This community is for professionals and enthusiasts of our products and services.
Share and discuss the best content and new marketing ideas, build your professional profile and become a better marketer together.


Raytracing calculation: Expert mode parameters

Kurt Bieri

The justification for having several parameters is that the program aims to offer the possibility of adjustment for all kinds of specific scenarios. 

In general, only a few of the values will need to be changed by the user, as a function of the project. In what follows, a number of guidelines 

are set out to provide you with the necessary background for performing the appropriate parameter adjustment.

Note: the option of changing individual parameter settings is provided as a convience but also constitutes a challenge. 

It is, of course, vital to have consistent value settings. If a high image quality is chosen and at the same time an insufficient spatial resolution 

is set for the indirect light calculation, the results will most probably not be satisfactory.

Expert mode

Scene geometry: Parameters in this category do not generally require user intervention, since they are only provided for very exceptional cases 

in which problems occur with exporting the geometry.

For the raytracing process, the scene data are converted into a special structure (octree).

The scene is subsequently divided into cubes and sub-cubes to provide information on which parts of the overall volume are occupied by objects and which not. 

An important criterion for this conversion process is, of course, the ratio of the smallest featured geometry size to the overall scene size. 

The underlying algorithm is already capable of handling a wide range of geometry ratios, however, so the octree resolution does not normally need to be changed. 

Only in cases in which huge scenes also contain a lot of small geometric detail, especially if this detail appears in a locally concentrated manner, might it be necessary 

to augment this value (e.g. from 16000 to 32000).

A further geometry-specific adjustment option affects the way in which the objects are stored internally. 

Given the fact that scenes are generally growing more complex from year to year, Relux Raytracer allocates all the objects a highly compressed data structure by default. 

Thus, scenes can be rendered now with the new version which were an impossible task for previous Relux Raytracer versions. 

This memory optimization does not normally need to be changed by the user. (For simple scenes, deactivating it can save around 10% of the calculation time)

Image Quality: For convenience, some parameters are still subsumed under the category of image quality even in expert mode. 

Apart from a number of internal calculation parameters, this value also controls the over-sampling rate and the image filtering, 

which is applied to smooth out rough borders and pixel effects in the final image.

Direct Illumination: With regard to the direct part of the overall illumination calculation, the options soft shadows and the associated value 

of the source subdivision in Relux Raytracer amount are currently adjustable by the user. 

For a realistic image appearance, the soft shadows option should remain activated. 

The subdivision in the Relux Raytracer parameter then determines how softly the shadows are to be rendered. 

A higher value for the Relux Raytracer unfortunately also results in an increase in the necessary calculation time.

Indirect Illumination: Calculating the indirect part of the illumination is undoubtedly the most complex task of a visualization. 

Within Radiance, the indirect illumination is also simulated with a ray tracing process, but in contrast to the simple direct ray tracing method, 

rays are sent out not only from the observer's point of view in this case but also from various points within the overall scene. 

These indirect rays sample the (indirect) illumination that reaches the point under consideration from the whole hemisphere above it. 

Benefit is also derived from the fact that the indirect illumination varies much less strongly than the direct illumination. 

It is thus possible, to a certain extent, to cache values from the calculation and to re-use them for points in the vicinity (interpolation method).

This general approach demands a set of specific parameters, of which the four most important ones can be set in Relux Raytracer:

The number of interreflexions, i.e. up to which depth indirect interreflexions are followed, the spatial resolution used for the calculation 

and interpolation process, the number of indirect rays which are sent out to sample the surrounding hemisphere, and, finally, a value for 

the background ambient illumination, which serves to simulate the part of the illumination which is not considered due to the fact that 

only a finite number of indirect interreflexions are traced.

Theoretically, an infinite number of interreflexions need to be considered in order to simulate the light flux exchange between the scene 

surfaces in a physically exact manner. 

To limit calculation times, especially for visualizations, one can work with just a few interreflexions and simulate the rest with an ambient background light. 

In this case, 2-3 interreflexions are recommended for practical application in normal scenes. Let us take a simple indirect illumination scenario as an example. 

With 1 interreflexion, the path luminaire -> ceiling -> workplane is sampled

With 2 interreflexions additional paths like luminaire -> ceiling -> walls -> workplane can be considered. 

The ambient illumination is generally scene-dependent; it is advisable to run a number of previews to establish suitable values for typical scenarios. 

Higher values for the number of interreflexions (3-5) are recommended for scenes with mainly indirect illumination, or if a high calculation accuracy is needed. 

More than 7-9 interreflexions are generally not advisable, since the sampled contributions decrease quickly with higher iteration levels and, 

in the end, fall below the general error of the applied stochastic ray tracing method. (It is clear that the ambient light should be reduced accordingly when using 

such a high number of interreflexions, or set completely at zero, in order to avoid erroneously adding light to the scene.)

Of course, the number of indirect interreflexions has a pronounced influence on the calculation time. 

For quick preview runs, the indirect calculation can be switched off completely (0 interreflexions). 

Keep in mind that all surfaces which are not illuminated directly by a light source are then rendered unnaturally dark.

The necessary spatial resolution is a critical parameter which is very much a function of the scene geometry and leads to artefacts in the image if set to inadequate values. 

The default value of 0.4 m is already a relatively coarse setting to keep calculation times to within reasonable limits. 

Scenes with a high illumination variance and/or fine detail might need a finer setting (such as 0.2 or, in extreme cases, as low as 0.1 or 0.05). 

It should be noted that this, unfortunately, rapidly leads to a considerably higher calculation time. 

On the other hand, with artificially illuminated exterior areas, for instance, where the indirect component part plays a more minor role, 

the value can often be increased (to between 0.6 and 1.0 m, for example) without immediately causing image artefacts.

The pronounced dependence on the scene geometry and the illumination configuration make it difficult to provide any further general advice. 

Given the major influence on calculation time, however, it is generally advisable to experiment with different settings.

The number of indirect rays is set to a default value which is adequate for standard scene types (please consult the following chart). 

What has been said above, however, similarly applies here – a large amount of scene detail and a high illumination variance could necessitate 

an adjustment (to between 800 and 1000, or more) in order to make sure that all the surrounding details are correctly sampled. 

With more or less empty scenes and fairly uniform illumination, the value can safely be reduced (to between 300 and 500, or less). 

The influence on the calculation time is less pronounced than for the spatial resolution parameter (although 1000 rays might sound a great deal, 

they still only permit a rather coarse sampling of the surrounding objects). 

Experimenting with different settings is recommended, too. Hence, in the case of scenes that require a fine spatial resolution on account of extensive geometry detail, 

but which also have only a low illumination variance, it may make sense to reduce the number of indirect rays so as to keep the calculation times within reasonable limits.