hope they don't fade away too soon...
How a HYBRID system can be cheaper than a digital system alone is beyond me, Ken.
Visual Bandwidth, Inc.
- Just a quick note here. With respect to all the cinematic concerns, (and of course pre-rendered will always be more cinematically beautiful than real-time, as both real-time and rendering capabilities evolve), what people are trying to achieve with real-time systems is NOT a cinematic experience. The most effective use of real-time systems that I've seen are with a live presenter and an interactive program. These will necessarily be different than a cinematic experience. Including the live presenter element breaks some of the immersive qualities of a cinematic experience. And this is fine, if interactivity is your goal. In fact, it could be said that part of the goal of an interactive experience is to break the cinematic 'trance' that often comes over audiences and cause them to think, rather than just view.
That said - I'm a big believer that both interactive and pre-rendered cinematic experiences should be part of our virtual 'bag of tricks' that we can deliver to our audiences. Let's have the best both worlds can deliver.
Just my $0.02
Todd K. Slisher
Vice President of Science Programs
Detroit Science Center
5020 John R Street
Detroit, MI 48202
It seems that the true utility of real-time systems is not to re-create the
cinematic experience but to evolve the immersive mediated experience beyond
a solely passive one. Granted, simply rendering pre-determined shots and
camera paths in real-time to replace their pre-rendered equivalent for most
material is currently not feasible (though many gaming engines are heading
in that direction). But watch AMNH's Passport to the Universe and then look
at the same shots rendered in real-time in Uniview - you'll be hard-pressed
to tell the difference.
That said, Source, Unreal Tournament, Crytek and others are great examples
of gaming engines that are increasingly being used to create Machinima
real-time cinema (see http://www.machinima.com for examples). Any real-time
engine can theoretically be adopted for use in domes (we've already helped
create a dome version of Unreal Tournament at http://planetjeff.net), so
it's really a matter of the value proposition to implement this
modifications. It's is very unlikely that the best real-time fulldome
engines, especially for non-astronomy applications, will be developed by
fulldome hardware companies - they will likely be modified engines that are
already being developed for gaming and visualization applications.
All of this was discussed at length last year at the NEI fulldome meeting
We theorized and waxed poetic about how we could us Will Wright's new Spore
game in the dome and how to interface numerous handheld devices with the
larger display (ie everyone having a wireless PSP and playing each other on
the dome). Do yourself a favor and watch
http://video.google.com/videoplay?docid=8372603330420559198 - you'll
quickly get the idea of why real-time environments could be incredible for
interactive education. CMU Entertainment Technology Center students
addressed these issues as well with their Interactive Dome Project
(http://www.etc.cmu.edu/projects/dome/) last year.
And Tom is right - just because it's real-time doesn't mean it's cheap,
especially for high-quality data collection, modeling, animation,
simulation, lighting, etc. But serious real-time capabilities do
dramatically increase possibilities for new forms of interactivity,
networking, education, exploration, and experimentation - in other words,
some of the reasons that many of us became interested in immersive virtual
environments to begin with...
- My quick comment on full dome resolution ..... the final resolution as it relates to the informational content on the final dome surface is dependent on a whole raft of factors.
The resolution of the raw frames that make up the content is only one part of the story, it only determines the limit of the visual quality of the final projected result given a perfect projection system. There are however lots of factors that reduce the effective resolution of the result reflected off the dome surface, some of these are:
1. Codec choice. It seems many (perhaps most) planetariums are not usng lossless encoding/playback in which case the pixels being sent to the projection hardware are at a lower fidelity than the original rendered material. Personally I've always been surprised that some/many systems even use mpg and variants .... surely with todays hardware we can move past such compromising technologies.
2. Projectors don't give a 1:1 representation of pixels on the projection surface, this is especially so for CRT technology but also true for digital projectors. There are all sorts of sources for this including use of analog signals, the reality of lens physics, focusing on a curved surface, etc.
3. Many multiprojector systems use digital warping to correct for the spherical geometry, this lowers the information content.....individual pixels get contrbutions from neighbours so are no longer independent.
4. Edge blending is rarely (if ever) perfect, the result is usually a blurring (often significant) of the image between the projection patches.
I'm sure it is easy to find examples of planetariums with significantly higher theoretical projected resolution but where content looks worse than planetariums with lower spec'ed projection hardware.
One should also note that just because raw pristine frames may have a certain pixel count, that certainly does not mean it is higher resolution (in the informational sense) than content with a lower pixel count. This is clearly the case for filmed material, but also for CG. The resolution as it relates to quality is dependent on things such as antialiasing settings, quality of the model geometry, texture resolution, and other factors related to rendering technique.
- Posted for Brad Thompson:
Sorry for the long diatribe but Paul and David touched lightly on
something that I've encountered over and over again throughput my
years sitting under a dome. This is a bit off-topic and my long post
isn't meant as a retort to anything anyone said here. There is this
myth about real-time technology that I seem to keep encountering over
and over, even from people who really should know better. The key to
Paul's statement is that you "can't re-create a CINEMATIC experience
this way at the moment." I'm guessing that by "cinematic
experience", he's referring to passive linear cinematic
storytelling. Sure, Uniview, Digital Sky, D3, Starry night, game
engines, etc. are out there and being used to create great and unique
experiences. However, they aren't currently the most powerful/
effective/efficient tools for passive linear cinematic storytelling
on the dome (aka a cinematic experience.)
For that to be the case, a system would have to be able to generate
The Enchanted Reef, Black Holes: The other side of Infinity, DarkStar
Adventure, Astronaut, or (insert name of your favorite prerendered
dome show here) frames in "real-time" without visual compromises. It
would have to be a more efficient platform to produce upon than the
applications that those shows were produced with (3dsMax, Maya,
custom supercomputer apps, etc.), and the resulting production would
have to be as portable as prerendered content is now. To my
knowledge, this doesn't even come close to describing any system that
currently exists or will exist in the forseeable future. I'm not
saying that this can never happen, but ever since high-end graphics
technology became a commodity item, Nvidia, ATI and the like have
been trying to convince us that the next generation of their
technology will bring cinema quality graphics in real-time. So far,
they've always been at least 10 years behind. On the tech side, what
happens is that new methods are pioneered in software labs at
universities or the R&D teams at animation or FX studios. These
methods and tools quickly show up in animation software, then in the
cinema. People's expectations are raised, and then finally, the
propeller-heads at the video card companies and the engine
programmers at game companies figure out ways to accelerate these new
methods via new hardware. I could mention Blinn's law here as well,
which is a long standing maxim credited to CG pioneer Jim Blinn that
states “any renderer, no matter how fast processors get, will always
take a couple of hours, because that's the tolerance level of artists."
Furthermore, I find that people often don't recognize that the same
graphics technology that powers all the cool real-time digital dome
systems out there, also powers 3dsMax, Maya, FinalCut, AfterEffects,
etc. I have a real-time viewport where I can spin around, set up my
animations, press play, and immediately see the results. If I've
kept things basic, I don't have to wait for a render to know what
things will look like. It's only when I start taking my visual
fidelity or complexity beyond what real-time graphics systems are
capable of that I have to do test renders to see the results. Also,
what it means to "keep things basic" is evolving at the same pace as
real-time graphics technology. Finally, these tools have been
hammered on and optimized for efficiency of production much more, and
by many MANY more people than the latest narrowly focused tool that
us dome-specific guys/gals invent.
In the end, it all comes down to the question of which tools will
help me tell my story the best (highest quality, fastest, most
efficient, etc.) in the dome. Software applications usually have to
make a choice between simplicity and power. If I only ever produced
"the sky tonight" type talks, then I'd probably choose something like
Starry Night, D3, or one of the other simple streamlined focused
applications for doing that (unless I'm concerned about wide
distribution.) If I wanted to do something truely interactive, then
obviously a real-time application is the only choice. If I want to
craft a linear cinematic experience with no limitations except my own
imagination, budget and ability, then the hundreds of tools that fall
into the "prerendered" category are the only logical choice.
Remember, all of these things are just tools to create the
experience. Understand and use your tools or else they will use you.
Brad Thompson - bthompson@...
Digital Animation & Design - Spitz, Inc.
-- "Hush, may I ask you all for silence? The dreamer is still asleep"
- I'll echo others' sentiments about leaving room for both real-time
and rendered technologies in our domes, in particular, Todd Slisher's
> I'm a big believer that both interactive and pre-rendered cinematicAgreed!
> experiences should be part of our virtual 'bag of tricks' that we
> can deliver to our audiences. Let's have the best both worlds can
In many ways, the two are apples and oranges. I think AMNH's attempt
to proceed down both routes shows how we value the differing
experiences they can offer, even though our 430-seat theater doesn't
give us the same kind of opportunities for real-time experimentation
that a smaller, classroom-sized theater would (which is why we need
to build a second theater, in my opinion, but that's another story).
However, the "apples and oranges" mentality doesn't do the
relationship justice. Our monthly real-time program, "Virtual
Universe," has given us an opportunity to explore topics in the dome
that have in turn informed our big-budget "space shows." One huge
benefit is that the same people who create the shows (e.g., Carter
Emmart and I) also share responsibilities for the monthly program.
So we get to experience first-hand audience reactions to our work,
tweak and modify the presentation both visually and verbally, and
basically play with our ideas before they appear in a more formal,
(Using my preferred term "narrative journey" for immersive
experiences, the difference is a bit like the slick, pre-recorded
tour versus the friend pointing out places of interest. A well-
informed friend, with enough experience, can provide some insight
into improving the pre-produced version.)
We also use our dome's real-time capabilities for testing and
creating flight paths. In particular, we were able to collaborate
with NCSA over the phone as we shared the same virtual environment:
they would make a flight path, share it with us, then we could fly it
together through an identical virtual space. But this idea could be
carried further... If a production work flow were designed
correctly, it seems that one could effectively create a real-time
version of a show (or portions thereof), test it with audiences, then
use that feedback to inform development of a final production. Seems
like it would be worth a try, at any rate. N.B., however, that real-
time needs must be considered specifically in advance of a
production; done correctly, they can proceed naturally from work done
for the rendered work, but that doesn't happen for free or without an
investment of time and resources.
Finally, I see potential for real-time show distribution. Because
flightpaths and models could be downloaded off an FTP site or sent on
a DVD-R, they represent a much more (and quite literally)
"lightweight" means of distributing content than shipping hard drives
around the world. As real-time engines continue to improve, there
are many things they do just as well as pre-rendered media (e.g.,
most solar system shows, some general astronomical content), so why
not save on the FedEx bill? Perhaps one could even disassemble the
show kits, in fine old planetarium tradition, to create mini-shows or
portions of a program that could be used in different contexts?
Just my $0.02.
Ryan Wyatt, Science Visualizer
Rose Center for Earth & Space
American Museum of Natural History
79th Street at Central Park West
New York, NY 10024