5. Calculation Quality Parameters
Direct light parameters (i.e. not diffuse):
· limit reflection
Number of direct light bounces on surfaces. More light bounces = longer simulation, more light bounces = greater spread of light
· source sub-structuring
If a direct light source (window / luminaire) is big in proportion to the first object it hits then the light source will be divided into more light sources. Calculated at every first bounce for all receiving surfaces. This parameter determines whether the direct source should be sub-divided and is a ratio of size of the source to the distance between the source and the surface. A value of zero means the source is never subdivided, effectively switching this off for maximum speed at the expense of accuracy.
· direct relays
The number of surfaces that the virtual light sources (mirrors & other special types of source in radiance – not currently utilised) will be bounced off.
· direct preset density
As the light bounces this is the number of bounced / scattered rays – depends on material (specular or diffuse). More light rays = longer simulation, more light rays = greater spread of light
· direct thresholding
Used to test for multiple direct light sources hitting the same place and speeds up the calculation by reducing the calculation effort spent on the lower intensity light sources. If pushed to its maximum it means that Radiance takes account of more light sources meaning that shadowing is accounted for in a better way – treating all light sources the same is expensive. Use to get good shadowing
· direct certainty
Second step of the direct thresholding calculation, direct certainty checks to see whether the choice of ignoring the lower intensity sources following an initial calculation was a good one, if not it will then put more computational effort into the previously de-emphasised lower intensity sources. Use to get good shadowing and a double check to catch wrongly chosen values in direct thresholding
· specular threshold
Direct rays come in and bounce off as a new light source if the surface is highly specular and above the threshold value chosen. If the surface is not highly specular or is below the threshold value the direct ray is changed into a diffuse reflection. So a low value will mean highly specular behaviour, a higher value will mean highly diffuse behaviour. Set this value lower than your choice of material specularity where you want that material to behave more like a new light source – i.e. highly reflective.
Ambient light Parameters or indirect light, i.e. diffuse sources of light (surfaces) not light sources
• ambient bounces
The number of surfaces that the light will be bounced off. More light bounces = longer simulation, more light bounces = greater spread of light
• ambient divisions
As the light bounces this is the number of bounced / scattered rays – depends on material (specular or diffuse). More light rays = longer simulation, more light rays = greater spread of light
• ambient super-samples
This increases the amount of attention in the calculation in areas of necessary detail – i.e. light sources or boundaries. Important in ‘detailed’ scenes. This is computationally expensive though and can be offset by increasing ambient divisions instead. Higher number = more rays studied per surface, more rays means more detail picked out of the scene.
• ambient accuracy
This is the maximum error permitted in the calculation between the amount of ambient / diffuse light hitting and scattering diffusely each time the ambient light bounces. It is a way of reducing the iterative computation by setting a point at which the calculation stops. Lower error = more of the light striking a surface is scattered, higher error = less of the light striking a surface is then scattered.
• ambient resolution
Grid size (resolution) used by the ambient calculation, i.e. larger numbers mean more calculation points where light will bounce on a surface. Too few can mean that objects are ‘ignored’. Same value is applied to all surfaces though so there is no means of setting for bigger / smaller surface sizes. Higher resolution leads to more accurate images but at the expense of time.
Other parameters:
· pixel sample rate
Controls the distance (in pixels) that the intensity of light is assessed and then interpolates the value of the light along that line ignoring every pixel in between. Imagine a pixel width venetian blind with a pixel sample rate of 4. There would be two blinds and two gaps across the 4 pixels but the sample would be at pixel 1 and then 5 (4 in between) if pixel 1 and 5 were blinds it would make all of the pixels 1 -> 4 blind, if 1 and 5 were gaps it would make pixels 1->4 gaps. Smaller values pick up detail, larger values miss detail.
· sampling threshold
If the difference as a fraction in light intensity at the pixels at the points described above are greater than the sampling threshold then we will add an extra point halfway through – this is in effect a ‘safety-net’ for a poor choice of pixel sample rate. Smaller values help to ensure that detail is not lost.
· limit weight
This value sets the point at which reflected light has no more power so stops bouncing. Smaller values mean light will reflect more often.
· ambient value (RGB)
Colour starting point for the entire scene for the calculation, 0,0,0 would be no light, 100,100,100 would be white light, 100,0,0 would be red light.
