Hard to remove noise & fireflies when using rounded edges on metals.

[The Pixel Artist]

I've noticed that when using the "Rounded corners" option on metallic materials (highly reflective/glossy/etc) it will produce a lot of glossy noise or fireflies that just won't clean up, even after a few hundred passes.

Not sure if there's a way to fix this or if it's just a bug.  Any ideas?

[Ondra]

those look like sun reflections, do you have sun visibility to reflections enabled?

[The Pixel Artist]

No, it's actually rendered with an HDRI.

BTW, is there a way to render a scene with the Corona Sun/Sky combo, but also use an HDRI for the background/reflection/refraction?  I use this setup in Vray, but couldn't get it to work in Corona.

[Utroll]

Just for reflection I would light with HDRI and plug override reflection in corosettings.

If you mean adding a sun plus or less in the position of the hdri sun to add some hard shadows, I think you can achieve that too but in the very sensitive parameters world of corona : tricky :) I could achieve that with an hdri having a global multiplier of 20 and a sun having it's power at 0,2.. it seems that the important thing about is in the corona sun tab : separate sampling (below the choice of sun color input, by sky, or direct temp. choice,..)

The sensivity of corona + the exposure/highight compression makes it not so easy to figure at first but if you start by lightning only with hdri, exposure at 0 and highlight comp. at 1, you can tweak the multiplier of your hdri till you get correct result, may be a bit underexpose and then you can start 'adding' the sun, start from 0,05 even, and going up...

I'm very willing to ear about this separate sampling parameter, I think it's usefull when corona sun is not used with corona sky map.. but someone might explain more may be ?

Now don't expect purist too welcome that way, it is FAKE to add a sun to a sun. Some might say that a correctly exposed hdri already contains a sun.. you can 'paint' with photoshop a pure bright spot on your hdr to force shadows directionnality.
(a bit like http://www.cgtechniques.com/tutorials/synthhdr.php you might find clearer sources about)

[Ludvik Koutny]

What's your PT samples value?

[The Pixel Artist]

What's your PT samples value?

Default values, I believe, as I'm actually not sure which the PT samples are (I'm assuming that mean path tracing samples).  I don't see that listed, is that the same as MSI or the GI multiplier maybe?

Either way, I turned everything back to default for this image.

[Ondra]

still, it looks like the HDRI has some bright spot acting as a sun. In that case it creates the usual problem with fireflies. As a test, try setting environment reflect override to something like constant color. The problems should go away.

[Ludvik Koutny]

Do not ever do that, ever....  setting Enviro override to constant color may solve the issue but it will significantly lower quality of your output. The problem here is insufficient amount of AA samples, so try to lower PT samples to something like 4, so more of the power goes to AA instead of GI.

Another source of problem could be using fine bump map in combination with small chamferred edges or rounded corners feature. If that is the case, then the result you are getting is pretty much right.

[Ondra]

I just need the image for debugging

[The Pixel Artist]

I just need the image for debugging

Ok, will do.

Do not ever do that, ever....  setting Enviro override to constant color may solve the issue but it will significantly lower quality of your output. The problem here is insufficient amount of AA samples, so try to lower PT samples to something like 4, so more of the power goes to AA instead of GI.

Another source of problem could be using fine bump map in combination with small chamferred edges or rounded corners feature. If that is the case, then the result you are getting is pretty much right.

Rawalanche, thanks for your help.  Now, forgive my ignorance, where do I find the "PT samples" you've mentioned.  Unless I'm just not seeing it, I don't see that parameter anywhere.  Is that maybe the same as MSI?

Also, reading in another thread I found during a search, you mentioned that metals should generally never use pure black nor a level of 1, but something like .55 for aluminum, .7 for steel, and .95 for mirrors, increasing the IOR to increase the reflectivity.  Now, my questions is, what should I set the Reflection color at, 255 or should it be something under 220?

[Ludvik Koutny]

I just need the image for debugging

Ok, will do.

Do not ever do that, ever....  setting Enviro override to constant color may solve the issue but it will significantly lower quality of your output. The problem here is insufficient amount of AA samples, so try to lower PT samples to something like 4, so more of the power goes to AA instead of GI.

Another source of problem could be using fine bump map in combination with small chamferred edges or rounded corners feature. If that is the case, then the result you are getting is pretty much right.

Rawalanche, thanks for your help.  Now, forgive my ignorance, where do I find the "PT samples" you've mentioned.  Unless I'm just not seeing it, I don't see that parameter anywhere.  Is that maybe the same as MSI?

Also, reading in another thread I found during a search, you mentioned that metals should generally never use pure black nor a level of 1, but something like .55 for aluminum, .7 for steel, and .95 for mirrors, increasing the IOR to increase the reflectivity.  Now, my questions is, what should I set the Reflection color at, 255 or should it be something under 220?

I think PT samples were renamed to GI samples multiplier in dailies. As for reflection values... good practice is to never go over like 240, but having 255 in reflection should not have nearly as destructive effect as having 255 white in diffuse slot :)

[vkiuru]

I just need the image for debugging

Ok, will do.

Do not ever do that, ever....  setting Enviro override to constant color may solve the issue but it will significantly lower quality of your output. The problem here is insufficient amount of AA samples, so try to lower PT samples to something like 4, so more of the power goes to AA instead of GI.

Another source of problem could be using fine bump map in combination with small chamferred edges or rounded corners feature. If that is the case, then the result you are getting is pretty much right.

Rawalanche, thanks for your help.  Now, forgive my ignorance, where do I find the "PT samples" you've mentioned.  Unless I'm just not seeing it, I don't see that parameter anywhere.  Is that maybe the same as MSI?

Also, reading in another thread I found during a search, you mentioned that metals should generally never use pure black nor a level of 1, but something like .55 for aluminum, .7 for steel, and .95 for mirrors, increasing the IOR to increase the reflectivity.  Now, my questions is, what should I set the Reflection color at, 255 or should it be something under 220?

I think PT samples were renamed to GI samples multiplier in dailies. As for reflection values... good practice is to never go over like 240, but having 255 in reflection should not have nearly as destructive effect as having 255 white in diffuse slot :)

If this is true about the reflection value, maybe it would be a good idea to make the default value something other than 255 white? The though that it could cause problems, even minor ones, never crossed my mind.

[Ludvik Koutny]

If this is true about the reflection value, maybe it would be a good idea to make the default value something other than 255 white? The though that it could cause problems, even minor ones, never crossed my mind.

Well, then you have to care also about changing output of any reflection map that has white areas on it. So it's easier to just set reflection level to 0.9, and then it does not matter what color or map is used to drive reflectivity ;) That's why those level values are there after all, to define how strong white color is ;)

[vkiuru]

If this is true about the reflection value, maybe it would be a good idea to make the default value something other than 255 white? The though that it could cause problems, even minor ones, never crossed my mind.

Well, then you have to care also about changing output of any reflection map that has white areas on it. So it's easier to just set reflection level to 0.9, and then it does not matter what color or map is used to drive reflectivity ;) That's why those level values are there after all, to define how strong white color is ;)

Yeah makes sense.

[Ondra]

maybe it would be fastest if you could post the scene setup with camera, environment map, and one of the problematic objects