On 15-3-2019 22:39, Alain wrote:
> Le 19-03-15 à 07:54, Thomas de Groot a écrit :
>> On 15-3-2019 9:35, Thomas de Groot wrote:
>>> On 14-3-2019 13:28, Thomas de Groot wrote:
>>>> On 14-3-2019 8:43, Thomas de Groot wrote:
>>>>> The small issue I was talking about is the thin line of light 
>>>>> visible between the glass/lead and the window frame. It could be 
>>>>> the window being a tiny bit too small but I guess it is light 
>>>>> "leaking" at angles. Something similar is shown at the angle 
>>>>> between floor and wall. I dimly remember this phenomenon was 
>>>>> discussed many years ago and I do not remember the way to correct 
>>>>> that, except, I believe, by using stochastic anti-aliasing. I shall 
>>>>> test that in any case.
>>>>>
>>>>
>>>> As I supposed indeed, stochastic antialiasing (with no_cache) solves 
>>>> the issue.
>>>>
>>>
>>> The downside of this is that combining stochastic aa with photons, 
>>> will dramatically decrease render speed. A rough calculation for the 
>>> image above, with a crude setting of aa (+a0.1) and with photons 
>>> enabled, shows that it will take at least 24 hours to render. :-(
>>>
>>
>> Not precise enough I am afraid. I need to give you more relevant info:
>>
>> + The issue mentioned above was solved using UberPOV, without photons, 
>> and with no_cache on in the radiosity block. Command line: +w640 +h640 
>> +am3 +a0.01 +ac0.995 +r3 +bm2 +bs8 +wt6. Radiosity's count was lowered 
>> to 10. Render time seemed acceptably "low" (I didn't render a complete 
>> image).
>>
>> + The same settings using UberPOV, this time with photons on, the 
>> render time soared up to the estimate of 24 hours+.
>>
>> + Using the latest version of 3.8, and with identical settings for 
>> +am3 on the command line (no_cache is /not/ available in this 
>> version), both with or without photons available, the render time was 
>> "low" i.e. acceptable for a test and within the hour. However, the 
>> issue mentioned above was *not* solved. I tried to increase 
>> dramatically the count value in the radiosity block but to no avail 
>> where the issue was concerned.
>>
>> I would like some expert's view on the matter please.
>>
> 

Thank you Alain, much appreciated. some answers:

> The most obvious issue that I see is where the wall meet the floor. It 
> looks like you are getting radiosity samples from the back of the wall.
> Some possibility :
> Make the wall go slightly lower than the floor.

That is already the case.

> Place another box inside the wall at the floor's level.
> Add a plinth in front of the wall. It can make the scene more realistic 
> at the same time.

That is certainly something to do.

> Scale up the whole scene by some factor, like 10 times. If using spacing 
> for the photons, increase it by the same factor.

That is an interesting idea to follow up upon. I seem dimly to remember 
now that this was the best solution given in the past. This time, I am 
going to write that down in my notes book. ;-) [and then forget about it 
of course]

> 
> If you have some problem with the window, try making the outer edge 
> broader so that it goes into the red frame.

That also is already the case. I shall see what happens when scaling the 
scene up.

Thanks again!

-- 
Thomas

Post a reply to this message