overcast skies

radzilla
1

Hi Carsten.
I have completed my initial simulations on the daylighting facade, and for sunny skies the simulation look very good. Much better than when done with vanilla radiance (if you want me to send you any images then let me know). I have used a photon port on the outer pane of glass to direct photons into the room and simulation times are not too long. I do inevitably have a couple of questions though:
My simulations work well with sunny skies, but when i try an overcast sky the results look terrible. Very blotchy as if not many photons are making it into the scene. I have tried upping the number of global and caustic photons but this does not help. Is there a problem with photon mapping and diffuse skies. I have looked though all the literature i could find but saw no reference to photon mapping being innappropriate for overcast conditions.
Second question is the speed of radzilla. It compares well with vanilla radiance when neither are compiled with any special compile flags however I can speed up my vanilla installation considerably with compile flags but they have no effect on radzilla. Do you know of any compile flags I could try to speed things up a bit.
Lastly I have tried the -apm mkpmap option described in Roalnds manual to set the maximum number of bounces before a photon is discounted. This gives me an error in radzilla though. Has it yet to be implemented or does it have another name.
And really lastly many thanks for providing this software.

Ryan

ยทยทยท

--
Dr Ryan Southall,
Senior Research Associate,
The Martin Centre,
6 Chaucer Rd,
Cambridge.
CB2 2EB.
Tel: (01223) 331714
Fax: (01223) 331701
Email: [email protected]

2

Ryan Southall wrote:
..

done with vanilla radiance (if you want me to send you any images then let me know). ...

this would be nice, as I'm thinking of putting some images together which especially focus on the addon-features to classic Radiance..

I have used a photon port on the outer pane of glass to direct photons into the room and simulation times are not too long. I do inevitably have a couple of questions though:
My simulations work well with sunny skies, but when i try an overcast sky the results look terrible. Very blotchy as if not many photons are making it into the scene. I have tried upping the number of global and caustic photons but this does not help. Is there a problem with photon mapping and diffuse skies. I have looked though all the literature i could find but saw no reference to photon mapping being innappropriate for overcast conditions.

Unfortunately, the photon map so far has not received much attention, so no discussion and knowledge gathering comparable to that about classic Radiance on radiance-online has taken place. In addition, the pmap-master himself, Roland, apparently has left the scene.

So altogether everyone is still a bit on his own concerning the use and the pitfalls of the pmap implementation.
So far, I've sucessfully used the photon map for localized artificial light sources, I haven't had the time to check things out for wide angle 'source' type lightsources whicgh are used for modeling the sky. I hope I can tell more after having made some tests myself...
Maybe it helps increasing the number of samples per steradian used for calulating the photon emission probabbility for sources with an attached intensity distribution (-apdp parameter, default is 1000)

however I can speed up my vanilla installation considerably with compile flags but they have no effect on radzilla.

This is just a cause of the current crude and provisional installation process. The upcoming release will provide a central place where to put all sorts of compiler flags each user might want to try..

Lastly I have tried the -apm mkpmap option described in Roalnds manual to set the maximum number of bounces before a photon is discounted. This gives me an error in radzilla though. Has it yet to be implemented or does it have another name.

Maybe both :slight_smile:
it is already implemented but I've apparently forgotten to implement the option parsing for it. For mnemonic reasons it might be advisable to rename it, too, as 3 character options so far stand for the different maps (caustic, global etc). The whole lot of pmap options is indeed everything else but user-friendly..

so long

Carsten