dear all
hi,i am a building physics university student,doing my thesis on
daylighting. i have a problem using radiance in my analysis, i draw my
simulation zone on ecotect and then export it to radiance,the problem is,
each time i do the analysis the result is different. do you know what can i
do?
thanks
sincerely
Shailan Zareiy

Hi Shailan,
Radiance uses probabilistic sampling of the model, which means that sample
rays are different for each run. You will never be able to get exactly the
same result every time, but by tightening the simulation parameters you can
reduce the difference between runs considerably. Better simulation
parameters means longer simulation times - it's a typical tradeoff between
precision and duration (you can have noisy results quickly or precise
results tomorrow).

I suggest looking at the descriptions of simulation parameters in this
tutorial:
and those starting on page 10 of this tutorial:

I'm a big fan of convergence testing of simulation parameters to build
confidence in results. Without convergence testing I have no way of
knowing what parameters are good for a model. John Mardaljevic did some
convergence testing in his thesis (see figure 3-26 & 3-30):
http://www.iesd.dmu.ac.uk/~jm/lib/exe/fetch.php?media=resources:mardaljevic_chap3.pdf

It also helps to look at your model interactively in Radiance with rvu to
make sure that everything is how you expect. I think there is a version of
rvu that works on windows in the NREL radiance package:

Best,
Andy

···

On Tue, Oct 2, 2012 at 10:15 AM, shailan zareiy <[email protected]> wrote:

dear all
hi,i am a building physics university student,doing my thesis on
daylighting. i have a problem using radiance in my analysis, i draw my
simulation zone on ecotect and then export it to radiance,the problem is,
each time i do the analysis the result is different. do you know what can i
do?
thanks
sincerely
Shailan Zareiy

_______________________________________________
[email protected]

Thank you very much andy. I think your answer solved the problem.

···

On Tue, Oct 2, 2012 at 9:14 PM, Andrew McNeil <[email protected]> wrote:

Hi Shailan,
Radiance uses probabilistic sampling of the model, which means that sample
rays are different for each run. You will never be able to get exactly the
same result every time, but by tightening the simulation parameters you can
reduce the difference between runs considerably. Better simulation
parameters means longer simulation times - it's a typical tradeoff between
precision and duration (you can have noisy results quickly or precise
results tomorrow).

I suggest looking at the descriptions of simulation parameters in this
tutorial:

and those starting on page 10 of this tutorial:

I'm a big fan of convergence testing of simulation parameters to build
confidence in results. Without convergence testing I have no way of
knowing what parameters are good for a model. John Mardaljevic did some
convergence testing in his thesis (see figure 3-26 & 3-30):

http://www.iesd.dmu.ac.uk/~jm/lib/exe/fetch.php?media=resources:mardaljevic_chap3.pdf

It also helps to look at your model interactively in Radiance with rvu to
make sure that everything is how you expect. I think there is a version of
rvu that works on windows in the NREL radiance package:

Best,
Andy

On Tue, Oct 2, 2012 at 10:15 AM, shailan zareiy <[email protected]>wrote:

dear all
hi,i am a building physics university student,doing my thesis on
daylighting. i have a problem using radiance in my analysis, i draw my
simulation zone on ecotect and then export it to radiance,the problem is,
each time i do the analysis the result is different. do you know what can i
do?
thanks
sincerely
Shailan Zareiy

_______________________________________________