Camera path mode and scene scale problem

[RE:FORM_STUDIOS]

I have my scene setup in meters, with 1 system unit = 1 m.
Using IR + LC with camera path mode on, the result is bad, the GI solution looks like the caustics reflected off a swimming pool in sunlight!

Switching my units to Generic Units... the GI solution is perfect...

Why would this setting change make any difference? Especially seeing as I'm not actually changing the system unit scaling. Is this a bug?

p.

[RE:FORM_STUDIOS]

Actually hold on... I the solution wasn't perfect... so it might not be related to the scene scale...
once I work out the culprit I'll report back on this thread...

[RE:FORM_STUDIOS]

Do I need to increase LC samples depending on animation length? ie 1000 samples for 25 frames, 4000 samples for 100 ?

[vlado]

You need to increase them based on the camera path length, not so much because of the animation range itself.

Best regards,
Vlado

[RE:FORM_STUDIOS]

Thanks vlado, I was going crazy for a moment there trying to work out why an animation which was working great for a 25 frame sequence was looking so bad when rendered at 100 frames (same rate of movement for the camera, ie 4 times the distance.)

Is there a method for determining the best sample rate, something akin to Lele's LC calculator, that I can use to best determine the number of samples required?
If there's not mathematical approach, is there any "rules of thumb" I can employ?

Should LC mode be set to Screen or World when using camera path mode?

[RE:FORM_STUDIOS]

After coming across some more info on another thread, I'm beginning to think the problem may lie more in my IR settings rather than LC. Increasing my IR to -3/1 improves the solution a bit, but not enough.

for a fourfold increase in camera distance, would I need to increase my IR map from -4/0 to say 0/4 ? I assume the increase required depends on the interp frame setting? So for an interp frame rate of 3 I'd need 7 times the amount of IR samples to achieve a similar result?

Sorry for the scattergun approach to my questions, I'm just trying to understand more about how the different sample rates influence the final result when using camera path.

Thanks again
p.

[vlado]

Each increase in the min/max rate increases the number of samples by 4, so you would only need to go from -4/0 to -3/1.

Best regards,
Vlado

[3DMK]

Should LC mode be set to Screen or World when using camera path mode?

I always set the mode to world as mentioned in this Vray tutorial

http://www.spot3d.com/vray/help/150S...ials_imap2.htm

You can notice that in places where the camera gets close to a wall in the scene, the light cache samples are smaller and more noisy. This is because the light cache Scale is set to Screen by default. This is fine for still images, but for animation we would like a more even sample distribution. To achieve this, we will need to use the World option.

Cheers

[RE:FORM_STUDIOS]

One of the problems seems to be the samples swimming around... does using world units ensure static sample placement?

[RE:FORM_STUDIOS]

When rendering the scene with a white material over-ride, (prepass with white override too), the scene looks great. Going back to normal materials seems to screw up the prepass . The GI solution starts swimming and flickering again. My IRMaps go from 16mb per frame with a white material, to just 3Mb per frame with the correct materials. I assume this is through rays terminating earlier due to lower darker materials. If so how best to tackle this problem?

Clr Threshold? I'm using 0.3/0.1/0.1 for my thresholds (clr/nrm/dst),

or higher DMC settings?... currently adaptive_DMC, 1/8, noise-thresh .01

I guess it must be one of these thresholds becase with a white-card material, there are enough samples to get a good result.

I thought I could test the whitecard prepass with the normal materials, but the rendertime goes through the roof. Switching off reflection/refraction seems to fix this slowdown. Why would this happen (the increase in rendertime far exceeds what I would expect from switching on reflection/refractions). I assume the GI animation mode doesn't like having different materials to deal with than were used in the prepass?

[3DMK]

One of the problems seems to be the samples swimming around... does using world units ensure static sample placement?

yes, 'world' units seems to lock all the samples down as I always use 'world' scale & have NEVER had any sample swimming problems in my animations.