PDA

View Full Version : Physical render random color shifts...


interactiveBoy
03-27-2012, 01:05 PM
We've done several projects using physical renderer in C4D, and in every one of them there have been random frames that have a completely different color range. For example, in most frames the mood is dark and contrasty as it should be, but then for one random frame, everything looks blown out. I thought it had to do with a net render glitch, but last night I ran a render on just my station.

In this animation the only animated object is the camera.

Notice how one of the frames is completely blown out and grainy.

What gives? We've had this issue on every project that has used physical renderer, costing us many hours of trouble shooting.

interactiveBoy
03-27-2012, 01:51 PM
So we've been running some tests, and it appears that motion blur in the physical renderer is causing this. Motion blur is the number one reason we are using the physical renderer.

This is really frustrating. Has anyone else had this issue? What are the work-arounds?

bunter
03-27-2012, 03:14 PM
Have you sent the file to Maxon tech support? They are very helpful and responsive in my experience.

chi
03-27-2012, 03:16 PM
is it possible to post a stripped down version of the file?

interactiveBoy
03-27-2012, 04:25 PM
I have sent the file to tech support. We'll see what happens. Previously, on other projects, it was too difficult to strip it down.

I'll update this thread when I hear back from them.

Thanks!
:thumbsup:

LucentDreams
03-27-2012, 05:40 PM
my initial suspicion was a computer on the farm with a weird LUT, but if your sure it's the motion blur causing it I'd look at when it happens, is is the first frame a specific computer renders? Which physical sampling mode is being used, fixed, adaptive, does baking animation help etc.

interactiveBoy
03-27-2012, 06:15 PM
As I mentioned, this was rendered on a single station. I suspected net render at first also, but last night I rendered overnight on my station only, and the problem persists.

I'm gonna wait and have tech support look at it.

interactiveBoy
03-27-2012, 10:38 PM
Well I got a response from tech support, but they didn't really look thoroughly into the issue, as their answer does not make sense. :banghead: Anyone here care to have a look at the file?


One of the frames that blows out is 488. Please render to picture viewer - frames 487 and 488 in the least, so you can see the difference.

Things I've tried: Turning motion blur off prevents the blow out.
Turning off the square spot light prevents the blow out.

If it were happening on every frame I would be able to blame it on one or the other...but it is randomly happening. (Well, random in the sense that I can't figure out why some frames are fine and others are not)

Tech support said it was the emitter with the lights in it blowing out the scene. I disagree. It may be contributing, but why only on certain frames?

Very frustrating to say the least. I think physical renderer has some major issues with color shifting and lights blowing out randomly. This has happened on other projects as well.

EricM
03-27-2012, 11:33 PM
Don't know yet what's causing it, but on my computer, your scene exhibits the same flash that you descibes on frame 488 with MB on.

interactiveBoy
03-28-2012, 12:25 AM
Thank you Eric for checking it out. I'm hoping to get several sets of eyes on this file and maybe at least be able to come up with a reproduceable bug that can be done in another file following certain steps.

We recently did another project that did not have emitted lights or anything out of the ordinary and there are several frames that shift color for no apparent reason. (Notice I say "apparent") I'm not totally ruling out user error here, but even so, it seems that there's more to using physical render than meets the eye...and certain things will just not play well together.

chi
03-28-2012, 03:19 AM
I don't think tech support was totally wrong on this one.
when the light/particles are born/die they are getting a crazy movement...as if the particle ID changes...this would make the renderer think that the particle has traveled a great distance over 1 frame.

and if you look at when errors start cropping up, it seems to happen where most of your particle/lights are dying, around the 480 mark.

Try setting the particle life to be as long as the project.

I tested this over a trouble area and it seems to work fine.

Shademaster
03-28-2012, 07:38 AM
I have noticed that brigtness values over a couple of hundred percent generate black clipping like spots. I had the same problem with a Sun reflection in water.

Don't know a fix for this, maybe dial down the brightness?

interactiveBoy
03-28-2012, 01:33 PM
Chi - Thanks so much for your time and for explaining what's going on here. So tech support was right only in that it involves the emitter in some way. However, why does the problem go away if you kill the square spotlight? If you leave the file as is (no change to the particle lifetime) but disable the square spotlight, the scene no longer gets blown out on frame 488. There seems to be more to the problem than meets the eye.

So, anyway, I think what you're saying is that the renderer thinks the particles are moving a great distance, and therefore the motionblur is randomly blowing out frames? Is that correct? I'd call this a bug. How do you have that information? How are you able to determine that the particle IDs are changing? I didn't realize you could do that with the standard emitter.

This color shift has happened in projects where there are no emitters at all. In one particular project, it is a COLOR shift, not just a brightness change in the lighting.

Physical renderer is BUGGY. Plain and simple.

Thanks for the work-around. I have extended the lifetime of my particles to compensate for this buggy behavior. However, what am I gonna do when I need to have a scene where my particles die DURING the scene?

By the way - if I sound negative...it's unintentional. I'm just trying to get to the bottom of this. I appreciate your insight!!

:beer:

HolgerBiebrach
03-28-2012, 01:40 PM
I don't think tech support was totally wrong on this one.
when the light/particles are born/die they are getting a crazy movement...as if the particle ID changes...this would make the renderer think that the particle has traveled a great distance over 1 frame.

and if you look at when errors start cropping up, it seems to happen where most of your particle/lights are dying, around the 480 mark.

Try setting the particle life to be as long as the project.

I tested this over a trouble area and it seems to work fine.

Yes, its certainly because of the ID-change of the particles. When particles die all other paticles get a new ID. Instead of letting the particle die you can assign them to another Paticle-group.
I used to have similar problems....now I allways usea "dead-particle"-Group.

vid2k2
03-28-2012, 02:33 PM
Physical renderer is BUGGY. Plain and simple.



FWIW,
I rendered out frames 470 to 490 without any problems ...
v12 (without Physical renderer).

interactiveBoy
03-28-2012, 02:55 PM
FWIW,
I rendered out frames 470 to 490 without any problems ...
v12 (without Physical renderer).

Right. It is a physical render issue

interactiveBoy
03-28-2012, 02:56 PM
Yes, its certainly because of the ID-change of the particles. When particles die all other paticles get a new ID. Instead of letting the particle die you can assign them to another Paticle-group.
I used to have similar problems....now I allways usea "dead-particle"-Group.


How does one do this? I'm using the standard particle emitter. This is not thinking particles.

HolgerBiebrach
03-28-2012, 04:03 PM
this is only possible with Thinking particles. Standardparticle can not be assigned to a particle group. You need to redo your particle-setup with TP I guess.:blush:

chi
03-28-2012, 05:17 PM
Chi - Thanks so much for your time and for explaining what's going on here. So tech support was right only in that it involves the emitter in some way. However, why does the problem go away if you kill the square spotlight? If you leave the file as is (no change to the particle lifetime) but disable the square spotlight, the scene no longer gets blown out on frame 488. There seems to be more to the problem than meets the eye.

So, anyway, I think what you're saying is that the renderer thinks the particles are moving a great distance, and therefore the motionblur is randomly blowing out frames? Is that correct? I'd call this a bug. How do you have that information? How are you able to determine that the particle IDs are changing? I didn't realize you could do that with the standard emitter.

This color shift has happened in projects where there are no emitters at all. In one particular project, it is a COLOR shift, not just a brightness change in the lighting.

Physical renderer is BUGGY. Plain and simple.



I guess from the information I had.
Flickers in the render, a frame range when they happen and then the tech support mention of the emitter I stripped the scene down to just those elements...plus a cube so I could see brightness.

Then I tried a render with the visible light off...everything worked.

so it was a visible light in an emitter.

so I rebuilt the emitter in a new scene so that I could look at what was happening to only a few particles.

Then I rendered again.

When a particle/light dies I noticed that there is a blur between the light/particle that just died and another particle/light that was still living.

So this showed a shift in object IDs that it didn't know how to deal with.

so from there I started working toward towards a solution.

This isn't so much a bug as a limitation, I don't think they ever thought that the standard particles needed to have unique IDs when they were written 10+ years ago or however old the standard particles are.

As far as creating a particle that dies...you will likely need to stretch the brain and start delving into TP.

@ holger - I would be interested to see a screen of a "dead" group. :)

interactiveBoy
03-28-2012, 06:07 PM
Man Chi - Thanks for your breakdown. While I agree that it is a limitation in regards to the particle emitter...I still am dumbfounded as to why it is happening like it is when those emitted lights are so far away from where the blow out is happening in my original scene. Also, the problem goes away if you disable the square spotlight in my scene (the one called computerLight) You can leave the particles completely alone (don't change anything from the settings I have in the example file), and just disable the computerLight. The problem goes away. That's why I believe there is much more going on here.

Your research and solution will indeed get me a work-around for this exact issue, in this exact scene. Unfortunately this same type of issue happens in scenes that don't even have an emitter.

I'm on the verge of abandoning the physical renderer because of this.

I'm very thankful for your help.

:bowdown:

interactiveBoy
03-28-2012, 06:23 PM
so I just tried making a new scene with nothing but an emitter, visible lights and a cube as you did Chi. Indeed that shows a streak between two lights as a particle dies.

(nice find!)

However, the same does not happen if I change the particle to be a sphere with a luminous texture. (See attached images)

This limitation seems to only affect visible lights? Sorry to sound like I'm repeating myself here, but it is a bigger issue. Again, I'm thankful for the insight and the way you found a work-around...but there are more problems lurking under the hood. :)

Tech support gave me a response that only touched the surface.

chi
03-28-2012, 09:14 PM
http://dl.dropbox.com/u/8306001/lights.PNG

looks like it is an issue with the lights in general.
killing a light causes all lights in the scene to shuffle around.

another workaround then might be to use a luminant disc for visibility + a light for illumination.

either way looks like an issue that couldn't really have been foreseen...but is unfortunate.

Phys. render is a growing beast and I still love it...even with it's quirks.

LucentDreams
03-28-2012, 11:22 PM
If you did want to convert the solution to TP yuou coudl use a matric object set to generate thinking particles, turn off the dying and then use the die group

interactiveBoy
03-29-2012, 12:21 PM
You guys rock. Thanks for your time looking into this.

I got comfirmation from tech support that this is a bug. I've offered to send them 2 other files that have similar/but different problems with color shifting on random frames with no emitters at all in the scene.

I'll start a new thread for that.

CGTalk Moderation
03-29-2012, 12:21 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.