PDA

View Full Version : Help with DP Spline Deformer



Chrusion
09-29-2014, 10:58 AM
I'm trying to do something similar to the 3D ribbons in this thread >>> 3D-Ribbons (http://forums.newtek.com/showthread.php?143496-3D-Ribbons), but instead of multiple ribbon meshes, I have a "ribbon" of points (point cloud) to which a gravel/stone mesh will be instanced. What I'm shooting for is a conveyor belt of gravel moving up the belt and falling off the end.

The point cloud 'ribbon' is modeled with the head at 0,0,0 and the tail a few meters on +Z. I added four nulls, arranged them with #1 at the head of the point cloud, #2 a meter down and a bit forward, #3, 2 meters down, and #4 4 meters down. They are children of the point cloud.

In the point cloud object props I add Spline Deformer node to Displacement Node, connect Delta to Input, open Deformer Props, click Add Children, and exit.

The point cloud is now distorted to smithereens. I ignore that and edit the pitch of the 4 nulls to get a correctly oriented smooth spline. I then click Record Spline Rest Position in Spline Deformer props. This results in the point cloud ribbon being sort of restored to normal, but it remains highly distorted on X and Y to a very narrow, straight shaft with about a half meter of the head being detached by about a half meter gap.

I then add a Scalar node connected to Shift Offset and envelope its value from 0 to 100 over 90 frames.

Scrubbing the timeline results in the point cloud NOT MOVING along the spline, but the straight portion of the clouds tail after the gap rotates like a spoke upward, while the "detached" group of head points moves downward by about 1/4 meter.

Stripped down scene >>> 124543

Need help getting the point cloud to move along the spline as I've seen numerous poly meshes do without any such "separation" or extreme deformation as mine is experiencing before even "entering" the spline. IOW, what am I doing wrong!!?

Chrusion
09-29-2014, 01:43 PM
I think I figured it out.

It appears you have to know the length of your object, then in the Deformer props enter desired number of nodes in the Create Item(s) field, divide length of object by this number of nodes (nulls) minus one and enter the value (X) in the Distance field, then click the Create Item(s) button. This creates child nulls along the length of the object (evenly spaced by Distance X) and auto Record Spline Rest Position. After this initial setup, you then manipulate the nulls (position/rotation) to create the desire path.

Not sure how the Load Pose From Spline button works (load a modeled curve object?), but assume the rest position has to be set manually or something. Wish there was a step-by-step tutorial on every option, param, and node input, etc.

Not sure which buttons to click an in what order when you already have a set of child nulls, positioned as desired, that have been "imported" via the Add Children button. This just screws up the object to be deformed, no matter what order I clicked the other buttons for Record Rest Position, Reset Item Position, Align Item to Spline, etc. Doing such guessing and experimenting does not work at all.

Chrusion
12-29-2014, 08:45 AM
Dennis... New problem...

DP Spline Deformer doesn't single-pass photoreal motion blur! Am I missing something? I don't want to activate multi-pass MB, because it multiplies render time by the number of passes selected and generates the dreaded banded or stepped effect, which the rest of the anim does not have up to this point. Meaning, I'm at the point where the cloud of points I'm deforming with the displacement node is in extreme closeup. The points are instanced with a rocky nugget object. They DO blur (single pass photoreal) only when the camera is moving... even when the motion of the two (deformed point cloud and camera) are the exact same speed and parallel to each other. Looks a bit like bullet time. But as soon as the camera came to a stop, the blur stopped and now all the nuggets quickly passing by the camera have no blur.

Frame 119 as camera comes to a stop
126330

Frame 120 with camera at a full stop
126331

Obviously the deformed points/instances have no SELF blur, as indicated by the lack of blur the farther they are away from the camera, but the effect of suddenly no blur on the frames after the camera stops is visually distracting.

If there is a method to get the points/instances to self blur, I'll just rerender the first 119 frames.

The node setup uses a Constant > Scalar with a linear post behavior envelop (plugged into Shift Offset of Spline Deformer node) to drive the motion of the point cloud forward thru the Deformer Nodes/Nulls. Is there another way to move the object thru the nodes that provides intra-frame motion info?

ernpchan
12-29-2014, 09:33 AM
What happens if you bake the deformation to mdd?

dpont
12-29-2014, 10:16 AM
126332

Very simple Spline Deformer animation,
single-pass PRMB, camera not moving.

Denis.

Chrusion
12-29-2014, 10:16 AM
Just tried scanning/baking to MDD and using your MDD Pointer node (replacinging the deformer output) hoping that reading back MDD motion would provide single-pass motion blur like your MDD Emitter node does for Particles... Nope.

I even tried jittering the Mapped Time envelope (1-frame saw tooth so that post behavior repeat offset created an inclined ramp with value = 1.0 at frame 30)... Nope.

Looks like your MDD and Spline Deformer nodes don't interpolate motion vectors or time on an intra- or inter-frame basis. Super Heavy Sigh of Sadness.

Any way to code that before my render deadline ends in the next day or so? LOL. :-P

Chrusion
12-29-2014, 10:22 AM
Not understanding your pix, Denis. I assume that's a moving "ribbon" cloud of points deformed by Spline Deformer as I've described and rendered as you say. So, why are your instances applied to those points blurring and mine are not? Or is that, instead, a tube mesh object wiggling in the middle? If the latter, the two can then be said they're not the same... polys vs. points + instanced geometry.

dpont
12-29-2014, 10:30 AM
This is just a tube mesh deformed by the node,
so PRMB works with the node for a classic nodal deformation.

Denis.

Chrusion
12-29-2014, 10:31 AM
Here's the stripped down scene to prove there's no PRMB on this use of Spline Deformer.

dpont
12-29-2014, 10:46 AM
Here's the stripped down scene to prove there's no PRMB on this use of Spline Deformer.

Not a specific limitation of the node,
but a limitation of PRMB, this works (actually?) with a mesh raytraced in the LW Renderer,
may be because points or point-polys are not raytraced.

Denis.

Chrusion
12-29-2014, 10:52 AM
Hmm... how are MDD scanned PFX particles, then emitted by MDD Emitter, with the same instanced object PRMB'd if the limitation is in LW? Instances are being volumetrically raytraced in both situations.

dpont
12-29-2014, 11:30 AM
The node itself is just sending deformed point displacement info to LW,
whatever the geometry, from single point or belonging to a poly,
PRMB needs several steps or intermediate infos for the interpolation
of samples in the renderer.

When I said not raytraced, I mean a different situation,
but not may be impossible to implement in LW.

Denis.

Chrusion
12-29-2014, 12:17 PM
But, doesn't LW just interpolate btwn frames? That is, look ahead to frame n+1 and do its stocastic sampling btwn obj. transform at frame n and n+1? If so, then doesn't LW poll the displacement node system the same way? If it doesn't, guess I'm screwed and will just have to live to stroboscopic, non-motion blurred gravel falling off this conveyor belt.

Chrusion
12-29-2014, 12:22 PM
But wait! If your deforming tube was blurred, then LW IS looking ahead, right? So, it must be something to do with the Instancer system, yes? Try that... instance something to your tube and see if only the tube is blurred and the instances are not.

dpont
12-29-2014, 12:42 PM
No for sure, instances on a deformed object aren't prmblurred.
actually in LW, the meshinfo system updates deformed/displaced mesh
only at current time, prmb can do that but it is an exception
and probably doesn't use this system.

Denis.

Chrusion
12-29-2014, 12:48 PM
Bummer. Thanks for the confirmation.