Incorporating analysis into design is critical to achieving high-performance buildings. Our Daylighting Visualization tool employs Radiance and Daysim to help architectural designers assess daylighting performance as they design, making it easier for them to meet their projects’ performance goals.
We recently reviewed and updated a critical aspect of our default setup, the parameters used when employing Radiance’s rtrace program. We sought a set of parameters that would produce results comparable to relevant and accepted benchmarks without significantly lengthening simulation time. The following outlines our research approach and presents the results of our studies.
We considered three buildings of various complexity: a simple box, a series of stacked boxes, and a more detailed multi-level office building.
For each of the buildings, we conducted five sensor-based Radiance simulations using different sets of Radiance parameters, and estimated the illuminance on March 21 at 10am. (A sensor-based simulation assess illuminance at discrete points at the workplane. The simulations delivered in Sefaira Daylighting Visualization are sensor-based.)
The following five Radiance parameter sets were used to estimate illuminance for each of the three buildings:
- The proposed Sefaira parameters (which have since been incorporated into the product).
- Parameters sourced from Radiance’s default “Medium” rendering setting.
- Parameters sourced from a publicly-available Daysim 3.0 tutorial.
- Parameters sourced from specialists’ input.
- Parameters sourced from early Radiance documentation.
All other elements of the simulation (e.g. material properties, location, etc.) were kept constant. Read on to see the results from each of the simulations
Note that the point illuminance values estimated in the simple box model are very consistent across all five simulations:
This graph shows sensor point values for the whole model. (That’s possible with a small model, as there is a reasonably low number of sensors to fit on the graph). Zooming in to the orange boxed region, we can see that some variation does occur:
The variation is slight, with the greatest outlier being the parameters sourced from early Radiance documentation. The proposed Sefaira settings track well against the other three benchmarks.
In addition to illuminance, we compared simulation run times. Sefaira was a clear winner, as it was much faster than the three benchmarks it tracked so closely against:
The stacked boxes model contains quite a few more sensors, so we took a segment of the results to illustrate how the various groups of settings compared with each other.
Again, the “Radsite ‘Accurate’” settings produce outlying values, while the other four track fairly well against each other. There is a bit more “noise” or variation produced from the Sefaira settings; this is an opportunity for further refinement, possibly associated with scene size and scene description (i.e. we will look beyond just Radiance parameters to smooth this variation).
Timing data was again promising, as Sefaira outperformed the other options:
As the models grew in size and complexity, so too did the variation among the parameter groups:
Input from industry specialists at the 2017 International Radiance Workshop suggested we should consider managing models such that smaller Radiance scenes are created.
Again, Sefaira was the fastest:
Our main takeaway from this third and final study was that the Sefaira settings adequately balance speed and precision: they are tracking closely enough to benchmarks to provide the level of precision necessary for early stage analysis, and they support simulation times that are necessary for iterative design.
If you have any questions, or have comments about the default setup for Sefaira Daylighting Visualization, please contact firstname.lastname@example.org.