Proper Spot lights?? Area spots? Attenuation? Collimation?

[aaron_ross]

Hi,

I am new to V-Ray, so maybe I am missing something. But there does not seem to be a way to control Spot lights in V-Ray. There are a number of severe limitations compared to Arnold.

1. V-Ray does not appear to support 3ds Max Photometric Spot distribution for any light Shape type other than Point. I'm only able to get super hard shadows and lighting. Using a V-Ray Disc light is not an acceptable substitute, as there is no mechanism to control hotspot/falloff.

2. No Attenuation for any V-Ray lights. There are only two decay options: inverse square and none. Artists rely on computers to do things that can't be done in the real world.

3. No mechanism for simulating a collimated beam. Real-world spot lights are constructed with reflectors and lenses to focus light rays so that illumination travels farther. The decay rate is still inverse square, but the focusing hardware causes the decay to effectively begin at a virtual light position behind the actual source of illumination. Arnold has a Lens Radius parameter that simulates this phenomenon. Without the ability to simulate a collimated beam, it's impossible to get realistic renders of any light fixture that incorporates a lens or reflector. And since all spot lights have lenses and / or reflectors, V-Ray is completely unable to reproduce the optics of a spot light.

So we can't get soft lighting while also controlling hotspot/falloff. We also can't control the decay rate, either by increasing or decreasing effective decay. Without these features, the greater speed of V-Ray is useless to me. My only choice is to wait longer for an Arnold rendering that actually represents real-world lighting conditions, and/or gives me full control over non-physical decay rates.

Thanks,

Aaron

[vlado]

For the moment you can use a VRayIES light, which will be able to properly cast area shadows for IES profiles. It is not recommended to use the 3ds Max photometric lights as they generally have performance issues.

You can also use the 3ds Max Spot lights with the shadow type set to VRayShadow (with the "area shadows" option enabled). They have greater control over the hotspot/falloff/attenuation.

Will make a note of your other requests and we will look into them.

Best regards,
Vlado

[^Lele^]

Area lights have a directional parameter which seems to be doing exactly what the lens radius does in arnold, when used in conjunction with a disk light radius.
The attached images are a disk light shining over a plane, so what you see is a 2d slice of the volume the light covers.

See image 1 is a light with size-invariant intensity units (lm.): small radius, lower directionality.
Image 2 is higher directionality and a bigger radius (which is what truncating a cone cone does).
The third, proof of the concept, is a laser: small emitter size, very high directionality.

I believe the samples match the light distribution found in the arnold manual quite closely, and here too the emitted energy hasn't changed, but has been distribuited in a different way.

[aaron_ross]

Thanks guys. Lele, I knew that the directionality of an area light would give a collimated beam. The problem with that approach is that it doesn't allow control over hotspot/falloff, and only gives a constant luminance over the surface of the light's area. Also, in my experience with the similar feature of Arnold, the result is more noisy than a Spot light with a non-zero Lens Radius.

Vlado, I knew about IES lights, but of course that raises issues with art direction and dependencies. It's awesome that you've provided controls to vary the photometric file's properties, but users are still stuck with having to find an IES profile that matches the needs of the scene. That's reasonable for arch viz, but not a preferred workflow for other applications such as film and animation. We're always going to be somewhat limited by the IES profile. For example, I would wager that there are no IES profiles available for limelight-type instruments with very long throw and a very sharp edge.

I tried the 3ds Max legacy Spot light technique, but this has numerous issues. First, legacy lights are not physically accurate. Second, there's no way to create a collimated beam, because we can't set a negative value for Decay Start. Third, Area shadows don't affect the light *suffusion*, only the softness of shadows. It's always a point light source with zero size. Last, Area shadows don't work on GPU.

Regarding 3ds Max Photometric lights, I don't have enough experience with V-Ray to have noticed any performance issue, but I'll take your word for it. The Photometric lights are very useful in many situations, so I hope you can fully support all of their features. I would be willing to wait for longer render times if necessary.

Frankly, I was very surprised that there is no V-Ray native Spot light. This seems sorely needed.

Thanks,

Aaron

[^Lele^]

Thanks guys. Lele, I knew that the directionality of an area light would give a collimated beam. The problem with that approach is that it doesn't allow control over hotspot/falloff, and only gives a constant luminance over the surface of the light's area.

You can texture the light as you please, and get falloff for color and intensity across the emitter, something you can't achieve with a spotlight (see the vraySoftbox texture).
The Arnold light with hotspot/falloff controls, and a lens radius, is a spotlight too.
The cone truncation you seem to be looking for can be replicated identically, if with different controls, with a standard max spot, using the near falloff, and a choice of falloff types and light intensity to fill the cone.
The only difference is the truncation is flat in arnold spotlight, and round in the standard max spotlight.

Or are you talking of something else, and i misunderstoiod you?

Also, in my experience with the similar feature of Arnold, the result is more noisy than a Spot light with a non-zero Lens Radius.

That's because the spotlight is a point, not an area light.
No need to worry with us, anyways: our engine is adaptive and will converge the image to the noise threshold you set.

We're always going to be somewhat limited by the IES profile. For example, I would wager that there are no IES profiles available for limelight-type instruments with very long throw and a very sharp edge.

You can make your own.

First, legacy lights are not physically accurate.

They are point lights in Arnold too, though. Hence no better in that respect.

Second, there's no way to create a collimated beam, because we can't set a negative value for Decay Start.

I don't get this, could i see a picture, or render, of it?
And if it exists in Arnold, could i also see a screenshot of the light settings?

Third, Area shadows don't affect the light *suffusion*, only the softness of shadows. It's always a point light source with zero size.

It's exactly the same in Arnold, is it not?
If we are talklng of the light i linked you the help page of, i'm unsure as to what difference there should be with light soffusion, as i see none, myself.
But maybe i am not understanding what it is you're saying.

Last, Area shadows don't work on GPU.

If we're comparing GPU engines, though, things may differ a bit (on both sides, to be sure.), as the various limitations of coding for GPU make it a very different game altogether.
In Arnold, f.e., no light filter/modifier (gobo, barn doors, blocker or crucially decay) works with GPU, to this day.
What does, and what doesn't, along with what may and will work, is often dictaded by architectural choices, so it's a bit of a hit and miss, across the market.

[aaron_ross]

Lele,

The Arnold Spot light can have area or volume. It is not always a point light source. The same goes for the Photometric Spot distribution in any Shape mode other than Point. It's not "just" a spot.

If a light actually has a size, the result is soft shadows and suffusion. Suffusion is a light scattering effect that is different from soft shadows. Suffusion can't be reproduced by a point light source. The effect is softer direct illumination and more scattering in GI.

The 3ds Max legacy Spot light can't reproduce a physically accurate collimated beam with Near Attenuation. Well, technically I guess it's theoretically possible to stumble on the exact settings that would work for a particular scene. But with no point of reference to what's actually accurate, it's extremely improbable to stumble over the correct settings. The effect should be an inverse square decay where the virtual light source is behind the light object. In other words, intensity starts decaying at a position behind the light. The overall illumination is diminished, and the decay is still quadratic, but the relative amount of decay over a given distance is reduced. The result is a more consistent level of illumination over distance. Increasing overall illumination to compensate results in a physically accurate spot light, mimicing the optics of a reflector and/or lens.

Aaron

[^Lele^]

We'll have to agree to disagree on what an Arnold spotlight does.
I see no trace of the stuff you're talking about: the spotlight radius *only* influences the shadowing, *not* the lighting, identically to what happens with max spotlights and v-ray area shadows.
The manual pictures, the text written there, and all my tests seem to indicate that the light flux is entirely unchanged by the spotlight radius.

Radius
If Radius is set to zero, then the cone of light from the spot will emanate from a notional point source. With a non-zero value for radius, then the light source will behave more like a spherical source with the radius specified (although again with the light emitted being constrained to a cone). The larger the radius, the softer the lighting from this light source (zero radius will give hard shadows). See also the Radius parameter in the point light.

The physical accuracy of what you talk about in terms of decay and light position is also entirely lost to me (as of that too there is no visual trace i can see in my tests, nor any mention of it anywhere in the Arnold docs.)

The max spotlight has inverse square falloff too, on top of lienar and none, and both can be used with near and far ranges.
There is no stumbling on settings, as one defines precisely start and end (both with visual cues, and with scene unit values), and those do not touch the light distribution between the near and far ranges: the two are entirely independent.

IES area illumination is a wholly different kettle of fish to a point-source light, albeit with area shadowing capabilities.

I will need visual examples, and ideally files to work with, as i wouldn't know how to come up with a proof of concept to showcase what should be implemented to devs.

EDIT: as proof of what i was saying, here's a spot in a volume with near decay at 0, 25 and 50 scene units.
So, to put the light back to where the original source was, it's enough to move the fixture back on its locak Z axis by the same amount of units.
You'll also notice light distribution (set as inverse decay, starting @1 unit) is unchanged, which allows for fine control.
EDIT 2: I added also the images with repositioned light source.
My conclusion is that the Arnold "lens radius" is nothing more, nothing less, that the combination of these two effects: a near decay plus a back-transform.
That it's automatic, and thereby changes the light cone, is -in my personal preference - an unwelcome side-effect.
I find the method with max spotlights superior for control.

[aaron_ross]

Arnold Lens Radius does not apply Near Attenuation, it just moves the start point for quadratic decay behind the light shape. The Arnold documentation is not terribly great, but I tried to explain it in this video, which I have unlocked so everyone can view it without a LinkedIn Learning account.

https://www.linkedin.com/learning/em...6Syft5s4NHlXdu CScW4-PAPDVH6Ppi7trswyLoaUHrCIcJJkyfFT_PPZmj8ymr_S6Iumsm kaSN6ZZYeTJDKPwJ0_vhHIySM3ltZDi1IkHOPcx_D5JdWvCLI6 Rf5K7lrafbTxoSdOoNAgwfif1pE7vBt0H2YV1jEsMp7c5L1tol DYTnQ_AVXEoJhZW0aosO6iWpk1OoQjfTxIOX1EFyuz_kpWjvWJ dG2NEC6DizobAFrDSaZOIbbCOiiUSIh3zqHKOzbBF9QsNY7e55 TDRkM21f8fjQ2LZNqgiZY7Q2ZdF4_30aTNNr6oY7Kzc3GOX8S8 H3t3LyejK7NlsZjrEtA6bHCpVz17cRZaGjhRr_NeOMTVRQrNKm vlq-bQ-WrNYGK-8eXrFDIycPwTSkRehFNgS7j-6AZgG22Z2C8JhzKUhryuoCy5mYtP7O1LcoQfhreWYRUTsj3mL3 0aH7Nn9dZ3lauR1Jqoms0LGOHxiow_5Iracb-n8m4hCvlYg0vjKwJf4EKctc7ySgGaKcK9WWMowrRlLopCFoXdG b_vV_eGMcsr-_IIa_7wFi7WWu4yOydCa1iYUKP7SAJod0KIRr0oTpKptF01vpW hgjGmYesKx2tqu3c7cmHFPXeg9tTZtH5jxowd2parQql-lPCmvO2SqjKfDlZ6yAELx-9Jpl2yeOx7fcwd3gk03o00IB14nojt1PUKwAqfKTCSGJDtM2W2 R0

The idea behind a separate shape Radius and Lens Radius is to isolate shadows/suffusion from decay. Again, the Arnold Spot always gives inverse square decay, but if the virtual light source is behind the light shape, the relative decay over a given distance is diminished. The Spread parameter of an area light does the same thing, but not as well. For example, if you want to simulate a video projector, the best way is with a Spot light.

Near Attenuation is great to have in a pinch, but it's not the same thing and requires more effort to get the same results.

Regarding suffusion, take a look at this. These are two Arnold Spot lights with identical parameters except for the Radius. The position of the lights relative to their respective teapots is identical, and the camera is precisely midway between them. These lights are set to normalize intensity, so they're both sending out the same amount of energy, but the distribution is different.



I've disabled GI to make this more obvious. Notice that the lighting is different on the two teapots and on the floor. It's not just the shadows! The actual shading is different. The one on the left is softer, flatter, with less contrast. It is a little bit subtle, but if you take spot readings of pixel intensities you will see that there is a measurable effect. The hotspot in the center of the right-hand teapot is brighter than the hotspot in the center of the left-hand teapot.

The light on the left has a Radius of 50 cm. The cone doesn't show that effect -- it looks like a point light source -- but it is actually more suffuse illumination. The light on the right has a Radius of 0.1 cm. Notice that the floor looks different, too. That's because the effect of suffused light from an area is more obvious when the illuminated surface is parallel to the light rays. A higher Radius causes many rays to be fired out from different locations, giving softer light and shadows. This is happening even though the rays are all ultimately constrained by the cone angle. That cone angle is effectively cropping an area light.

So, SIZE MATTERS. Suffusion is a real thing. Spot lights can have area. Area does not just effect shadows, it effects the quality of illumination.

[^Lele^]

Nope, that's only due to the *shadowing* size being excessive for the object scale.
Nothing at all to do with light flux or light source size.
It's a point light, stays a point light (unlike, f.e. the sperical point light. that does become a spherical area with radius > 0, in Arnold.).

Attached, a max spotlight with sharp, and then area shadows of increasing size.

[aaron_ross]

Lele,

Take a look at this image.



I've reduced the Radius parameters of the two Arnold Spot lights to 1.0 cm and 0.1 cm. All other parameters are identical. The shadows are nearly the same; very sharp in both cases. But the incident light on surfaces is different. This is evidence that it's not merely excessive shadow size causing the difference in shading. Increasing the Radius of an Arnold Spot light causes greater suffusion / scattering / spread, even though the distribution is constrained within the Cone Angle.

Again, an Area light is qualitatively different from a point light. And an Arnold Spot always has an area. In fact, it's not possible for an Arnold spot to NOT have area. If you set the Radius to zero, you get no light at all.

Regards,

Aaron

[aaron_ross]

I am able to get acceptable results for light suffusion and shadow casting from a 3ds Max Photometric light using V-Ray shadows,.

The only Shape type that seems to work with Spotlight Distribution in V-Ray is Rectangle. Sadly, the optimal shape of Disc does not seem to be supported.

I get believable results by setting the Rectangle Shape dimensions and the V-Ray Shadow dimensions to similar values.

It does not appear that Photometric lights are supported on the GPU.

Currently there is no way to achieve a proper collimated beam that can be fully art directed. This is due to limitations in both the Photometric light and the V-Ray plugin.

[^Lele^]

To the left, a red spotlight looking at the camera, with radius 25.
To the right, a point light with radius 25.

[kimgar]

How exactly are you guys controlling a no-decay vraylight's far attenuation with a falloff/gradient texture?

I'm at a point now where the standard max spotlights are creating weird mesh artifacts, especially on render nodes and need to use something else.

The vray lights render perfectly as a disc with some distribution to simulate a spotlight, but I just cant figure out how to control the far attenuation

It has been mentioned a Falloff map in distance mode as light texture, but I just can't figure it out. Distance blend along local Z axis? Local Z of vraylight would be negative?

I tried a rotated gradient ramp and almost got somewhere, see attached image. But it just won't cut off the light where I want. I think I got my local axis confused, but how is this mapped along light Z axis and how far does the gradient go?
I would expect the light along the plane to go from black (nothing) to red, to green to black again (far attenuation).

The colors are just for testing, but how do you map this to control far attenuation?

[^Lele^]

We don't currently support arbitrary falloff (or near and far ranges) on area lights.
It's unlikely anything like the above will work as expected.

Is the standard spotlight issue a known, and unfixable one?
If not, i'd be keen on getting a scene.

[kimgar]

Ah, thanks for clarifying, I got confused by Vlados reply in an old post "I think we'll skip that. What we will probably add,
is the ability to use a mask texture for the lighting, which simply modulates the light intensity depending on the texture.
Then you could use a Falloff map in distance mode, or a VRayDistance texture to do all kinds of fun stuff. This should be much
more flexible than a mere attenuation."

I guess I missed the keyword "probably", ok, the vraylights render so much nicer, cleaner and faster in this scene though,
I'll try to figure something out, I think I'll have a look at the vray ies lights.

About the standard spotlight issue, it's probably not unfixable, just a pain to debug since it renders fine locally and only happens when a sequence of 10 frames or
so is sent to backburner. Weird. Also, I am on max 2019, so the thing might be fixed already, but that doesn't really help me much at the moment.

I sent you a stripped down scene anyway, in case you still want to see it.