SSS - What value the HG phase function field uses?

[Meelis]

Hi

What value the HG phase function field uses?

If uniform, is it average of all used measured wavelengths, or measured @ some fixed wavelength?

How to i get R G B values from that?
In Indigo is R 635-700nm, G 490-560nm, B 450-490nm?

[CTZn]

Hi Meelis

What value the HG phase function field uses?

The phase function values must be in the range [-1;1].

If uniform, is it average of all used measured wavelengths, or measured @ some fixed wavelength?

Good question... I assume uniform scattering is for all wavelengths and in all directions. Guesswork.

In Indigo is R 635-700nm, G 490-560nm, B 450-490nm?

No, it's the other way around (swap R <-> B). Maybe absorbtion is a factor of confusion, as opposed to transparency ?

I don't know a source of physical references for this parameter.

It defines the direction of scattering, negative values produce backwards light scattering. You can achieve polychroic glasses with that parameter, though they are cpu-intensive materials.

[Meelis]

I was just wondering whats good value for:
<phase_function>
<henyey_greenstein>
<g_spectrum>
<uniform>
<value>0.97503</value>
</uniform>
</g_spectrum>
</henyey_greenstein>
</phase_function>
Because this value is quite different per given wavelength.
1 source stated measured values G 0.94 @ 500nm and G 0.85 @ 750nm). And the values are lower @ less absorbing wavelength's. Another source told 0.972 is an adequate approximation for the effective scattering phase function @ wavelength of 514.
But is it adequate Indigo uniform value?

[CTZn]

Oh, that uniform ? No it's not suitable for what you want to achieve.

There is a possible confusion going on here:

Uniform phase function per se takes no value, unlike that uniform spectrum you exposed. Talking about Hervey-Greenstein phase function, uniform is not the spectrum type you want to use. If you have a set of values at different wavelengths then you want to use tabulated data:

Code: Select all

<phase_function>
	<regular_tabulated>
		<start_wavelength>0.4E-06</start_wavelength>
		<end_wavelength>0.7E-06</end_wavelength>
		<num_values>2</num_values>
		<values>
			0.94 0.85
		</values>
	</regular_tabulated>
</phase_function>

This igs code will split the visible spectrum in 2 parts, and assign a specified value for each.

The number of values only depends on the quality of your ressource, the more the smoother.

G 0.94 @ 500nm and G 0.85 @ 750nm). And the values are lower @ less absorbing wavelength's. Another source told 0.972 [...]

All these values are comprised between 1 and -1, they are valid phase function values. The example code above uses the two first values you gave.

If you have only two values but need an higher accuracy regarding the splitting wavelength, you can duplicate the values (and augment num_values) in order to obtain smaller fractions of the exposed specctrum:

Code: Select all

		<num_values>10</num_values>
		<values>
			0.94 0.94 0.94 0.94 0.94 0.94 0.85 0.85 0.85 0.85
		</values>

The exposed spectrum bandwidth is defined ranging from 400nm to 700nm (range:300nm. you can change that), and we have 10 values: each slice will be 30nm wide.

Note that this example above is strictly similar to the following peak spectrum:

Code: Select all

<peak>
	<peak_min>400</peak_min>
	<peak_width>180</peak_width>
	<base_value>0.85</base_value>
	<peak_value>0.94</peak_value>
</peak>

When I talk about fun with phase function, it's by playing with these two spectrum types to define it

Hope that helps !

[Meelis]

Thanks, that cleared things a lot.


------
EDIT

And don't forget to add the <henyey_greenstein> <g_spectrum> lines

Code: Select all

				<phase_function>
					<henyey_greenstein>
						<g_spectrum>
							<regular_tabulated>
								<start_wavelength>0.35E-06</start_wavelength>
								<end_wavelength>0.75E-06</end_wavelength>
								<num_values>3</num_values>
								<values>
									x
									x
									x
								</values>
							</regular_tabulated>
						</g_spectrum>
					</henyey_greenstein>
				</phase_function>

[CoolColJ]

didn't know you could use spectrum tablulated values on the phase function

I have a feeling this might be the reason why my Dirty Water material when using SSS goes all multi-coloured

[neo0.]

The Indigo site really needs a glossary for those of us who are well.. less experienced in physics.. Things like uniform scattering, tabulated spectra, etc.

[Borgleader]

The Indigo site really needs a glossary for those of us who are well.. less experienced in physics.. Things like uniform scattering, tabulated spectra, etc.

You could turn this into a learning experience and start building it yourself! Ask not what... Indigo can do you for you, but what you can do for Indigo?

[neo0.]

The Indigo site really needs a glossary for those of us who are well.. less experienced in physics.. Things like uniform scattering, tabulated spectra, etc.

You could turn this into a learning experience and start building it yourself! Ask not what... Indigo can do you for you, but what you can do for Indigo?

Donate servers so they can increase the maximum size for attatchments?

[Meelis]

didn't know you could use spectrum tablulated values on the phase function

I have a feeling this might be the reason why my Dirty Water material when using SSS goes all multi-coloured

If SSS value is too hi for a large volume then it can get colourfull noise that maybe gets clean after years,
especially with dispersion on.

[CoolColJ]

didn't know you could use spectrum tablulated values on the phase function

I have a feeling this might be the reason why my Dirty Water material when using SSS goes all multi-coloured

If SSS value is too hi for a large volume then it can get colourfull noise that maybe gets clean after years,
especially with dispersion on.

I found dropping number of bounces to 1000 and under helped a lot. Otherwise the image is clean by 9000 samples, but caustics and SSS still not converged with 10000 bounces!!
And with lower bounce count the indirectly seen caustics show up pretty quickly and are mostly shown by 2000 samples. After that it's just a matter of waiting till 4000+ samples to get them smoother and brighter etc

I've tried a lot of unbiased renders, and many will not show real caustics at the bottom of the pool, and even if they do it takes forever to show. To be fair Indigo used to be like that too