1. Didja accidentally blow through the whole, "We're using our real names" thing on registration? No problem, just send me (Mike) a Conversation message and I'll get you sorted, by which I mean hammered-into-obedient-line because I'm SO about having a lot of individuality-destroying, oppressive shit all over my forum.
    Dismiss Notice
  2. You're only as good as the harshest criticism you're willing to hear.
    Dismiss Notice

On the Softness of Shadows

Discussion in 'UNIT M3' started by Martin Hoffmann, Apr 26, 2020.

  1. @Mike Verta: I just watched the part of your redpill video about teaching Draco how to draw. I think it's worth taking a look at the concepts presented in the book "Drawing on the right side of the brain" by Betty Edwards. As far as I know the theory of the roles of the brain hemispheres as she presents them has largely been debunked, but the concepts she teaches still work all the same. I think for a beginner artist it's a useful mental model to think of these two modes. The further you progress the further you'll wander away from that kind of thinking automatically, but I found it to be an eye opening and encouraging boost when I started to learn how to draw. Especially the excercise of copying another drawing from reference, then flipping the reference on its head and doing it again. Here are two videos where someone does those excercises from the book and talks a little about the concepts behind it. The second one has the excercises I was looking for. I didn't watch both videos in full, but at a glance they seem to be OK summaries:

    Here you can see a couple before and after examples of people who took a course with the author and in a couple days they make very visible progress:

    A couple things in the book I would consider unnecessary crutches, like building the kinds of viewfinder or tracing tools that she proposes. I just ignored all those. The core concepts are very simple: learn to not "think in labels" of what things are, just focus on "what they really look like", and copy only the visual information in detail. So you can probably get most of what the book has to offer just by reading/watching a summary and doing similar excercises like for example the ones in the second video.

    That kind of drawing skill alone isn't super useful on its own (very impressive to non-artists though), but I think it's a great stepping stone to keep building on, and it's a very easy way to get over the first hurdle on that Journey, and build confidence in your ability to learn drawing or painting.

    I didn't know where to put this post, but I thought here's as good a place as anywhere, feel free to move it, if you want.
  2. Great, I hope it helps!

    By the way, I think when you explained why the shadow of a pole is fuzzy in the middle of the street but sharp at the bottom, you gave the wrong explanation. That effect comes from the size of the sun, which is big enough to make the light rays noticably non-parallel, and the further the shadow is away from the casting object, the further spread out the shadow penumbra is (I had to look that word up).


    I think you mixed it up with the explanation for why in very thick fog farther away objects get blurred.
    I know you'd have gotten them right if you hadn't been blazing so quickly through the examples. I just thought I should - in the spirit of fighting entropy :) - point it out, so that others don't memorize the wrong answer.
  3. #3 Mike Verta, Apr 28, 2020
    Last edited: Apr 28, 2020
    No, at that distance with parallel rays the effect is caused by particulate matter in the atmosphere. At 93 million miles away, the sun's rays do not "see" the difference of 15 feet. The effect you're talking about requires local, smaller light sources. It's one of the reasons that sunlight is so difficult to reproduce. We try with giant 10k lights super high-up on cranes, but it never really works. The effect you're referencing here, however, comes into play every single day when taking photos or lighting sources, where you can change the nature of shadows by either changing the distance from the subject, or the relative size of the emitter. In VFX, we almost always have a special "infinite distance/parallel rays" emitter to simulate the sun, because no local light source of any size will properly reproduce the effect. This is also why, conversely, this effect does not happen on the moon, where you always get evenly sharp shadows.

  4. Interesting, I'm happy to learn something new. Maybe it's a bit of both? It's hard to find pictures shot on the moon that solidly support my claim, because they're either blurry or shot at weird angles, but I think I found one here:


    Look in the foreground and compare the sharp shadow of the white bag with the shadow of the leg of the landing module at the same z-depth in the picture. The closer object casts the sharper shadow. No atmospheric scattering involved.

    You're right too that the "glow" around the sun that we can see in the sky and that is caused by the atmosphere is likely still big and bright enough to affect the look of the shadow. But still, I argue that the bigger part of the shadow penumbra on sunlight is caused by the enormous size of the sun. If the rays were truly parallel, you would see it only as a tiny dot on the horizon, but when the light is filtered by the atmosphere enough during sunset that you can look at the sun and clearly see its edge, you can clearly make it out to be an object in your field of view that has a noticable size to it. It's not just a tiny dot like a star on the night's sky. That also means you can be standing behind an object and have it only partially cover the sun from your point of view. And the further you are away from that object, the farther you can move your eye sideways before "sun fully covered by object" turns into "sun just became fully visible", and I would argue that makes up a considerable part of the width of the penumbra of the shadow that that object is casting onto your face. The umbra of the shadow sees none of the rays of the sun, in the middle of the penumbra half of the rays hit the ground (like seeing a half-covered sun from that point of view) and the penumbra ends at the point where all of the rays from the sun hit the ground (sun no longer covered from that point of view).

    I don't use vray, but I think the equivalent of what I'm talking about is the size parameter, and you're talking about the turbidity and ozone parameters I assume?


    Correct me if I'm wrong, but the main reason I'd see to put up a super bright light on a crane is to increase the distance between the light and the scene you are shooting, because that makes the effect of the inverse square law of light less visually pronounced - relative to the scale of the scene/set - because light falloff isn't linear.

    And since putting something up on a crane is nothing compared to going out into space, you're right, you can never immitate that effect on earth. I would argue in VFX the lack of distance based light falloff of a "sunlight" type light source is just as important as its close to parallel (but not fully parallel) light ray direction.
  5. #5 Mike Verta, Apr 28, 2020
    Last edited: Apr 28, 2020
    Keep questioning, keep thinking, it's always good. V-Ray's sun size approximation is not physically accurate. There are other accurate algorithms out there which directly account for turbidity, ozone, and other atmospheric parameters. As part of the internal testing team for Maxwell render, we did this stuff every day for 10 years, which is why that engine remained the ground truth standard for path tracing renderers until its development team formed the new Bella render. The sun's rays are effectively parallel, not literally parallel. Similarly, the moon, while not having an atmosphere, does have moon dust which is believed to sort of hover just off the surface at times, and of course can be kicked up by astronauts and what not. Sort of to your point about "sun size," there is no atmosphere to change the "size" of the sun on the moon, and yet some shadows are sharp, and sometimes they're a touch soft, even between objects of negligibly differing size. Dust, is certainly part of the why, though to add to the confusion, angle and nature of geometry can play a role, too:

    Here's a sun simulation where only the sun angle has changed. The nature of the geometry and the slightly-less-than-perfectly-parallel nature of the rays contribute to the effect as well, albeit not as pronounced as with additional particulate matter to scatter rays.


    Most often the shadows on the moon look almost artificially sharp, but still that's usually the way we render it. Because of the electrostatically hovering dust, though, sometimes we'll soften them just a little bit to match the occasional reference photo that exhibits that quality like you posted. In most modern rendering engines we can disable the inverse square law or selectively alter virtually any aspect of light propagation. The typical implementation of a physical sun simulation is primarily to replicate the near parallel nature of the rays and the turbidity and ozone in the atmosphere which generates the soft shadow we're talking about. If you keep your eye out for it, this phenomenon is actually easy to see varying degrees of, if you're ever in an area with especially strong pollution, you can see a much more pronounced version of the effect, or if there happens to be a lot of smoke in the air. One day, there was one of the wildfires about a mile from our house, and Draco and I were taking our usual coffee walk in the morning, and he pointed out that the telephone pole on the corner where I first pointed this out to him exhibited this phenomenon, greatly exaggerated. We can make an argument that atmospheric conditions can artificially seem to inflate the size of the emitter, but we still crash into the fact that that emitter, regardless of its perceived size, is 93 million miles away with nearly parallel rays, and the 20-foot height of a pole doesn't represent a significant enough distance for emitter proximity to be an issue. Grab a flashlight and a pepper mill and we can see this effect all day, for sure.

    And, incidentally, the reason we hang 10k lights big up on cranes is because even though they have fresnel lenses on them, the increased distance just helps the overall coverage, and yes minimizes the proximity effect between light source and casting objects. It just helps fake it a little better. Like anything else in nature, light propagation and shadow are insanely complex phenomena, and if we're even noticing these things in the first place, we're way, way ahead of the game.

    P.S. I'm trying to think if there's a "home version" of this effect I can do as a sort of lab-day for Draco. We have a fog machine and a few Fresnel lights. I wonder...
  6. I will, even though I must look like an insufferable smart-ass by now. But you still haven't convinced me yet...

    Yeah, but the sun is also 1.39 million km (or 864,000 miles) in diameter. You can calculate the angle between a ray coming at you straight from the middle of the sun to a ray coming from the edge of the sun with basic trigonometry. Here's a crude sketch and calculation (not to scale obviously because of the extreme dimensions):


    Here's a link to a calculator where you can easily enter the numbers:


    20 foot is about 6 meter. Assuming a 45° degree average angle of sunlight you get a shadow that is also 6 meters long. The distance between tip of the pole and the tip of the shadow is about 8.48 m. Now if you put 8.48 into the calculator for side b, and 0.26° for angle A, you get side a = 0.038 meter. Meaning at the tip of the shadow of the 20 foot pole, at 45° average angle of incoming sunlight, the distance between the center and the edge of the penumbra of the cast shadow is a little under 4 cm. So if I got this all right, the 20 foot pole has a 7.7 cm wide penumbra, just from the size of the sun alone. Feels about right to me from what we're used to seeing. And all other factors of atmospheric scattering that further cause the shadow-casting light rays to not be parallel, additionally add to that effect. Meaning on a cloudless clear sky (or on the moon) a lot of the effect is not caused by atmospheric scattering, and with diffusing objects in front of the sun, like a cloud, smoke, fog, etc. we quickly get into an area where most of the effect is caused by the scattering and the edges of the shadow soon become so blurry it's hard to even make it out as a cast shadow.

    That's what I meant with "Maybe it's a bit of both?". I disagree with the shortened explanation for why the shadows soften further away for the casting object to be "because there's stuff in the air". In my humble opinion the explanation needs to be "Because the shadow-casting lightrays are not parallel." And then you can further elaborate with "They're already not parallel because of the size of the sun, and any light scattering medium they pass through contributes to further redirect some of the rays and increase the degree to which they are not parallel when hitting the shadow-casting object, thus widening the penumbra of the shadow."

    I'll be happy to be proven wrong and learn something new, but I don't see how my math or logic is totally off here. Please enlighten me, pun intended. :)
  7. P.S.: I had replied to the version of your post before you edited it. From the new version I'd think we're less in disagreement than after reading your first version. The two example images you posted would illustrate the exact thing I'm talking about, even without any form of atmospheric scattering in my opinion.

    That we can agree on for sure!
  8. Not looking to convince you of anything, believe me, and if I'm wrong, wouldn't be the first time, that's for sure. We know the limit of science's understanding of the phenomena, because obsessive physicists have implemented every last nuance of these dynamics in physically-accurate sun simulations. On the simplest level, particulate matter scatters light rays. The more matter it passes through, the more it scatters, affecting the prenumbra. It's like any other disruptive medium. Similarly, size, distance, angle from an emitter to a casting object, and nature of the object geometry can also affect the prenumbra. A quick "gotta run" skype with a simulation expert yielded this somewhat cryptic additional info: "Sun rays are diverging and converging, converging .53degrees. Not 100% parallel." Atmospheric scattering is real and observable; ground level scattering is real and observable. We can certainly affect this phenomena deliberately by adding particulate matter to the air. Anyone with a balcony in Los Angeles can tell you it becomes covered in exhaust soot and dirt at an alarming rate; i.e. there is tons of ground-level particulate matter and particulate matter softens shadows. Ain't the only thing. My one-line answer was at best incomplete, and the phenomenon is affected by all these parameters; the only remaining question is by how much. I live half my life in Florida, half in the filthy air of Los Angeles; Florida's shadows are sharper, overall, and even this effect locally (telephone pole shadow) is greater in LA, which has a lot more shit in the air at ground level. So obviously I can neither deny the physics of light propagation, nor the easily-observed impact of local particulate. Perhaps this is why the science of sun simulation isn't "settled"....
    Martin Hoffmann likes this.
  9. Awesome, thanks for checking in with the expert, much appreciated! Then at least my math and reasoning for the direct light from the sun was right. I calculated ~0.266° for center to edge of sun, doubled and rounded you get the total spread of 0.53° that your guy cites. Incidentally Blender also has that as default for the light emission angle parameter on sun lamps. (Not that it means much, since many of its other defaults are totally arbitrary.)
    The air pollution in my area isn't that bad, so it may very well be that in your area the atmospheric scattering from particles in the air makes up the bulk of the visible effect, so we were looking at different examples from the get-go.

    Do you have a projector? You could project an image with a test grid and different color swatches, and then shine the projected light through fog, different liquids, or tinted transparent materials. Or use mirrors to "add" different color swatches to each other, mimicing the classic textbook example for RGB additive light mixing with 3 different colored flashlights. You could let Draco make predictions about how it will affect the projected image on the wall and let him verify his assumptions in the experiment.

Share This Page