Ambient Bounce parameter recommendations/studies


Have a look at sections 3.3.1 and 3.3.2 of John Mardaljevic's thesis.

Also, John is presenting a Seminar on the Radiance Ambient calculation at the Radiance workshop in August. I'd suggest you consider attending since it is only across the bay from your office (workshop info:



On Jul 7, 2011, at 9:45 AM, Chakraborty, Judhajit wrote:


Can anyone recommend any literature/study regarding setting up optimum ambient bounce (AB) parameters for spaces based on the accuracy of the results compared to the degree of its complexity/ floor plate sizes/ time taken to finish a rendering etc?
We are doing a study comparing a single space with the following daylighting strategies using both Radiance and Daysim:
a) Simple fenestration with only low e glazing (single/double/ triple insulated glazing etc)
b) Fenestration with blinds (manual/automated)
c) Light shelves
d) Combination of all of the above.

Any help/recommendations in this regard shall be highly appreciated.


Judhajit Chakraborty
Built Ecology Specialist
405 Howard Street
Suite 500
San Francisco CA 94105
T. +1 415 398 3833
D. +1 415 402 5815
F. +1 415 433 5311
E. [email protected]

A specialist service of WSP Flack + Kurtz

Radiance-general mailing list
[email protected]

Hi Judhajit!

In addition to previous replies, I would recommend these two:

1) study how rad is setting the ambient settings based on the information given in the rif-file - it is usually quite successful doing so. Code is open source...

2) Mark Stock did a lot of experimentation, especially focusing on scenes with large amounts of geometry objects which tend to stretch the ambient cache's efficiency. He was especially interested in rendering times and the visual appearance, still a great documentation by Mark at

Besides that - it is good practice to do some work on finding out when you settings lead to "stable" results. So for each scene type, it may be wise to vary ambient settings from "high performance" to "high accuracy" for different ambient bounces (-ab), at least once for sunny and once for cloudy skies. You can do so e.g. by rendering illuminance on the floor plane one pixel wide perpendicular to the facade. Plot the results, and you will find

a) at which number of ambient bounces increasing -ab does not lead to significant increase in recorded illuminance

b) at which "accuracy" you get a reasonable (usually rather smooth) plot - very low accuracy tends to give a lot of spikes in such a plot, even if the average may be acceptable

Having these tests done will probably give you a much better feeling about the parameters in your rpict/rtrace command line and should be kept with the results :slight_smile:

Cheers, Lars.