Un-dispersion of dispersion

Forums Gallery Un-dispersion of dispersion

This topic contains 30 replies, has 7 voices, and was last updated by Jeremy Hill 1 month ago.

Viewing 15 posts - 1 through 15 (of 31 total)
  • Author
    Posts
  • #3836
    Jeremy Hill
    Jeremy Hill

    We have all seen pictures where a beam of white light (let’s call it a laser) is sent through a prism, producing an album cover:

    Pretty, but hard to see what’s going on, so let’s get a little ambient light.

    The reason for making this scene was to answer a question: we see a dark room with a rainbow on the far wall, but what does the laser see? Luckily we can quite easily test this:

    So let’s get another view, first without ambient light, then with:

    And finally, with no ambient light, so we have in the scene only two lambertian surfaces with grid texture, one beveled prism of SF57 glass, one beam of white light hidden from camera, and one camera, let’s see if we guessed right, what the laser sees:

    Of course! Though from most angles we just see a dark space with a rainbow on the wall, in order to reach the laser’s “eye”, light must have traveled from the laser, been refracted twice on a per-wavelength basis to reach the back wall and produce the rainbow pattern — and then be lucky enough to have followed precisely the reverse of this path to arrive once again at the camera, where it shows the space lit by white light.

    So there you have your dose of bella-fueled optics geekery for the day. 🙂

    #3840
    Avatar
    Tella

    Quite amazing sir!

    #3842
    Avatar
    Thomas An.

    Very creative thinking !
    Basically you are sitting where the laser is and looking through the prism is if it was a window.

    The converse of that idea is have the double prism scene and flip the second prism so that it negates the first (1st spreads and 2nd contracts the rainbow); and a white beam projects to the wall.

    #3843
    Philip
    Philip

    Fascinating!!!!!!

    #3848
    Jeremy Hill
    Jeremy Hill

    Thanks guys! I am going to clean this scene up a bit and put it on the sample scenes page, because it is remarkably fun to play with.

    Reason being, not only is apollo good at finding such paths, but the emitter used is not really as shown in the diagram — I only showed a Thomas An. Collimator™ in the diagram so that you could get an idea of where it is, but in fact it is using an emitterProjector node to restrict the emitter’s aperture, which makes this very efficient; for example, here I have rendered it for 30 seconds on my haswell laptop just to give an idea:

    There are some other interesting situations as well; here is the laser off, ambient light on, and a grid texture placed on the back wall:

    Or laser & ambient off, with a white emitter plane on the back wall, instead:

    At some point I will try the more traditional double-prism recombination, but this way was just too easy (and it did indeed occur to me the way I said, wondering what the laser sees), since by parenting the camera and laser to the same transform, they are guaranteed to stay in perfect alignment, regardless where you move them.

    That said, I did do another type of test in a similar vein, but have not spent the time yet to get it perfectly aligned:

    #3849
    Avatar
    Thomas An.

    More powerful than Maxwell already.

    #3850
    Oscar Cano
    Oscar Cano

    🙂

    #3851
    Jeremy Hill
    Jeremy Hill

    Hadn’t occurred to me before to restrict emitter aperture with the emitterProjector instead of a physical collimator, on the single-wavelength cornell box .. so here is a 1-minute test render of that.

    Obviously such an optimization is not useful in all scenarios, but it sure seems to work pretty nice here. 🙂

    (in case anyone wonders what is up with the lower half of the image, there is a sheet of complex ior glass placed there, so the camera is looking through that to see the lower half of the scene — and if you are completely unfamiliar with what this is, you can find it described in Thomas’ list of psychotic renderer-torture scenes)

    #3852
    Jeremy Hill
    Jeremy Hill

    To wrap that up, here is 30 minutes rendering, 7 threads, cpu is a 4930mx.

    #3853
    Eric
    Eric

    Nice 🙂

    #3854
    Jeremy Hill
    Jeremy Hill

    Heh, I recall you asked about lasers early on .. and these are them. 🙂

    #3855
    Avatar
    Thomas An.

    Holy … 😮 You can do the “Single Wavelength Indirect Cornel” in 30min. It was supposed to be insanely difficult. Now it’s done on a six year old CPU.

    How does it do with the “Internal reflection test” scene using this emitter aperture trick ?

    #3856
    Jeremy Hill
    Jeremy Hill

    I have not tried it, but I’ll try to find the time. 🙂

    #3861
    Avatar
    Thomas An.

    It’s easy for anyone unfamiliar with this scene to dismiss it as just another boring Cornell box.
    Let’s put this in perspective for a second: You are looking at a caustic through glass of a caustic through dispersion.

      -A beam passes through a prism and spreads to various wavelengths (essentially it passes through glass and becomes colored caustics).
      -One of the colored caustic (the green wavelength) passes through a slit and is reflected by a cylindrical mirror. This is a second level caustic.
      -The green second level caustic lights the entire Cornell box and at the bottom half it passes through glass to make it a third level caustic.

    In retrospect I should have made the cube shinny metallic so that it’s caustics seen through the glass pane would be fourth level caustics.

    #3863
    Avatar
    Thomas An.

    I have not tried it, but I’ll try to find the time. 🙂

    On second thought, the technique may not work on that scene, because a perfect laser (running precisely parallel to the back walls of the glass box without touching) would not leave an imprint on the graded cards. On the original collimator the beam fans out by a few degrees, just enough to graze the cards (inside and on top of) the glass box.

Viewing 15 posts - 1 through 15 (of 31 total)

You must be logged in to reply to this topic.