Raytracing a Black Hole | Hacker News Reader

Raytracing a Black Hole

rantonels.github.io

329 pointspablode33 comments

I worked at dneg when this very shot was being rendered.

From what I remeber there were a few steps to this being rendered, the first part was a massive particle sim. I was a systems engineer, so my job was to keep the system running.

That particle sim ate at least 7, if not more hard disks. What ever it did, it destroyed 8tb HDDs. The worst part was we couldn't take that file server offline, as the render needed to finish. But, not only was the render hammering and eating disks, the raid array was desperatly trying to rebuild the groups at the same time.

As soon as the group was rebuilt, another disk would fail.

so much joy. As soon as we had some slack, we migrated away from that server and gave something less stressful to do.

AFAICT the parent comment is referring to Double Negative (dneg). The visual effects company responsible for the black hole shots in the movie Interstellar.

I so love this kind of behind the scene story

You're talking about a different black hole render. This is an original project that runs on consumer hardware.

....yes, yes I am.

Sometimes I think about what kind of yak shaving it takes to put together some menial technical detail, and I just throw the idea out, if it doesn't feel like the implementation is clean enough.

If someone told me that in order to take a picture of some natural landscape, I'd have to throw 7 brand new 8TB hard drives in the garbage, in order to capture a 64 megapixel image, I'd say "not worth it."

It challenges one's natural disinclination towards waste. I think that's a pretty normal gut response. Gifted with an a priori awareness of certainty of return on investment, it makes sense dive in, get knee deep, ignore the intuition that wells up around thoughts of sunk cost fallacies, and push through the hard parts. But when you're not a fortune teller, that risk aversion tends to help, more than harm.

But, you know, maybe that's the kind of thing that separates people like me from true success.

To put this in more context:

at the time there were 32 file servers, each with a 60 disk array attached. Thats almost 2000 spinny disks.

At that scale we'd expect about one failing disk every two weeks+. We were covered by a 24/7/365 4 hour warranty. This means that a replacement disk will appear within 4 hours, anytime of the year.

From what I understand the old disks are taken back, examined, refurbished and sent out again. I've heard that its often head alignmet, and or track setup issues that cause the most problems.

But yeah, there is waste....

+in practice we had a lot more disks than this, as we needed fast online backup. so we'd normally have at least two failed disks a week.

And in the coming years astronomers hope to image a black hole directly (in radio wavelengths). The few closest supermassive black holes (in center of galaxies) have their event horizon angular diameters just big enough to be resolved by VLBI (very-long-baseline interferometry). See [3] for details and a simulated image.

[1] https://en.wikipedia.org/wiki/Event_Horizon_Telescope

[2] https://eventhorizontelescope.org/

[3] Seeing Black Holes: from the Computer to the Telescope (https://arxiv.org/pdf/1804.03909.pdf)

I believe this is supposed to happen this year, last I heard was this summer but they haven't finished the analysis yet. The data was collected back in April, petabytes of harddrives flown to a central processing location.

That puts the other comment talking about 7x8TB harddrives being devoured in rendering the Interstellar scene a bit into perspective

On a related note, there's a paper by the Interstellar special effects team and Nobel laureate Kip Thorne describing how they rendered the black hole in the film and the tradeoffs they made:

https://arxiv.org/abs/1502.03808

It's amazing that they put so much attention to detail in the same movie where a botched docking maneuver causes an orbiting spacecraft to fall out of the sky, and in which a black hole opens the way for programing mechanical watches from the past.

It is and I hate that movie for it. In theory I should love it, but the disconnect between trying to be realistic and totally ignoring reality, killed it for me. I usually love sci-fi movies, realistic or not. But I mean, if you're going to put so much work into making a main element of the movie seem physically correct and realistic, why wouldn't you put at least a little effort into making the rest of the movie somewhat believable. The movie goes between metaphysical, realistic and action-movie physics every other scene.

Please don't kill this movie for me. It still has rewatch value for spectacles. :)

Compare Jean-Pierre Luminet's https://arxiv.org/abs/1804.03909 which discusses Interstellar starting at the bottom of the third page of the pdf.

For anyone interested, here are some videos of what it's like to fall into a black hole together with some very good explanations:

- https://jila.colorado.edu/~ajsh/insidebh/index.html

- https://physics.stackexchange.com/questions/26185/what-will-...

> Here we have an infinitely thin, flat, horizontal accretion disk extending from the photon sphere (this is very unrealistic, orbits below 3rS are unstable. More below) to 4 radii, coloured checkered white and blue on the top and white and green on the bottom. It is evident, with this colouring, that we've encountered another case of seeing 100% of something at the same time.

> For this image, I moved the observer up a bit, so he can take a better look at the disk. You can see two main images of the disk, one of the upper face, and one, inside, of the lower.

Wait, am I missing something? I don't see any green in the pictures.

Cool, now I have an intuition for how a black hole distorts lines of sight. Basically it's like a black ball hanging in space, with a "halo" that shows you a distorted image of everything that's behind the ball. For example, if there's a hula hoop around the ball, the near side of the hoop will look more or less normal to you, but the far side will look like it's warped, like it's refusing to go around the ball and instead shows up above or below it (or both). And the innermost part of the halo shows infinitely many images of the hoop (and everything else around the ball), stretched thinner and thinner.

I love this site, the method is very clever. I did a re-implementation of this at one point for fun. One of my favorite things to do with this simulation is to actually plot the path of the photons as they approach the black hole (well, leave the black hole but you reverse time for ray tracing purposes). Conveys the idea pretty well that there's something funky going on with spacetime. E.g. here https://imgur.com/OPHLMJL (just a screen shot from back then, it's much more fun with sliders that led you move things around).

I was rather confused by the section describing a "white and blue ... white and green" image that doesn't actually contain any green.

Eventually...

> The lower surface is blue and not green because I'm lazy

Being lazy is fine (I've rendered zero black holes myself), but he could at least be consistent.

This is pretty awesome to see since I did something related to this to simulate gravitational lensing around black holes back in 2000 under the guidance of Ronen Plesser at Duke University. Back then however, the computing resources I had available were far more primitive and I can't even imagine how long it would have taken me to render something like this back then. That said, this is way more advanced than what I was doing as well.

I had never thought about it but in a sense black holes work as a mirror. Some light will sling shot around the back of the hole and back in the opposite direction.

This is called gravitational lensing https://en.wikipedia.org/wiki/Gravitational_lens

And some light will orbit for eons before either going in or escaping.

Eons from whose point of view?

Not from the light. It’s always instantaneous

Constant, not instant.

You always get some back scatter from a glass surface, so any magnifying glass is much more like a mirror than a black hole.

> Also stereo pics are still cool

Not so much when they're in the stare-into-the-distance configuration rather than cross-your-eyes, and the images are uncomfortably far apart for your inter-ocular distance.

FTFY: https://i.imgur.com/jBhWrUY.png

A fascinating topic. Another fascinating surface is a projective plane, that I've never managed to grasp. Any suggestions on how to visualize a travel on the surface of the projective plane?

Greg Egan wrote a Java app to simulate the view near to and inside the event horizon for his short story The Planck Dive. The view from inside, looking back, at around 2M is particularly interesting

http://www.gregegan.net/PLANCK/Tour/TourApplet.html

http://www.gregegan.net/PLANCK/Planck.html

I did my own some time ago, and it can do Kerr black holes, too ;) No accretion disks, though, but should be possible to implement.

https://github.com/fizyk20/gr-raytracer

This software (and the site itself) is good fun, but you have to build it yourself and feed it good parameters. Unfortunately it is not actively developed (I've been in touch with the author in the past).

Reply on news.ycombinator.com