PDA

View Full Version : vray LWF, new discovery?


crystal3d
02-04-2011, 07:56 AM
before getting into any false hopes, i feel i should confirm this.


today, i ticked off , "clamp output" 1.0" and after 6 months , glossy reflections became fine again, they were so dimm and interpolated, now they are fine, but some white (probably aliased" dots appeared on renderings, so i set the" clamp output to 2.0" and raised DMC to 1-8, and that solved it.
now i wonder if it will cost me something else in the long run or is it a reasonable work-around for that?

achoury
02-06-2011, 07:39 PM
Hi
I'm not sure if this is the same as mental ray since you can restrict the value further than 1.0, but in general the "clamp" is just a simple function that allow you to go from HDR to LDR or from float to 8bit (0.0 to 1.0 range) in other word any negative values become 0 and any higher values than 1 or greater than 1 become 1.0 or if you like clamp = "normalized value".
I'm curious too what clamp 2 is for !
/Rachid

noouch
02-07-2011, 11:17 AM
The clamp value defines what the maximum pixel value should be. It's useful to avoid individual overbright pixels where a single sample is an insanely bright value and therefore more or less overrides the other samples. If you're calculating a prepass, it could very well be that these overbright samples are required for a correct result.

crystal3d
02-08-2011, 02:55 AM
if tired a fly through sequence with clamp out put 2.0, and antialias values had to go sky hight in order to handle the wide color range , image was so grainy in DMC AA,1-8,
it was 1-64 for the AA to output a noise free image...

my conclusion is , true HDR needs alot of antialiasing ,thus feasible for stills and 90% animations are infact LDR.

CMIIAW!!...(correct me if i am wrong)

crystal3d
02-08-2011, 02:58 AM
If you're calculating a prepass, it could very well be that these overbright samples are required for a correct result.

so clamp should be set on, "render time"only? the calculations should be with out "clamp output" but i guess it is already like that, clamp kicks in only in rendertime as i observe...

michaelaaut
02-11-2011, 12:15 PM
I wouldn't say it's a new discovery... that floating points values are beyond 1 isn't quit unusual.

I would optimize shaders, and I would check all the light values plus the physical camera settings, if used. Rendering first with an override grey material and see if floating values beyond 1 appear as well.

CaptainObvious
02-12-2011, 03:40 PM
Clamping is basically a way of filtering out too-extreme values. Suppose you're sampling the same pixel four times and end up with three full black values (zero), and one that's, say, 10 000. Then the final value of the pixel is 2500. With me so far?

Now, imagine that you've set the clamp level to 1.0. What happens now is that the three samples that are zero, they stay zero. However, the single much-too-bright value will be clamped to 1.0 instead of 10 000, and the new average will be 0.25 instead of 2500.

This is a great way of dealing with the "fireflies" that can occur in rendering, but it does mean you A) lose most of your super-white dynamic range and B) you kill small bright details.

A good workaround for these problems is to set the clamp value to something greater than 1.0. If you clamp at, say, 5.0 instead then you can strike a much better balance between maintaining dynamic range and details, while still controlling some of the worst fireflies.

crystal3d
02-16-2011, 05:28 AM
Clamping is basically a way of filtering out too-extreme values. Suppose you're sampling the same pixel four times and end up with three full black values (zero), and one that's, say, 10 000. Then the final value of the pixel is 2500. With me so far?

Now, imagine that you've set the clamp level to 1.0. What happens now is that the three samples that are zero, they stay zero. However, the single much-too-bright value will be clamped to 1.0 instead of 10 000, and the new average will be 0.25 instead of 2500.

This is a great way of dealing with the "fireflies" that can occur in rendering, but it does mean you A) lose most of your super-white dynamic range and B) you kill small bright details.

A good workaround for these problems is to set the clamp value to something greater than 1.0. If you clamp at, say, 5.0 instead then you can strike a much better balance between maintaining dynamic range and details, while still controlling some of the worst fireflies.

thus doubling the min-max value of antialising, 1-30 seemed to work in my case but it was not feasible for my animation...

Frubes
02-21-2011, 04:23 PM
Sub pixel mapping can help with overbright pixels. You may also want to check to see if you are using the "highlight glossiness" on your materials. This will introduce a fake highlight into your shaders which will more often than not cause superbright pixels. We try to avoid them at work as oversampling the shaders in order to irradicate them is inefficient.

CHRiTTeR
02-21-2011, 07:18 PM
nothing unusual here, those overbright hard pixels even happen in real life digital cameras ;)

Respawned
02-21-2011, 08:49 PM
I think it's best is to try and set up the highlights first without using clamp, then only enable it as a last resort.
I have however used it on purpose sometimes to get an over exposed photo look, like in this render from my gallery (I used clamp and sub pixel mapping).
http://features.cgsociety.org/newgallerycrits/g13/210113/210113_1246279334_large.jpg

CGTalk Moderation
02-21-2011, 08:49 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.