Flickering test log

Decided to start a log for tackling the Final Gathering flicker problem.


Aug 5
So I found more areas of flickering yesterday as I played my animation to potential employers (noooes!) lol. I am beginning to think it's how I set my SSS - will look into this.


July 31
--------------------------------------------------------------------------------------------------------------------------------

TAKE 6 | NGF_s002_q001_s001 (skeletal hand) 

Render settings: Freeze FG map (built based on every 10 frames)

No other changes (except SSS samples 256).

Rendering using Batch Render

Render time per frame: ~2 min
Results: Flickers.


TAKE 5 | NGF_s002_q001_s001 (skeletal hand)
Render settings: Freeze FG map (based only on frame 1)
Increased shadow rays from 12 --> 64 (but this was still flickering using command line)
Rendering using Batch Render

Render time per frame: ~2 min
Results: NO FLICKERING!! But less accurate FG. Ah well, worthy compromise!

Oh...em...gee. If this is what was wrong imma cry.

I'm using Maya Batch Render instead of command line rendering right now and there seems to be no flickering in the first 20 frames that have rendered. I noticed the FG map seemed to be recalculated for every frame despite having set it to Freeze under the render settings in the command line rendering feedback. Could this be it?!!!!!!11111 After days of throwing out renders?!!

If this works, I'll try plugging in the FG map I built every 10 frames for greater FG accuracy (instead of only using the first frame which might be sketchy). But I need to read more about FG - if I'm rebuilding every 10 frames, won't that show the flickering? Will have to test...

Still doesn't solve my problems for NGF_s002_q002_s001 though - I might try to build a map based on every 10 frames and see how that goes.

I got a bit sloppy with my testing though and changed too many variables at once, will have to go back and try to isolate the problem!

TAKE 4 | NGF_s002_q001_s001 (skeletal hand)

Result = still flickers just as much as the original...

Hmm. I am stuck........

July 30, 2011
--------------------------------------------------------------------------------------------------------------------------------
TAKE 4 | NGF_s002_q001_s001 (skeletal hand)
Render settings:

Ok well I've deduced it's the combo of FG and SSS (surprise) because there is no flickering when I render SSS without FG. I did intend to compromise but it really doesn't look nearly as nice without FG.

TAKE 2 | NGF_s002_q002_s001 (cellular)
Render settings:

Render time per frame: 8 minutes

Result: Lambert doesn't flicker anymore but SSS still does. Increased SSS samples to 256 but still flickers.

TAKE 1 | NGF_s002_q002_s001
Render settings: No change except I replaced the Blinn with a Lambert
Render time: ~3 minutes per frame
Result: It flickers in a completely new way! :( It looks fine when the camera is not moving, but once the camera moves, there is a ton of flickering... I mistakenly thought that swapping the shaders would fix it, as there is no flickering in NGF_s002_q002_s004. But I just realized that although that shot uses the same environment, there is also no camera movement.

Back to the drawing board. :(


NGF_s002_q001_s001
Turning off FG seems to have removed the flickering. Compromising time. :(

--------------------------------------------------------------------------------------------------------------------------------

July 29, 2011
TAKE 3 | NGF_s002_s001_s001 (skeletal hand)
Render settings: Freeze FG map, Set Optimize for Animation, Max/Min radius to 3/0
Result: Still flickers despite having a frozen FG map.

I am so confused... perhaps it's not the final gather?! Testing this now by turning off final gather altogether. I used a SSS shader like all of my other shots though, so not sure why this one is acting up so much. Wonder if it has to do with the Displacement...  Nevermind, SSS flickers in every shot... just varies in terms of how noticeable it is.

Next steps:
- remove displacement
- change SSS to lambert

--------------------------------------------------------------------------------------------------------------------------------
Sometime in June...

TAKE 2 | NGF_s002_s001_s001

Render settings: Increased FG Accuracy from 80 --> 200 (FG rebuild ON)
Result:  Still flickers, did not notice a difference.

TAKE 1 | NGF_s002_s001_s001

Render settings: Set to Optimize for Animation, Max/Min radius to 6/0.6 (FG rebuild ON)

Result:  Still flickers, did not notice a difference.

Will try to freeze the map next...

--------------------------------------------------------------------------------------------------------------------------------
RESOURCES

This guy's nightmare sounds all too familiar...

http://forum.mentalimages.com/archive/index.php/t-5325.html

But his solution in the end re: shadows probably isn't what I have. Will give it a shot if necessary.

Freezing final gather maps

Tips

Calculating min/max radius

Unsolved threads :(
http://forums.cgsociety.org/archive/index.php/t-506421.html

Try changing filter type
http://www.digitaltutors.com/forum/showthread.php?18801-SSS-Flickering-please-help-!

Support for my suspicion regarding scale conversion or radius:
http://forums.cgsociety.org/archive/index.php/t-684211.html

Final animation submitted

Here is my final animation I submitted for graduation!


I am still tweaking though! Never truly done, is it? :P

August schedule

1. Fix credits
2. Improve anatomy of nerve in Margot drawings
3. Re-render flickering hand - freeze FG map
4. Re-render flickering cellular environment - replace Blinn with Lambert shader
5. Fix DoF for cellular environment??

Final sign-off meeting!

Final Committee Meeting
Topic: final animation
Committee Members present:
Prof. Linda Wilson-Pauwels
Prof. Dave Mazierski
Michael Corrin
Geraldine Longo

Animation changes

NGF_s001_q001_s001
Make the 3 illustrations insets instead of full screen shots
Joyce: This looked really bad when I tried it, because having three insets and the Margot drawing created way too much visual noise. It was especially strange because there were now too many hands on screen and there isn't enough contrast between the images to differentiate between the insets and the main image. It might work better if they were more schematic, but due to time constraints, I am unable to redo the drawings.  So I am keeping it the way it was before!

NGF_s002_q001_s003
- "increasingly sensitive..." make impulse speed up faster and more obviously

NGF_s002_q003_s003

Cut out silence btwn NGF complexes being taken up in Margot shot.

Use insets to set up periphery and central (spinal cord)

Use zoom box to exit from molecular to go to Margot shot. Then when say "distance sites" label DRG. Then make transition in inset for molecular shot to new molecular shot, and do zoom box back in.


NGF_s002_q001_s001

Soften mask on Margot's chin to hide more of the nerve.


Labeling changes


NGF_s002_q001_s003

collateral nerve branch - leaderline shouldn't be at tip


NGF_s002_q002_s001
Keep histamine, etc - keep on screen, maybe fade out
Could also have the whole list show up first - would solve problem of having the pink particles spewing out


NGF_s002_q003_s001
Labels for cations in channel shot

Put arrow flowing through ion channel


NGF_s002_q004_s001

Label nucleus, cell body in DRG.


NGF_s003_q001_s001

Slower blend from pill to Margot holding pill
Too much blur on the hand


NGF_s003_q001_s003

Lost NGF label on motion graphic


Credit changes

- Nicholas Woolridge: Director (take out program), BScBMC, MScBMC
- Lab members - not lab
- Narration instead of voice talent for Shelley
- Alphabetical order of acknowledgments, take out titles
- Charles Lumsden - department of medicine, faculty of medicine
- QPRN Scholarship
- Change credits technical assistance --> instructors
- Marc --> title is senior lecturer
- Professor emerita for Margot

----------------------------------------------------------------------------------------
Vesalius Trust presentation ideas
- NGF binding to trkA
- DRG with pronociceptive
Say something about Toronto program just finished and animations will be submitted next year.
- Do breakdown of a few frames
- DoF and visual cues to maintain attention,
- Visually framing different areas on screen using DoF and palette.
- Design of composition rather than technical breakdown.
- NGF_s002_q004_s001: Inset part use limited colour palette and DoF to focus attention to raise certain elements to foreground.
Colour palette  - break down
Show snippets for each environment
 NGF-trkA vesicles formation
NGF_s003_q001_s003 - could be another shot


Possible Pfizer contact

David Gold - Pfizer

Mitac - funding agency

Getting there...

...I think! Sign-off meeting this Thursday...

I've been updating under the "to-do list" tab for anyone who's checking up on my progress. ;)

Particle depth

The linstep method of creating a depth map for nParticles doesn't seem to work for cached nParticles. The black-white shows up when you first enter in the expression, but once you change the frames it disappears! I guess it makes sense though because I'm plugging the rgbPP linstep stuff into the Runtime Expression, and if the particles are cached, then it's not really "running" anymore??

Anyway, Marc (my hero) came up with a different method of generating a depth map.
1. Create a depth render layer to generate the Set Range node.
2. Connect the Set Range node OutValueX (only x because only want one value?) to npCloudVolume - incandescence r,g,b

This works well with using the Measure tool to set up the Set Range oldMinX and oldMaxX.
I used Mentalray to render and used a UseBackground shader on the other objects.

Goodbye Hardware Render Buffer! :(

I should start creating a blog tag called "Marc is magical" or something. :)

Mudbox fail

UPDATE: Finally got it to work! I had to redo the textures on a combined mesh though.
Settings in Mudbox > Subdivision method > 32 bit EXR (rgba) file.
Make sure the plugin to read OpenEXR is loaded in Maya
Put in the file texture and make sure Alpha as luminance is checked off under Color Balance.
Maya appears to reject 32-bit float TIFFs and throws up a bunch of error messages but mentalray still renders it. No problems with EXR though.

The texture isn't as detailed or bumpy as I'd like, but I'm not willing to sacrifice render time at this point in the game.

-----------

I think Mudbox has become my most hated program in the span of 10 hours. I got my skeletal hand texture down in Mudbox - but frustratingly, have not been able to export it properly. The raytrace method creates insane amounts of artefacts in the Displacement Map. The Subdivision method just refuses to work with my object (containing multiple meshes). Even when I managed to get the subdivision method to work by going back to the original .obj and combining the mesh, and redoing the texturing - I ran into new issues. Maya can't read 32-bit float IFFs (or other file formats) properly. Considering Mudbox is an Autodesk product meant to work with Maya, I am pretty frustrated with this.
If I use 8-bit, Mudbox creates a displacement map in which the background is grey. Furthermore, Mudbox enjoys crashing every time I export if I keep "preview as bump map" checked.

Ahhhhhh!

I tested Mudbox two months ago on a simpler object and it seemed to work when I exported and brought the map into Maya. Otherwise, I would not have waited until now to texture the hand - I did not anticipate all these issues (silly me, I should know better by now).

Look at this beautiful displacement map Mudbox gave me through Raytrace method. I believe this is blooper-reel worthy...

Then I got pretty annoyed and decided to actually model the bumps (not a good idea lol)

Basically my options are:
1. Paint textures by hand (trying this in PS, but my UV maps aren't very good because I didn't anticipate this so I will probably have to re-UV map, very time consuming)
2. Apply a procedural texture - this has potential, but I will have to really work on it to get something that looks good.

This sets me behind schedule quite a bit. I'm thinking I may just
1. Actually model some larger grooves in just to indicate it is an arthritic joint
2. Use bump vs. displacement to save on render time, though it doesn't look as good

Mudbox, it could've been beautiful... :(

Will sleep on it and think some more.