## Re: [fulldome] Re: 4k or larger Star Fields

Expand Messages
• ... Ideally you would want a projection system that can handle both daytime and nighttime scenes. Our visual perception has a range of at least 10 orders of
Message 1 of 11 , Jun 12 11:01 PM
Chris Anderson wrote:
> Tom Casey wrote:
>> On Jun 11, 2008, at 6:23 PM, Chris Anderson wrote:
>> > How does this relate to full dome? In a perfect world, we'd like a
>> > digital version that duplicates this sort of resolution. So I thought
>> > I'd do a little back of the envelope calculation:
>> While this sort of math always fascinates me, in the real world we
>> found that any of our arithmetic rationalizations for creating our
>> digital starfields was, in the end, pretty meaningless.
>
> Well, that's why I started with the caveat that it was "back of the
> envelope."
> It was intended as a "let's see if we're in the ball park" sort of
> calculation.
>
>> The pixels
>> to optical comparison is pretty much in the eye of the beholder as
>> they say. A star cannot be a single pixel and look good projected.
>
> True. But perhaps this should read "_With current projection technology_
> a star cannot be a single pixel and look good projected."
>
> Ultimately, to produce the most realistic star field, we need a technology
> that projects sub-arcminute stars whose brightnesses span a dynamic
> range of ~2.5 million. While that's not possible today, it may be in the
> not-too-distant future. (I suspect that to pull off the dynamic range,
> we're going to need a very bright source (laser?) projecting onto a nearly
> black dome.)

Ideally you would want a projection system that can handle both
daytime and nighttime scenes. Our visual perception has a range of at
least 10 orders of magnitude; that is, our light perception from the
faintest intensities (roughly 10^-5 candela/sq. meter) to the
brightest (10^5 candela/sq. meter and greater) is different by a
factor of 10^10. The Sun is actually much brighter than 10^5
candelas/sq. meter, and I have seen claims that our sensistivity at
the high end is actually around 10^14 candelas/sq. meter. However
sources that bright should immediately blind you so we can ignore them
in our specifications for the ultimate virtual reality display.

At any one time, we can perceive a dynamic range of only about 5
orders of magnitude. That is the ratio of the brightest to the
faintest within a scene is only a factor of 100,000. The only HDR
displays that I know of are the Brightside monitors, which actually
achieve this ratio, with a maximum brightness of 3000 candelas/sq.
meter. Staring at one of these monitors can literally be blinding.
BrightSide displays can also show true black so they would be ideal
for the black night sky purists. Thus by re-normalizing the
brightness of a scene to match the display output (hence not showing
absolute intensity correctly), this technology can show both dim and
bright scenes quite well.

The images that BrightSide monitors produce are pretty stunning.
Comparing one to a normal CRT is like looking out a window and staring
at a postcard of the same scene. However although BrightSide has been
acquired by Dolby, they are still aiming for a very high-end market,
if only because their monitors are practically hand-built. They are a
long ways off from mass producing these things to the point where the
prices can drop significantly.

When such technologies might appear in the projector market is
completely unknown. Of course any projection onto a dome will still
be contrast limited because of cross bounce. For this to not be an
issue, you will have to wait for some new generation of flexible
OLED-based display. Those don't exist widely in even standard dynamic
range form, so it'll be a while before the HDR ones appear.

>> Stars of various magnitudes end up needing to be a soft "blob" of
>> pixels that the viewer's eye resolves as a star of a certain
>> brightness... different blob, different magnitude.
>
> An imperfect compromise, to be sure.

As long as your "star" appears as a point-source, with a size that's
below the eye's resolution limit (1 arcminute), and if your
surrounding background is pure black, then you should be simulating
how a real star in the sky would be perceived. I'm not sure how
realistic rendering via "soft blobs" is. Although the popular
planetarium programs like to use variable sized "fuzzy" star textures
to show the brightest stars, informal surveys suggest that the more
point-like your star, the more realistic it will look from the night
sky perspective.

(I also suspect that at low light levels, the use of scotopic vision
-- mostly rods instead of cones in your retina -- must have some

A quick back-of-the-envelope calculation will show how many pixels you
need to be eye-resolution limited. Assume Nyquist sampling so that
each pixel is 0.5 arcmin across. Since there are 1.1644E8/pi^2 sq.
arcminutes per steradian, a hemisphere will require just under
300,000,000 pixels.

...

But visual perception is a whole lot more than just pixels to match a
resolution goal. Even our sense of the three-dimensional world cannot
be completely simulated via only a stereoscopic display; there's a lot
more to it than that. Until the day that you can stream stimuli via
electrical signals directly into your brain, any virtual simulation
will always have subtle discrepancies when compared to the real thing.

--kachun

Dr. Ka Chun Yu / Curator of Space Science
Denver Museum of Nature & Science
2001 Colorado Blvd., Denver, CO 80205
P: 303-370-6394 / F: 303-331-6492
E: kcyu@...
W: https://scientists.dmns.org/sites/kachunyu/
----------------------------------------------------------------------
Your message has been successfully submitted and would be delivered to recipients shortly.