We spent a lot of time validating results from Licaso, especially in contrast to the tried and true calculations from AGi32 and ElumTools. (We also validated the results with DaySim/Radiance, see our blog). For the purposes of this article, AGi32 and ElumTools behave identically.
The best way to compare results is to look at the Illuminance Pseudocolor rendering with a scale applied.
Here is a comparison between AGi32 and Licaso on Dec 21 at 11:30.
The illuminance (power law) scale maximum is set to 30,000 lux.
As you can see, the radiosity solutions are almost identical.
Licaso -- Dec 21 11:30
AGi32 -- Dec 21 11:30
Comparing point-by-point results is a bit tricky since the illuminance meters (calc points) for Licaso work a bit differently than AGI32 and ElumTools.
See Images A & B below.
- AGI32 uses "point-based" illuminance meters which gather all the lumens which are incident at the point (Image A).
- Licaso uses "area-based" illuminance meters which gather all the lumens which are incident on the area (Image B).
Point-based meters are "sampling" the environment. The samples are aggregated to approximate things like the average, uniformity, etc. The advantage of point-based meters is that they "tell the truth". The disadvantage is that they only give information about the point itself -- we have no idea what is happening between the points.
Area-based meters are "harvesting" the environment. The advantage of area-based meters are that a group of them can gather all the lumens which are incident on an entire calculation plane (there is no such thing as "in between" the meters, see Image B). The disadvantage is that they only give information about the area and it is not possible to ascertain exactly what is happening at a given point in space.
Put another way, point-based meters are really good at giving you things like the maximum (so long as the maximum does not fall between points), where the area-based meter is better at calculating total accumulation, which is what we want for annual daylight simulation.
Compare values in the images below that are:
- Entirely outside the sunpatch.
- Entirely inside the sunpatch.
- Partially inside the sunpatch.
Additionally, we can cause the values in AGI32 (Image A) and Licaso (Image B) to converge by increasing the density of the calc points (e.g. use a 0.25 ft spacing instead of 2 ft). We can also make it easier to compare the results by increasing the mesh density of the surfaces in the model, especially the daylight transition surfaces (windows, skylights) and surfaces which receive direct sunlight (floor, etc). Of course this increased resolution can cause increased calculation time.
Image A -- AGI32 "Point" Illuminance Meters
Image B -- Licaso -- "Area" Illuminance Meters (indicators visible)
- Licaso does not use Adaptive Subdivision so if you are running a comparison you should disable Adaptive Subdivision in AGI32 or ElumTools.
- Using an extremely small point spacing (try and stay above 3 inches) for Licaso is not recommended and could result in accuracy issues due to surface mottling (common in radiosity with very small elements).
The boring details:
- Our "9 Rooms" test has windows facing N, S, E, W and the center room with only skylights.
- The site is the University of Colorado at Boulder (~ 40 N, 105 W).
- Using the nearest weather station (Broomfield Jeffco Boulder CO USA 724699)