Hi all,
after going again through my previous globular patch code with special diligence, doing some further careful size comparisons (see below), and after some code modifications, here is the new resulting patch for download and testing. Again, the patch archive contains both a standard patch file and the individual modified C++ files as of the current
rev 4606 for a simple replacement.
++++++++++++++++++++++++++++++++++
http://www.celestiaproject.net/~t00fri/images/ ... _r4606.zip ++++++++++++++++++++++++++++++++++
There was nothing much wrong previously, but there were a number of striking issues to be understood (like the above apparent M 13 size discrepancy)...Nevertheless I added several substantial code modifications after further thinking and understanding what is going on:
From my previous direct tests, there was little remaining doubt that the
globular scale factor was correct. So there must be other reasons for the considerably smaller apparent globular sizes from my code, as compared to some photographs, notably from big telescopes. Some reasons I mentioned already previously, like many very faint cluster stars that connot be rendered in Celestia etc.
- Clearly, if the radial globular luminosity distribution was falling off too steeply, the resulting globulars would appear too small as well.
After plenty of further thinking and reading, I decided to replace my previously used radial King profile for 3d space by the (2d) skyplane projection of it. This form is mostly used by researches (King, Harris,...) in earthbound comparisons with globular observations. It indeed seems like the best approach, notably when the observer is far away from the globular. I had used the projected form already for the central cloud texture for that reason. One immediate benefit is that the skyplane projection is considerably shallower, hence more off-center cluster stars will become generated and then rendered.
Yesterday, I first derived the modified math for the corresponding random generation of sprite stars according to the skyplane projected profile. Fortunately, all the previous "high-efficiency" tricks turned out to work as well, and that most effectively. I put up a commented Maple file, in case someone is interested how the used algorithm was actually derived.
http://www.celestiaproject.net/~t00fri/images/King_profile.pdf
- A modification which actually brought most improvement as to the apparent globular sizes was to make the size of the cluster stars proportional to the brightness parameter (adjustable via the ' [ ', ' ] ' keys)! This simple modification now allows to further enhance the brightness of the cluster stars in sync with that of the other stars in Celestia. The result is quite striking: if the magnitude threshold is set high enough via the ' ] ' key, many more peripheral stars are rendered now and the diameter of the globulars appears to substantially increase! This confirms the first suspicion I had and thus delegates the choice of brightness to the user's gusto.
I recommend automag=ON with Auto magnitude limit at 45 degrees: 7.5
This is good for both galaxies and globulars!
In addition, I used a new procedural texture profile for the individual cluster stars:
Instead of the previous Gaussian
Code: Select all
lumi = exp(-2.0f * (u * u + v * v)) - exp(-2.0f);
that is shallow around its maximum (u=v=0), I now tried this exponential type
Code: Select all
lumi = exp(- LumiShape * sqrt(u * u + v * v)) - exp(-LumiShape); (u^2 + v^2<=1)
This shape has a sharp & small bright center. The bright (big) cluster stars even appear to have sort of a halo around the small bright core. The appearance seems way more "starlike". The orange Red Giant stars now look much smaller (but brighter) to the observer flying through the globular core (Hey Martin!).
Note that the geometrical size of the cluster stars is strongly reduced via the convolution with the modified star texture shape! The reduction amounts to a factor of ~ 0.2 times the geometrical size! Despite better visibility of dim stars, the effective cluster size is rather smaller than before...
Let me know how you like these new cluster star features! It's easy to restore the old ones, if people prefer them 
. Altogether: one main reason why the clusters looked too small, was due to my desire to make the geometrical star sizes as small as possible, in order to come closer to the actual diameters... But that prevented quite a few peripheral stars from being rendered!
- Actually, since the globulars appear now substantially larger, I returned to only ONE globular size definition, the mu25 isophote radius! Hence the radius and apparent (angular) diameter values you read on the top left of the canvas now again correspond to the size that is graphically indicated by the 4 red selection cursor triangles. That's much more elegant, of course.
Please give me feedback about this choice, too!
Finally, let me discuss some crucial further size tests I did:
A) First consider M 4. Let us compare with this accurate summary of the various sizes of M 4 depending of how it is observed:
http://www.seds.org/messier/m/m004.html Here is an important summary quote (with given values cross checked from other sources):
SEDS page wrote:Its (M 4's) angular diameter, seen on deep photographs, is about 36 minutes of arc, more than that of the Full Moon; this corresponds to a linear diameter of about 75 light years. On typical photos it appears somewhat smaller at about 26', and visually it was estiamted at 14 arc minutes. Its tidal radius, determined by the distance where tidal gravitational forces of the Milky Way Galaxy would cause member stars to escape, is estimated at 32.49', or about 70 light-years, so that this globular gravitationally dominates a spherical volume 140 light-years in diameter.
M4 is one of the most open, or loose, globulars, as its classification in concentration class IX indicates. Its compressed central core was measured at 1.66' diameter, or linearly 3.6 light-years. Its half-mass radius is 3.65' or about 8 light-years, so half the cluster's mass is concentrated in an inner spherical volume of 16 light-years diameter.
In the first image of M 4 and the Moon from Celestia with my new patch, I have added various circles corresponding to the various possible ways of observation.
In this shot, the app. magnitude has been set only to VISUAL, i.e. m=6.5. The green circle corresponds to the visual size estimate of M 4 (14'), and you can see how well it matches!! Note also (pink!) the tiny central core diameter of only 1.66'...
Next, we turn up the brightness (' ] ' key) to maximum (with automag=
ON, of course) and consider the next test image:
You can see that NOW the M 4 stars almost completely fill up the 36' circle corresponding to the mu25 isophote radius! Hence in this "telescope view", the size of M 4 and the full Moon are certainly comparable.
In the two images you also can see that I have reverted to the mu25 isophote radius for the 4 red selection cursor triangles (as announced above).
B) In the next image of M 22 with brightness at maximum, you can see how well the orange mu25 isophote circle is now filled with stars. For globulars (like M 22) with relatively low King concentration c, the peripheral stars are particularly conspicuous, since the luminosity profile is comparatively shallow.
If brightness is
down to a recommended m=7.5, the apparent M 22 size is clearly somewhat smaller.
C) The final comparison image concerns the apparent size of M 13 relative to some environment marks, notably compared to the distances to the two bright stars
Hip 81848 and HIP 81673. I was totally confused 2 days ago, when I had discovered the corresponding beautiful (but
highly overexposed) above photo, taken with a BIG telescope. The previous M 13 size in Celestia was seen to be much smaller...
Again note the orange mu25 isophote radius circle along with the tiny pink circle corresponding to the core radius. Brightness is turned up to maximum and indeed you see that a substantial portion of the mu25 isophote circle is filled with cluster stars.
So things seem pretty consistent now! Note however, that for these comparisons I used 65536 stars rather than only 8192. That's one "abstraction" we have to live with, but I wanted to demonstrate that the underlying math is correct.
Let me know...
Fridger
