One of the things that customers love about Sefaira’s daylight application is the speed of analysis. Most models, even quite large ones, can be set up to return results within a minute. This makes it possible to embrace a cyclic workflow of test – > adjust – > use feedback – > test again, which is what designers need.
However, one downside of doing quick analysis is that we also get asked, “How can Sefaira’s calculations be legitimate if it’s so much faster than other methods?” So we decided to document and present our processes at this year’s IBPSA USA conference (where all the really serious building physics enthusiasts go).
This article summarizes what we presented in straight-forward terms, in order to explain how we can provide good, accurate calculations at a fraction of the speed expected from daylight analysis.
Radiance inputs are key
Most daylight analysis software, including Sefaira, uses Radiance as the underlying engine (Daysim, which we also use, is an extra layer on top of Radiance). Like all rendering software, Radiance requires the specification of a wide range of interrelated inputs for simulations to run. At Sefaira, we spent a lot of time researching combinations of inputs that provided comparable results to inputs recommended by other respected sources. We then tested those combinations for speed.
The chart below shows the variations in simulation time observed for our 10-story commercial office model for a range of different, well-regarded radiance inputs. Note that there is no standard for what the “right” radiance inputs should be as it is still largely subjective.
Examples of models studied
Simulation times for different rendering parameter sets
As seen here, the same model can take over 100 x longer depending on the Radiance inputs used. There is also no correlation in simulation times between other respected sources of inputs.
We’re able to do multiple things at the same time
If you have a 10-story building, it takes a lot longer to simulate the whole building in Radiance than to just simulate a single floor. However, manually breaking a building up into individual floors, analyzing them separately and then collating the results is generally not that practical. It’s not exactly ten times faster, but almost.
When someone starts a simulation with a multi-story building, we look for different floors in a model, break them up into separate simulations, run them in parallel and then collate the results at the end automatically. Each simulation runs on a different cloud machine. This allows us to make the overall experience much faster than running a whole single model at once, with no change to results.
Sefaira’s method vs. the Whole Building method
Results are as good as other methods
To make sure these processes are generating good results for our customers, we’ve benchmarked different solution methods against each other in terms of results. Results are measured over an array of points. The chart below plots the light levels measured at each point for all the different methods we compared ourselves with.
Light levels measured at each point in comparison with other methods
The level of agreement between results is generally within a range of ± 15%. Sefaira results are typically more conservative than other outputs but are also neither typically at the top nor the bottom of the range.
We’re not happy with 60-second simulations – we want to get our analysis to be as close to instant as possible. We’re also looking at how to do other types of Radiance simulations, like renderings, really fast. Whatever we do, we’ll be making sure our process is open and well documented so everyone continues to have confidence in the validity of the results produced.
To find out more about the information above, download our paper here.
If you’re looking to do daylight analysis on a project, get in touch at firstname.lastname@example.org. Our product team would love to hear about what you’re trying to do, whether or not it’s in Sefaira.