wall splotches

[OPEN_RANGE]

here is link to a part of an image that I renderd at 2K.

http://openrang.server269.com./test-...-splotch01.jpg

I'm hoping that someone can suggest what I might adjust to make the wall and ceiling more even in color. they have a slight mottled look that I'd like to eliminate if possable.

used Primary bounce -IR map, Secondary map - QMC, preset high, HSph subdiv- 20, Interp. samples - 20, Image sampler - Adaptive subdiv.

Thanks in advance

[vlado]

You need either more Hemispheric subdivs for the irradiance map, or more interpolation samples, or both.

Best regards,
Vlado

[studioDIM]

On top of what Vlado said, i'd suggest you to try start the min rate of the irmap at something like -6 (yeah, that big a sample) and refine to whatever you have in your image.
Coupled with more samples, it should well clear the image of any splotch (those you have hare high-frequency ones...)

Lele

[OPEN_RANGE]

On top of what Vlado said, i'd suggest you to try start the min rate of the irmap at something like -6 (yeah, that big a sample) and refine to whatever you have in your image.
Coupled with more samples, it should well clear the image of any splotch (those you have hare high-frequency ones...)

Lele

Vlado,

I increased HSph for 20 to 50. I have read not to go higher than this.
Increased Inter samples from 20 - 30

This seems to help.

Lele,

not sure what you mean by "refine to whatever you have in your image"

Thanks for your help!

[Da_elf]

i usually go with 60/30 for HSph/intermp and 0.3/0.3/0.3 for the rest

[RErender]

Who said no more than 50 for hsph ? I usually set it from 80-100

[percydaman]

120 and 60 here when I have a real bad case of the splotches. I use a dirtmap for finer GI detailing anyways.

[OPEN_RANGE]

Thank you all for your advice on this.

Percy,

Can you explain where you put the dirt map. Is it used as a bump map?

[percydaman]

no its something you would have to render a completely different pass for. Stick it in the diffuse slot of a standard material and then set the mat to 100% self illum. You can then use this as an override in your scene. No lights or GI needed to render this pass. Id do some forum searches if your still confused.

[studioDIM]

Lele,

not sure what you mean by "refine to whatever you have in your image"

I meant to leave max rate to whatever you had: -2, -1, 0 anything

Lele

[Clifton Santiago]

You could also try increasing your Global Subdivs multiplier which would effectively increase your Hsph. subdivs. I usually set mine at 8 and keep my Hsph. subdivs around 16; which means I have a multiplied Hsph. subdivs of 128. In 1.47 this was always high enough (to get rid of splotches as described), but I've found that in 1.5 I sometimes have to double that for interiors, e.g. raise Hsph. subdivs to 32.

[studioDIM]

So to show you a simple example:
The first image is the "High" preset, default hsubdivs and interp samples.
Took 18 seconds to render the GI, and do observe how the sampling is almost evenly done (bar in the most difficult places where it's visible a somewhat denser sampling).
The second is exactly the same settings as the first one, only i set min rate to -6. Eleven seconds (that's what, almost half the time?), and sure as hell you won't get any splotch on even surfaces. Notice how this "helps" vray do a better job at sampling, leaving it more free to choose, in each subsequent pass.
Sampling (whatever your subdivs will be) will be put a lot more in the problematic areas, and the simple ones will mostly be left alone.




Lele

[Clifton Santiago]

hey, nice tip Lele. I might try that out. Though, my IR calcs are usually so fast now that its hardly an issue.

[OPEN_RANGE]

Wow. Great feedback from all of you. I raised the HSph subdiv to 50 and Interp samples to 30. That took care of the splotches.

Lele,
I'm going to try setting the min rate to -6. The examples and times show this clearly works well. It appeared your examples are using light cache for secondary bounce. Would this apply to QMC for secondary bounce?

My, very limited, understanding is that the seperation between min and max determines the number of passes made before final rendering. It would seem that more passes = more time. Your example shows otherwise.

Does rendering at higher vs lower resolution require adjustment to the min- max, HSph subdiv and Int sample settings?

[studioDIM]

it's QMC for secondaries (plus a gamma of 2.2 in color mapping).
Since the passes at -6 are coarser, they take little time in evaluating vast areas.
Then, once the sampling is "good enough" the next pass will simply reuse the coarser pass' sample and skip over.
Hence the speed gain.

Of course, it's entirely wasteful is you have a very detailed image with little or no flat expanse, and the standard behaviour may apply better (more passes= more time).

Lele