Discussion of possible rendering of Pulsars

[Malenfant]

Don't get me wrong, I think this would be interesting to see in general... but not now, and most likely not in this program. There are a lot of other things that need to be sorted out in Celestia, and I think we need to draw a line between what Celestia needs be able to illustrate and what it doesn't need to illustrate.

I just think that redshifting and gravity distortions are not something that need to be shown in Celestia.

[t00fri]

Fridger,

Don't forget that objects are illuminated by reflected light as well as direct light. I suspect Chris might be thinking about how to implement Saturn-shine on Mimas (just a random example)

Having the code for multiple light sources in place will make this possible, and, I suspect, might have been easier to implement initially for direct illumination.

And, of course, there are exoplanets being found in multi-star systems.

Sure enough Selden.

But with the pulsars we are talking about 1500 most fascinating objects, not 10 or 20...All I am saying is that the multiple-star illumination stuff took Chris a long time to develop and debug, while the number of scientific applications is way smaller than what one could visualize with 1500 pulsars initially and in a second stage perhaps with lensing applications...and possibly black holes.

I just want to see a sufficiently large set of complete published multiple-star orbits with planets to which Chris code can be applied.... .

If --in sync with Chris-- we go for pulsars with some generic texture and other features, these require us to get so close that ignoring the vast light deflections would be an intolerable physical mistake.

Bye Fridger

[Malenfant]

But with the pulsars we are talking about 1500 most fascinating objects, not 10 or 20...All I am saying is that the multiple-star illumination stuff took Chris a long time to develop and debug, while the number of scientific applications is way smaller than what one could visualize with 1500 pulsars initially and in a second stage perhaps with lensing applications...and possibly black holes.

Again, you're not looking at the big picture.

If people go to a multiple star system with a planet and no multiple lighting, they're going to wonder why it's only lit by one star. As it is, there's plenty of educational value in multiple star rendering (which you're obviously ignoring) - you can use it to illustrate the effect of different luminosities at different distances from a planet, how illumination changes over time, the Flammarion effect (if the stars are different colours), even how the surface temperature would vary (if that was implemented realistically too).

And more importantly, multiple star rendering would render a commonly-encountered situation in the universe realistically. Multiple star systems are a LOT more common than pulsars and blackholes, which only form as a result of the deaths of the most massive, rarest stars anyway.

Meanwhile, there are no "scientific applications" at all when it comes to showing distortions around neutron stars or black holes in Celestia, for the simple reason that at best you're just showing a generalisation. We don't have a Mass term in Celestia, we don't even know the masses for most of these objects anyway, we don't know their orientations etc... so at best after a lot of effort we'd be showing a generalised distortion that took a lot of work to implement. Yeah, that's really useful.

I'm all for including the pulsar catalogue, because that IS actually useful. I just don't see any need to go further than that.

If --in sync with Chris-- we go for pulsars with some generic texture and other features, these require us to get so close that ignoring the light deflections would be an intolerable physical mistake.

We'd be showing them with generic textures for crying out loud, it's quite obvious that the things aren't being shown realistically. So making a fuss that the gravity distortions aren't there is ridiculous, because it's already unrealistic. How can you turn a blind eye to the appearance of the object while insisting that its distortions are shown accurately?! The only real purpose of showing the pulsar catalogue is simply to show where the objects are. Everything else is just gravy.

[t00fri]

But with the pulsars we are talking about 1500 most fascinating objects, not 10 or 20...All I am saying is that the multiple-star illumination stuff took Chris a long time to develop and debug, while the number of scientific applications is way smaller than what one could visualize with 1500 pulsars initially and in a second stage perhaps with lensing applications...and possibly black holes.

Bye Fridger

Again, you're not looking at the big picture.

If people go to a multiple star system with a planet and no multiple lighting, they're going to wonder why it's only lit by one star. As it is, there's plenty of educational value in multiple star rendering (which you're obviously ignoring) - you can use it to illustrate the effect of different luminosities at different distances from a planet, how illumination changes over time, the Flammarion effect (if the stars are different colours), even how the surface temperature would vary (if that was implemented realistically too).

And more importantly, multiple star rendering would render a commonly-encountered situation in the universe realistically. Multiple star systems are a LOT more common than pulsars and blackholes, which only form as a result of the deaths of the most massive, rarest stars anyway.

Meanwhile, there are no "scientific applications" at all when it comes to showing distortions around neutron stars or black holes in Celestia, for the simple reason that at best you're just showing a generalisation. We don't have a Mass term in Celestia, we don't even know the masses for most of these objects anyway, we don't know their orientations etc... so at best after a lot of effort we'd be showing a generalised distortion that took a lot of work to implement. Yeah, that's really useful.

I'm all for including the pulsar catalogue, because that IS actually useful. I just don't see any need to go further than that.

Not again the business with the mass, please. I have even quoted scientific review talks and original papers earlier where it was demonstrated that pulsar masses can be measured quantitatively and that we do have actual data. On the other hand no multiple star system can be set up in Celestia without ALL masses being known! These data are lacking.

Malenfant you are cutting off your own finger

[Malenfant]

Not again the business with the mass, please. I have even quoted scientific review talks and original papers earlier where it was demonstrated that pulsar masses can be measured quantitatively and that we do have actual data. On the other hand no multiple star system can be set up in Celestia without ALL masses being known! These data are lacking.

Oh, big bloody deal. We don't know what pulsars look like close up, we don't know their orientations either, and we lack lots of other data about them too. So your argument against the multiple stars applies equally to pulsars too.

Plus, Celestia currently supports multiple lighting. It doesn't currently support mass. And if we add mass for pulsars, then we should damn well add mass for everything else.

I also note that you ignore the actual point I'm making, as usual. It's funny, when you can't actually defend your own arguments you just resort to attacking arguments against them to try to make them look worse than your own.

Frankly, I'm thinking a simpler solution is to just not include any pulsar catalogues at all.

[t00fri]

As a co-author of Celestia I would definitely veto to render pulsar surfaces without taking into account light deflection. If Chris (not YOU ) decides against implementing light deflection, it's fine with me. BUT then
we MUST renounce rendering pulsar surfaces as tiny as 10KM in size for consistency!

This I disagree with . . . I think it's fine to proceed toward realism in small steps. With an all-or-nothing approach, development of Celestia wouldn't have gotten very far. I argue that we do as much as possible with the current renderer, then improve upon it as time permits.

--Chris

[t00fri]

....
Plus, Celestia currently supports multiple lighting. It doesn't currently support mass. And if we add mass for pulsars, then we should damn well add mass for everything else.
....

It was me who extracted all available binary orbits from catalogs, where all parameters INCLUDING their masses are entered, of course. So what are you talking about??

But these are just binary orbits where multiple illumination does NOT play any role! In the cases of >= 2 suns with planets we DO NOT know all masses, not even qualitatively. So there are effectively no applications left where the orbit parameters are fully known. That's what you were referring to with "important everyday applications".

[t00fri]

Frankly, I'm thinking a simpler solution is to just not include any pulsar catalogues at all.

Frankly, the devs so far have always found a viable compromise.

@Chris,

I would strongly be in favor to retreat to a "quiter place" where actually facts can be used as arguments, as usual.

[chris]

I can't say that I'm that well acquainted with general relativity, but I think that I understand it well enough to say why I think that depicting the gravitational deflection of light might be difficult.

Light follows paths called null geodesics. In flat space, unperturbed by large masses, these are straight lines. Given a point in space, you can figure out where it will appear on the view plane (i.e. the screen) by computing the intersection of the view plane with the line between the point and the eye. This is a very simple operation, and it runs very quickly on modern graphics hardware.

When space is perturbed a by a large concentration of mass, the null geodesic is no longer a straight line, so computing intersections of null geodesics and the view plane becomes much more difficult. I suspect that only for certain restricted cases will it be possible to calculate a nontrivial number of intersections at interactive frame rates:

1. Only light rays which directly reach the viewer are considered
2. The objects emitting the light rays are at distances much greater than the distance between the viewer and the distorting mass
3. There is a single distorting mass in a simple and highly symmetric configuration

An example situation where all three conditions are satisfied is an observer stargazing in the vicinity of a neutron star or black hole. The mass disitribution is as symmetrical as you can get: a sphere, unless it's spinning very rapidly. The stars are distant, so we know the direction from which their light is arriving. A star close to the large mass would emit light in all directions, and there are many paths that could reach the observer. Treating objects as being at an infinite distance greatly simplifies the problem.

With a single mass, we can get an approximate solution by computing the deflection angle of light rays due to the mass, then computing the intersection of the deflected rays with the view plane. If the viewer is close to the mass, we'd need more than the deflection angle--we actually need to calculate the intersection curving part of the geodesic with the view plane. I don't know how difficult this would be.

With multiple masses, you might be able to develop an approximate solution by calculating the deflection angle from the more distant mass, and then deflect that deflected ray. If the masses are sufficiently close that the geodesic isn't effectively a straight line by the time it passes near the second mass, the approximation falls apart, and we're in for disappointment if we hope render the scene accurately in realtime.

--Chris

[Malenfant]

Oh, ffs, I give up. Talking to you is like talking to a brick wall... but less productive.

I'm using practicality as an argument here. You're just insisting that your "facts" are more important and ignoring them. I swear your obsession and blind insistence that everything should be 100% physically accurate is going to destroy this program. Luckily Chris has more sense than just to listen to your opinions alone.

And masses are not supported in Celestia. You can include mass in the ssc, but it doesn't do a damn thing.

But whatever - go away to your "quiet place" where you can just impose your arguments on people without contradiction or dissent, because I think that's what you're really after here... you are clealry not comfortable when people argue with you because you think that only purely scientific arguments (and only ones that come from you) are valid. And that is not the way to debate anything.

[Malenfant]

As a co-author of Celestia I would definitely veto to render pulsar surfaces without taking into account light deflection. If Chris (not YOU Wink ) decides against implementing light deflection, it's fine with me. BUT then
we MUST renounce rendering pulsar surfaces as tiny as 10KM in size for consistency!

Another "flawed" argument from you. And who the hell said you have a veto in anything at all? Chris decides, not you. If you don't like it, go away.

But you'd actually rather not show any pulsars at all because you don't like the idea of showing a generic surface? Utterly ridiculous. You can get useful info out of showing their distribution in the galaxy, frankly whether they're rendered realistically in that case is completely irrelevant. I'd rather have some useful information than none at all, but evidently you're too busy sticking to your ridiculously impractical principles of "100% realistic or nothing at all". How very constructive of you.

In that case, why did you bother with the binary star catalogues? We don't know everything about those, why did you allow them to be incorporated into Celestia? And how about your galaxies? They're not all rendered exactly like they are in reality either. So by saying that you'd "veto" pulsars that don't have realistic surfaces, you're just being really hypocritical here. While your other contributions have been as realistic as the source catalogues they are still not 100% realistic, so pulsars certainly don't need to be either.

[t00fri]

I can't say that I'm that well acquainted with general relativity, but I think that I understand it well enough to say why I think that depicting the gravitational deflection of light might be difficult.

Light follows paths called null geodesics. In flat space, unperturbed by large masses, these are straight lines. Given a point in space, you can figure out where it will appear on the view plane (i.e. the screen) by computing the intersection of the view plane with the line between the point and the eye. This is a very simple operation, and it runs very quickly on modern graphics hardware.

When space is perturbed a by a large concentration of mass, the null geodesic is no longer a straight line, so computing intersections of null geodesics and the view plane becomes much more difficult. I suspect that only for certain restricted cases will it be possible to calculate a nontrivial number of intersections at interactive frame rates:

1. Only light rays which directly reach the viewer are considered
2. The objects emitting the light rays are at distances much greater than the distance between the viewer and the distorting mass
3. There is a single distorting mass in a simple and highly symmetric configuration

An example situation where all three conditions are satisfied is an observer stargazing in the vicinity of a neutron star or black hole. The mass disitribution is as symmetrical as you can get: a sphere, unless it's spinning very rapidly. The stars are distant, so we know the direction from which their light is arriving. A star close to the large mass would emit light in all directions, and there are many paths that could reach the observer. Treating objects as being at an infinite distance greatly simplifies the problem.

With a single mass, we can get an approximate solution by computing the deflection angle of light rays due to the mass, then computing the intersection of the deflected rays with the view plane. If the viewer is close to the mass, we'd need more than the deflection angle--we actually need to calculate the intersection curving part of the geodesic with the view plane. I don't know how difficult this would be.

With multiple masses, you might be able to develop an approximate solution by calculating the deflection angle from the more distant mass, and then deflect that deflected ray. If the masses are sufficiently close that the geodesic isn't effectively a straight line by the time it passes near the second mass, the approximation falls apart, and we're in for disappointment if we hope render the scene accurately in realtime.

--Chris

Chris,

yes, I can agree with your summary qualitatively. In fact, the situation is not unlike optical scattering in the presence of a (variable) refraction index.

We know that normal optical scattering can be pretty complicated in the most general case. Then we have to calculate the full Mie sum and that is getting hard when k*R>>1, for example. Fortunately, also in optics, there are a number of dramatic simplifications that allow you to write down good approximations for atmospheres and to arrive at surprisingly simple results for diffraction and refraction, /depending/ strongly on the value of k*R as well as the distance of the observer and the source, respectively, from the medium. Then familiar regimes appear like Fresnel and Fraunhofer etc....

Similar simplications also exist in lensing or light deflection of GR, along the lines you indicated. These simplifications have to be exploited, of course, in order to make things sufficiently fast. Yet, as pecularities of GR, we will have to take into account the 'horizon' and the 'photon sphere' etc....

As I wrote earlier, the main concern is to impose certain boundaries in the code where we know that the usual approximations are loosing their validity. That can be done since the theoretical framework is under control. Another matter will be how to "hide" these boundaries in a clever way to the users. But really, that does not concern me much at this stage...

Just as an aside: geodesic paths are VERY intimately related to a formulation in terms of action-angle variables. This is uniformly true for classical mechanics, quantum mechanics and GR . Long time ago, when I was still a "baby physicist", I wrote a long paper with my wife, involving a very rewarding use of action-angle variables. That's why I am still so fond of them...

Of course when proposing those variables the other day, I was thinking already of a geodetic description as required in GR. You did not respond to that post, but I suppose you read it at least...

Bye Fridger

[t00fri]

[
....
Another stupid argument from you. And who the hell said you have a veto in anything at all? Chris decides, not you. If you don't like it, go away.

OK, what do you think co-authorship is worth otherwise? Go and have a look in your version of Celestia, who its authors are. Anyhow, I suppose this is not of your business...

I'll also write a PM to Selden concerning your language above. I am not usually presenting stupid arguments.

[Malenfant]

OK, what do you think co-authorship is worth otherwise? Go and have a look in your version of Celestia, who its authors are. Anyhow, I suppose this is not of your business...

I'll also write a PM to Selden concerning your language above. I am not usually present stupid arguments.

Your arguments are frequently badly flawed since you only see things from your own limited perspective. Go right ahead and whine to Selden, it doesn't change the fact that my point is valid though.

And last time I looked, Chris had final say on anything. He can listen to other authors' opinions but you don't have any "veto" right, though you do try your best with your usual threats and tantrums if it doesn't look like you're going to get your way. But you have no more veto right on anything than any other developer who isn't Chris does. "co-authorship" means nothing at all beyond the fact that you've contributed something to the program.

My views on this matter are - as usual - fairly balanced. I think it's worth adding a pulsar catalogue just to show their general locations and I do think it's worth using a 'placeholder' texture since we obviously don't know what they look like. We can make a reasonable guess based on their temperatures though. This is good enough for now, and certainly as good as any other catalogues that have been implemented. However, I certainly don't believe that gravitational distortions and other relativistic effects are worth implementing at this stage, and I don't believe they should be a priority at all right now given other aspects that need to be worked on. I don't believe they are necessary and I don't believe they will add enough value to be worth the effort. MAYBE later on, once other more important aspects of Celestia have been completed it might be worth doing if it fits with the program's general direction. That is my opinion, and it is based on general practicality.

However, your views are not balanced or reasonable, or even particularly rational. You can throw "facts" around all you like, but there is more than just scientific accuracy at stake here, there's the practicality of coding it, of integrating into the program, of how actually useful it will be, of how often it will crop up, whether it's more important than other aspects of the code to warrant not spending time on those etc. That is the big picture that you fail to account for. Yet you insist - as usual - that we have to do it your way or no way at all. This is not the behaviour of anyone interested in compromise or reasoned debate.

[Cham]

Just an idea :

There's another feature of pulsars (and black holes) which could be implemented without the complexity of light deviation : gravitational redshift. The perceived color depends on the observer's position (Doppler shift depends on the relative velocity, and I'm not including it here). Suppose the object is emitting a typical wavelengh Lambda0. If we assume simple spherical symmetry (and neglect all effects of rotation and frame dragging), the wavelenght perceived at the location of some stationary observer is

Code: Select all

Lambda = Lambda0 * Sqrt[(1 - 2GM / r c^2)/(1 - 2GM / r0 c^2)]



where "r" is (roughly) the distance between the object's center and the stationary observer. Maybe this could be coded ?

[Cham]

Here's a web page with a good video showing a supernova followed by the formation of a pulsar, with the light bending effect. The video is very well done. I suggest you download the large version (broadcast quality).

http://www.spitzer.caltech.edu/Media/re ... 10v1.shtml

[ajtribick]

Hmmm... while I wasn't intending to be back so soon, I confess to a bit of lurking, and this thread seemed extremely interesting.

This may be useful: document describing the creation of a raytracer for a Schwarzschild metric.

However given pulsars rotate so fast (especially millisecond pulsars), I'm wondering whether you'd notice significant deviations from the view around a non-rotating mass. I'd suspect that raytracing with a Kerr metric would be a lot more difficult than a Schwarzschild one.

[t00fri]

Hmmm... while I wasn't intending to be back so soon, I confess to a bit of lurking, and this thread seemed extremely interesting.

This may be useful: document describing the creation of a raytracer for a Schwarzschild metric.

However given pulsars rotate so fast (especially millisecond pulsars), I'm wondering whether you'd notice significant deviations from the view around a non-rotating mass. I'd suspect that raytracing with a Kerr metric would be a lot more difficult than a Schwarzschild one.

Good you find this subject interesting. The thesis may well be useful for practical considerations.

I hope also Chris will eventually find light deflection and other such phenomena a bit more challenging than spacecraft obits

Bye Fridger