Cloth Simulation tutorials by Donovan Keith in R18 give different result
Posted: 06 August 2017 09:48 PM   [ Ignore ]  
Total Posts:  9
Joined  2013-10-16

Hi, I have been following the Cloth Simulation videos by Donovan Keith.
https://www.cineversity.com/vidplaylist/cloth_simulation/cloth_simulation_part_04

I am using R18 and have noticed that I am getting very different results to the tutorials, specifically where the belt tag is being used.

An example is during the Cloth Simulation, Part 04: Create a Drop Curtain Falling Onstage tutorial where the cloth is belted to 5 cubes, in the video the cloth hangs nicely from the belts, however when i follow the same steps in R18 my cloth hangs down very low and distorts the mesh much more than the example.

I am pretty sure I followed each step, however maybe I have missed something, or does R18 simulate differently to R16?

Has anyone else experienced this issue using R18?

I have posted two screenshots (the tutorial and my scene) and also my scene file.

File Attachments
dynamics_cloth_cineversity_tut_04.zip  (File Size: 750KB - Downloads: 5)
Profile
 
 
Posted: 06 August 2017 10:51 PM   [ Ignore ]   [ # 1 ]  
Moderator
Avatar
Total Posts:  4428
Joined  2011-03-03

Hi picturedrift,

Thanks a lot for the file, that helps to answer.

If you go to Part 4 and there to 06:40 [min:sec] you will see that some settings are set to zero: Flexion, Bounce.
Try it with these settings and with the iteration set back to 50, to get a similar result.

Perhaps download the provided files of part 4 and compare those with your set up, it will help you to see if the application is doing something differnt or if there is a difference in the parameter/settings. Having said that, you might re-cache the provided scene (save the provided and your cache), just in case to have both (your set up and Donovan’s) with the R18 algorithm. The changes are not big, but to get it as close as possible for any comparison. Keep in mind that these simulations are very complex each little variation will result in changes, hence another reason to cache solutions.

I have rendered your file in R16 and R18, as well both files that were shared in both releases. I switched the lights off and the Standard render on, to keep it short: there should be not a big difference among all versions.

All the best

 Signature 

Dr. Sassi V. Sassmannshausen Ph.D.

Photography For C4D Artists: 200 Free Tutorials: Texture, Panorama, HDRI, Camera Projection, etc.
https://www.youtube.com/user/DrSassiLA/playlists

Profile
 
 
Posted: 07 August 2017 02:22 AM   [ Ignore ]   [ # 2 ]  
Total Posts:  9
Joined  2013-10-16

Hi, thank you very much for the quick reply.

I adjusted my flexion and bounce settings to match but do not get the same result as the tutorial.

I am pretty sure i have the same settings but get a different result.

I also downloaded Donovans file and with the cache enabled I get the same result as in the tutorial video, however when i disable the cache and re run the sim i get a different result. (see attached screen shots)

I have attached my file again with felxion and bounce revised to zero. when you open my file in R18 do you get a similar sim to the tutorial or was it really stretched at the belt points like my screenshot?

File Attachments
dynamics_cloth_cineversity_tut_04_02.zip  (File Size: 1353KB - Downloads: 4)
Profile
 
 
Posted: 07 August 2017 02:26 AM   [ Ignore ]   [ # 3 ]  
Total Posts:  9
Joined  2013-10-16

sorry i think my screenshots didn’t upload as there is a 800x600 max resolution. here they are.

Image Attachments
donovan_without_cache.jpgdonovan_with_cache.jpg
Profile
 
 
Posted: 07 August 2017 10:30 AM   [ Ignore ]   [ # 4 ]  
Moderator
Avatar
Total Posts:  4428
Joined  2011-03-03

Hi picturedrift,

Thanks a lot for the image, that made it clear.

Yes, I can see that the initial dropping is quite different. I placed the “Influence” animation on it (Belts), so that wasn’t obvious anymore to me. The only other difference I was able to find was the frame rate, but that should not be of any concerns.

Here is the rendering in comparison that I did with Donovan’s v02 file.
I have placed the two preview renderings side by side and then on top of each other. This was the difference I wrote about above.
Clip:
https://www.amazon.com/clouddrive/share/u2UHEUaPFx7FtMQ9kXbnA7pexFfGdu28NmrZ3t0uZhf
This was the difference I have found. I think in the overlay, that was pretty clear, but not really significant.

The initial hanging/stretching R16/R18 is quite different, and I have no indication that any parameter is different.

To just change something seems the most simple suggestion at the moment, which means for example, to change the information available on the belt. I have painted some vertex maps. (In add mode with 5% strength. so each stroke adds a little bit, to keep it not too obvious, like a round gradient).

Scene file
https://www.amazon.com/clouddrive/share/xcpeLeWWVRtg4Q6kMvzLImVm8ksTf3Opv1r2kW6Efd6
I took the Influence animation out for this file, so we have a clear view to the problem…
You certainly noticed that I have some parameters changed, Flexion and Rubber, you might dial it down to your liking.

The problem that we see between the two versions is perhaps related to changes of the applications code. Simulations are complex, and to improve the solving time, the developer team has always new ideas, otherwise we would stuck with a slow app. With those simulations, where one decision influences the next (e.g., polygons) in space and time, the outcome might vary naturally a little bit.

I hope that works for you.

 Signature 

Dr. Sassi V. Sassmannshausen Ph.D.

Photography For C4D Artists: 200 Free Tutorials: Texture, Panorama, HDRI, Camera Projection, etc.
https://www.youtube.com/user/DrSassiLA/playlists

Profile
 
 
Posted: 07 August 2017 03:44 PM   [ Ignore ]   [ # 5 ]  
Total Posts:  9
Joined  2013-10-16

Thanks again for the reply.
Looking at your file, you painted vertex weights and assigned each weight to the corresponding belt tag. You also increased the iterations form 50 to 153.

I tried both of these processes and still do not get the correct result.
I also tried copying your vertex tags to my cloth geo but also didn’t get the correct result.

When i open your file and play the sim it looks right though.

attached 2 scene files
dynamics_cloth_cineversity_tut_04_03 is my original file with vertex weights painted

dynamics_cloth_cineversity_tut_04_04 is my original file with your vertex weight tags copied to my cloth geo

also screenshots attached

Image Attachments
vertex_painting.jpg
File Attachments
dynamics_cloth_cineversity_tut_04_03.zip  (File Size: 1374KB - Downloads: 5)
Profile
 
 
Posted: 07 August 2017 03:47 PM   [ Ignore ]   [ # 6 ]  
Total Posts:  9
Joined  2013-10-16

I could not upload the second file due to a mime type error
“Error Message:  The file you are attempting to upload has invalid content for its MIME type.”

so here is a link to it via dropbox

https://www.dropbox.com/s/pow8ek4gfh6xfgl/dynamics_cloth_cineversity_tut_04_04.zip?dl=0

Profile
 
 
Posted: 07 August 2017 04:57 PM   [ Ignore ]   [ # 7 ]  
Moderator
Avatar
Total Posts:  4428
Joined  2011-03-03

Thanks for the file, picturedrift. Very nice of you.

I had hopes that the “MIME” thingy was an annoyance of the past, obviously not. Sorry about that.

The iterations are not relevant for now. But check the amount of points I had in each belt. To have them not like a massive element, I added the Weight Tag as well, so they fade more, at least that is my intention to hide them that way, it is a mix of forces.

When you explore Donovan’s _v02 file, you will see that he added also some more points, and a weight map like a frame. Since we are two years further in application iterations I tried to adopt to the changes: much more points, perhaps too many.

The file I had shared was your file, not something I had set up newly. In the same way is the file below, just taken from your last post. I have attached the Cubes each times a little bit differently, so please check which “flow” of the fabric you like the most and apply it as you like.

Scene file
https://www.amazon.com/clouddrive/share/BOG3C7K2jMWSDFEj8OWl0HUGZFcXYQPCHO2lVDeninf

I hope all these extra steps are helping at the same time to explore this area a little bit more, fingers crossed you can see it in that way.

All the best

 Signature 

Dr. Sassi V. Sassmannshausen Ph.D.

Photography For C4D Artists: 200 Free Tutorials: Texture, Panorama, HDRI, Camera Projection, etc.
https://www.youtube.com/user/DrSassiLA/playlists

Profile
 
 
Posted: 07 August 2017 05:47 PM   [ Ignore ]   [ # 8 ]  
Total Posts:  9
Joined  2013-10-16

ah ok, I understand now. you need to assign more points on the mesh to each belt tag.

at 5:33 Donovan goes through this stage but only selects one point per belt tag.

So using R18 you need many more points associated to the belt and then paint vertex weights to get a similar result that only 1 point and no vertex weights would have given you in R16. That seems like a pretty big change to the way the cloth works, and a change that seems to make it more difficult to use….

but at least it makes sense now

thanks again for your patience and help

Profile
 
 
Posted: 07 August 2017 06:49 PM   [ Ignore ]   [ # 9 ]  
Moderator
Avatar
Total Posts:  4428
Joined  2011-03-03

Yes, picturedrift, he uses one point in his initial file and several points each time in his _v2 file.

You will find a lot of tutorials, if my memory serves me well, from the last decade+, where the instructor uses often even the whole “side” of a plane, to belt it. Such amount is and will be used, in fact the belt was initially created for such amounts.
I have some problems to see that as a big change, but everyone has of course his/jher own perspective to it, I agree. The more detailed belting seems to offer a better precision than before. But I guess we should stop here, I can’t add more information to this problem at the moment, and perhaps any extra time would be better invested in suggestion, as mentioned below.

If you find that this should be different, MAXON is certainly listening, and has even channel to receive such wishes:
https://www.maxon.net/en-us/support/suggestions/
Pelase describe exactly what you need and how it can improve your work. The more people write the more an idea will be implemented, or an existing feature will be fine-tuned. Thanks for your time! Very much appreciated. Thanks.

My best wises

 Signature 

Dr. Sassi V. Sassmannshausen Ph.D.

Photography For C4D Artists: 200 Free Tutorials: Texture, Panorama, HDRI, Camera Projection, etc.
https://www.youtube.com/user/DrSassiLA/playlists

Profile
 
 
Posted: 08 August 2017 04:41 AM   [ Ignore ]   [ # 10 ]  
Total Posts:  9
Joined  2013-10-16

Thanks again for your help.

Profile
 
 
Posted: 08 August 2017 10:35 AM   [ Ignore ]   [ # 11 ]  
Moderator
Avatar
Total Posts:  4428
Joined  2011-03-03

You’re welcome, picturedrift.

 Signature 

Dr. Sassi V. Sassmannshausen Ph.D.

Photography For C4D Artists: 200 Free Tutorials: Texture, Panorama, HDRI, Camera Projection, etc.
https://www.youtube.com/user/DrSassiLA/playlists

Profile