Motion Blur Issue/Question?

[kriscabrera15]

Hey guys am rendering out some helicopter rotors spinning, and i have a question. At the moment my settings are .5 frames and .5 interval. Now my question is if i am using a setting of .5 frames why when rendering the times line goes to say 5.75 frames instead of 5.5 when evaluating? am i reading this wrong?

Thanks!

[Morbid Angel]

because the offset is .5, so its not sampling from frame 1 to frame 2, but from frame 1.5. The proper way is to have the interval set to 0.

[kriscabrera15]

thanks dude so here is a question when cache out fast spinning object ie like helicopter rotors, what export settings would give the proper motion blur if you using alembic or fbx in order to get all the information in between frames? i am having a bit of an issue getting these to mblur right when sampling in between frames, rotors keep flying in space.

Also if the offset it at 0 then should your frames be set to 1?

[Morbid Angel]

If you mean duration, it should be set to 0.5, which means shutter of 180 (typical camera shutter), 1 would produce shutter of 360, not possible in reality (to my knowledge).
For export, especially for fast objects like rotors you would want to have about 6 subsamples to produce a smooth arch in the rotation. Typically I would set the geosamples for MB to 6, you can also override per object motion blur samples, so only rotors get 6, the rest can get 2 (less memory hog). So take your frame and divide by 6 you would get your interval for subframe. Typically 0.25 is more then enough.

[kriscabrera15]

oh am getting it a bit more, now when u say "take your frame and divide by 6" what frame are you referring to?

[Morbid Angel]

Well your frame is a float number, but usually its between 0.0-1.0 (duration) where for example if you have duration of 0.5 and interval center 0 then the motion blur will be sampled from 1.0>1.5. So you can take this value of 1 and divide by 6 (geo samples) and receive an geo sampling interval of 0.16666 (or round off to 0.16). For duration of 1>1.5 since its half you can multiply the 0.16 x2, and receive a geo sampling interval of 0.32, so this would be the number you need to get proper geo sampling with 6 geo samples in the duration of 0.5. At least this is what I think my math could be off though and I don't really both calculating it to this exact, you rarely have to, hence I just use the 0.25 cache interval for the geo caches to save me the headache, usually works well.

[kriscabrera15]

thanks for the knowledge man! once last question, so at the moment i have .5 duration and 0 interval and 6 geo samples, i exported an alembic with a low sample of -.2 and a high of .2 calculating at .5 frames. Now what happening is on some frames the propellers are completely going away from the helecopter. Am trying to figure out why but i cant. would you happen to have an idea? am thinking it an animation problem at this point. Like spinning way to fast??

[t.petrov]

I can give you one practical advise how to see at what frame times your samples should be.

The advise is the following:
1. setup your motion blur settings
2. put an animated object; a cube should be enough
3. render a frame
4. inspect the log for lines like:

[2013/Dec/2|13:48:52] V-Ray: Updating frame at time 1.75
[2013/Dec/2|13:48:52] V-Ray: Updating frame at time 2.00
[2013/Dec/2|13:48:52] V-Ray: Updating frame at time 2.25

This is for the duration=0.5, center=0.0, geometry_samples=2
And this is the output with the above settings, but center set to 0.25:

[2013/Dec/2|13:53:47] V-Ray: Updating frame at time 2.00
[2013/Dec/2|13:53:47] V-Ray: Updating frame at time 2.50

In the first case there are 3 lines logged, because V-Ray always samples the static/non-animated
objects at the start of the frame. V-Ray also knows how many samples to do for each objects and in
this case if I open a vrscene exported from this rendering I'll see that my object has only two samples
- one at 1.75 and one at 2.25.

On the other hand the formula which can be used to calculate interval between the samples should be:

sample_duration=frame_duration / (geometry_samples-1)

Hope it helps...

[kriscabrera15]

awesome! clears it up guys thanks for the help!

Thanks Morbid really appreciate it!

Ok so in my scene i am using .5 duration and 6 GeoSamples.
Doing the math below i cam up .083 rounded off to .1. meaning an interval of .1 That sound correct?

Sry for the 21 question wanna get down to figuring out this issue am having here

Thanks in advanced!

[Morbid Angel]

Sounds about right I was a bit off on my numbers, should have divided the 0.16 by 2 no multiplied.

The rotor disappearing, could be any number of things, for example encountering a gimball lock so at certain angle the propeller is not rotating properly. Did you try rendering without caching and see if you have this issue?

[kriscabrera15]

Thanks morbid!, yea that my next step is to render just the rig, make sure nothing is happening at cache time.

[Morbid Angel]

If its happening at the beginning and at the end, you need to pad your cache by a few frames like 5 frames at each end. Though, with alembic, I wouldn't be surprised if there are issues, we had so many problems trying to get the transforms right, it doesn't handle layered transforms really well.

[kriscabrera15]

Well it's actually happening in between frames, so when vray is calculating motion blur you could see the prop in the viewport going off in space. Trying to grab the rig and see if it renders fine

[Morbid Angel]

Well it's actually happening in between frames, so when vray is calculating motion blur you could see the prop in the viewport going off in space. Trying to grab the rig and see if it renders fine

This sounds like gimball to me. One way to fix it is to change the rotation order of the object which is driving it. You can also change the time slider to allow you to scrub in subframes, and you should see the weird motion appear in the viewport.

[kriscabrera15]

cant get ahold of the rig but the issue is the way the pivots are on the rig, its causing the models to fly off in space when caching subsamples.