Blotches

[Joppi]

Need a little bit help again!

I´m working on a really big interior scene. The building have large glass stripes on every side (four sides). My scene is lit by four big rektangular lights. The problem is that the renderings take pretty muck time, wich is ok. But i got "blotches" (leopard-pattern) on the inner roof no matter wich settings i use. How to control that.

I´m very interested in feedback of my general settings as well, since i´m new to Vray

[dalomar_cg]

Re: Blotches

The weird thing about IR is that sometimes a "better" setting may not always lead to a better result. Specifically I'm talking about Min/Max rate. You're Max rate is good, but the Min rate is actually too detailed. This is causing a significant placement of samples on large flat surfaces (ceiling) and the slight difference between each of the samples is causing the splotchiness. Go back and set the Min rate to -4 or even -5 and see if that helps (it might actually speed things up a little too).

[andreas]

Re: Blotches

Hello Joppi,

i would suggest you try way higher numbers for hemispherical subdivisions for an interior scene like this I would try 230 or so...

best regards

andy

[Micha_cg]

Re: Blotches

I disabled multipass for IrrMap allways and take in acount the max rate only. Your interpolation samples are quite high, so you lost details - try 20..30 (render faster). Like Andy wrote, try higher IM subdivs. Also avoid glass in the windows.
Lower max rate like -2 is faster, but you lost little shadow details - you must decide, what is needed.

General: if you use a small range for the image sampler like 1/6, than you get a material based subdiv control. So, if a material looks noisy, than higher mat subdivs works. Your current 1/100 is an universal setup and you can't control the blury noise of single objects, only global by the noise threshold.

[dalomar_cg]

Re: Blotches

I disabled multipass for IrrMap allways and take in acount the max rate only.

Incorrect Micha. Multipass will still take into account both the min and the max rate. The only difference is that instead of calculating a pass for each sampling rate, all of the calculations are done at the same time. Having a IR calculation that only samples at one rate, would leave an even blanket of samples across the whole image...no adaptation.

[dalomar_cg]

Re: Blotches

And here's an example where only the max rate is referenced

[Micha_cg]

Re: Blotches

Now I'm confused. Do you mean, if I disabled multipass, than there is a difference between -4/-1 and -1/-1? How looks you last image with -1/-1?

Some months befor I rendered a scene with -3/-1 multipass and -3/-1 no multipass - the difference wasn't a big surprice, the multipass show some disadvantages too (for example a light leak).
Than I tested multipass -4/-1, -3/-1 and -2/-1 with the result: if the wrong relation is set, than the render times are longer than no multipass (-3/-1) and a high min rate caused blotches, because the extrem undersamples pass cause a wrong, confused sampling. The best relation was -3/-1.
So, my experience, single pass works stable for speed and quality. Multipass dosn't show me great advantages and could cause problems, so I don't used it anymore.

[dalomar_cg]

Re: Blotches

There will be no difference between having multipass enabled or disabled. All that does it changes how v-ray progresses through the calculation, NOT the end result of the calculation. Before you said that having mulipass would only reference the max rate. If that was truely the case, then you would have an even blanket of samples as seen in the 3rd image.

As for setting a good min rate, too low (-9>low) will cause the interpolation of the IR samples to be influenced by samples that may not be relevant to the specific area of the scene. Too high of a min rate (-2or-1>high) will cause unnecessary oversampling that will cause the render times to take longer, very little adaptation over the scene, and splotches on large flat surfaces due to variations between the samples (this can be solved by more hsubds which causes things to take longer).

The best settings for min/max rate are usually between 4 and 5 steps apart. For hiRes work sometimes 6 steps is okay for fine details. Fewer steps has the potential to be inefficient and introduce splotches due to oversampling. More steps will cause artifacts due to density bias.

There should not be any practical difference between the IR maps calculated with or without multipass.

[ALTO]

Re: Blotches

Hello Joppi,

i would suggest you try way higher numbers for hemispherical subdivisions for an interior scene like this I would try 230 or so...

best regards

andy

From my point of view there is absolutely NO reason to use this kind of value, for "hard scene" i use always max 70/80 with perfect results!

Joppi, i'm sure there is something wrong with your lighting setup not with your settings (well try lower values in Image Sampler like 1/10, and maybe use -4-2 for IM), what kind of material are you using for the GI Environment? Where are exactly your rectangular lights?

Try to remove glass and lights, replace your Skylight with a simple color (blue or white) and see what's happen... i'm sure you will find the problem!

[dalomar_cg]

Re: Blotches

From my point of view there is absolutely NO reason to use this kind of value, for "hard scene" i use always max 70/80 with perfect results!

Agreed...once you've got 80-100 hSubds there isn't too much that adding more is going to do. And of course the expense of adding more is time.

[Joppi]

Re: Blotches

Thank you guys!

I´m now testing you´re advises. Very grateful!

Every window opening have glass for reflections,
and the rectangular lights are at the same size of each side at the building.
Background is just white, and a reflection layer. No GI lightning. Standard camera.

The scene also contains about 100-120 furnitures & objects (cars etc) that are imported as blocks. Works perfect, exept some views that have large areas of "blackish" materials, with Displecement pattern (wich i now changed to bump, since I have a rendercrash every second time...)

I´ll post a pic on the scene later on.

[andreas]

Re: Blotches

From my point of view there is absolutely NO reason to use this kind of value, for "hard scene" i use always max 70/80 with perfect results!

Agreed...once you've got 80-100 hSubds there isn't too much that adding more is going to do. And of course the expense of adding more is time.

well... I am not so sure... I had several indoor scenes in the past were I got blotchyness in areas, which were only lit by indirect light for example and I had to pitch the hemispherical subdivisions up to 230 or something to get rid of it...

I will try to post a scene tonight, were this behaviour becomes obvious. Of course if there is another / better solution I am always willing to learn

best regards

Andy

[dalomar_cg]

Re: Blotches

I'm not disagreeing that it will do more (it obviously has to), but the question is whether that is the best approach. I think that having hSubDs that high indicates that some tweeks can be done in other areas which may wind up solving the issue and may be faster than increasing the hSubDs that high. There's always multiple ways to get to a good final result, the question is to whether one approach is more effiecient/faster than another. If you don't care about that, then its just a matter of preference as to what you use.

[andreas]

Re: Blotches

I'm not disagreeing that it will do more (it obviously has to), but the question is whether that is the best approach. I think that having hSubDs that high indicates that some tweeks can be done in other areas which may wind up solving the issue and may be faster than increasing the hSubDs that high. There's always multiple ways to get to a good final result, the question is to whether one approach is more effiecient/faster than another. If you don't care about that, then its just a matter of preference as to what you use.

of course we are talking about the best way to archive a certain level of quality in the least time possible, as you said efficiency. In this case I would narrow it down for the approach using lightcache and irradiance map.

Maybe it would be cool to have several "problematic" but still simple scenes to illustrate certain problems that typically can occur when using vray.
The scenes could than also be used as a kind of benchmark for different approaches, were we can compare quality and render times.

I remember many scenes where I had to use completely different setups to get them to work in an reasonable time. Surprisingly often unclean / blotchy problems are not or not only material or light related, but are due to certain geometric configurations in the model.

best regards

Andy

[dalomar_cg]

Re: Blotches

(I've decided to precede my answer below with the short statement...I think that would be a great idea, and would be very useful. It would be very challenging to set up well and make sure that it becomes a sharing ground for good information. A "render-off" to get the best results and shortest time on a given scene would be very interesting and would be a great way to learn provided users shared the information about how they put it together rather than just posting a visopt. Below I am merely sharing my thoughts about possible pitfalls that this idea might be suceptable too.(sorry for the weird intro))

I think that would be a great idea...but rarely is a given thing the result of one setting, or is there "one fix" for a given issue. To complicate things even further, there are certain "flavors" of working with vray that have each one using different base settings and having different tweeks. A tweek that may work for one "flavor" may have a very adverse affect within the context of another flavor. (the best way I can describe this is two workflows; one that relies on QMC for Primaries and AdaptiveQMC for AA, and another one that relies on IR for Primaries and Adaptive Subdivision for AA. A good fix for the second setup (IR and Adaptive Subdivision) may be to decrease the QMC noise threshold. However applying that same adjustment too the first set up (QMC and AQMC) may cause the rendering to slow down to a hault. Same change, but different effects depending on the rest of the settings...To make matters worse, people probably won't change the setting they were just told (or saw) would fix there issue, so the will go back and try to adjust another setting that may actually be fine or have no affect.)

Essentially, learning and understanding the interaction between different settings is something that happens through testing, testing and more testing. The "success" of an end result is always a combination of every thing...geometry, materials, lighting, and settings. My fear in having a scene that people can look too and say "it worked there, so it should work here" is that it might get taken as an absolute, which is far from it. Personally I'm of the mantra of "Teach a Man to Fish", so I'd prefer people ask a question (and hopefully give it some thought in the process) and get a good answer, then have the visopts from a possibly unrelated scene and wonder why its not working for them.