Max Brightness Value

[Dade]

@B.Y.O.B.: at what point do you call the get film luminance method in BlendLuxCore ?

Sorry, I don't understand the question.

It is a question for

[Racleborg]

I have just managed to get it to do it again by erratically switching the clamp value on and off before it reaches 1 sample.

[B.Y.O.B.]

@B.Y.O.B.: at what point do you call the get film luminance method in BlendLuxCore ?

After 10 seconds of rendering:
https://github.com/LuxCoreRender/BlendL ... al.py#L143
(I call Film::GetFilmY: https://github.com/LuxCoreRender/BlendL ... er.py#L219)

@B.Y.O.B. is it an uninitialized variable ? Where does it come from as soon after having just opened the file ?

Once the value is computed, it is saved in the property.
So if you only open the file without rendering, you are seeing the last value that was computed.
You have to disable clamping, start a render and wait 10 seconds, then the new value is printed in the console (and assigned to the property).

[Racleborg]

If I have a NAN value whilst using Metropolis, then I have to change to Sobol and re run in order to fix it. Once fixed, if I then switch back to Metropolis, re-run and then Esc within the first couple of minutes then the NAN value is back.

I hope this helps.

[Racleborg]

After playing around with this a lot over the weekend, it appears that the problem only occurs with Metropolis and not with Sobol.

[Fox]

I just found weird thing.
My setup is hdr map light and small heterogeneous volume object (camera is not inside it).
Object has jpg uv texture in transmission color in glass and in Glossy Translucent (i do add 0.02 and multiply by 0.97 to avoid 0 and 1 values).

When i remove the mix material, i get smaller brightness values.
Glossy Translucent only 72.274162
Glass only 1.723827
0.5 mix Glossy Translucent and Glass 340282346638528859811704183484516925440

[Dade]

After playing around with this a lot over the weekend, it appears that the problem only occurs with Metropolis and not with Sobol.

I'm going to check this, I was "absorbed" by a quite serious problem I have found in TILEPATH but I have nearly finished fixing all the related issues.

[Racleborg]

Dade

Thanks, if I can run any further tests for you to help with this issue?

Cheers

[Racleborg]

Just a quick update:

I am finding that Sobol is always working correctly, whereas Metropolis is always creating a NAN value.

[Dade]

If I have a NAN value whilst using Metropolis, then I have to change to Sobol and re run in order to fix it. Once fixed, if I then switch back to Metropolis, re-run and then Esc within the first couple of minutes then the NAN value is back.

I'm following your steps and trying to reproduce the problem here but it works all the times.

@B.Y.O.B.: waiting 10 secs, to update the suggested clamp value, is a procedure prone to errors. For instance, when working with GPUs, the film could have not yet been downloaded from the GPU ram at all. Could you use a condition based on "samples per pixel" statistic ? It is exactly the same problem of BCD denoiser warmup up period (and the same solution could be used here too).