TheDunadan Posted June 9, 2004 Share Posted June 9, 2004 Hi, I spent today a fair amount of time with CHOP's and whoa: Amazing ! But well, I guess everyone is already aware of that I don't know who else ever used Matlab / Simulink, but Houdini's CHOP's have quite a few things in common with Simulink. Simulink is a program that can be used for all kind of math simulation and it has as well a node-based graph system and you have various operators that work in a farily similar way to CHOP's in Houdini. Houdini isn't supposed to be a math program, but has anyone ever thought about giving Houdini the possiblity to have something such as a LaPlace CHOP that allows you the user to write functions in the complex variable S-Domain of the Laplace transformation ?! I'd love to see that, since the laplace transformation allows an easy way of describing & working with differential equations. Unfortunally the Apprentice Edition has no HDK, but the required algorithms are rather well known numerical problems and therefore shouldn't be that hard to implement. Would anyone else like to see that in Houdini ?! Jens Quote Link to comment Share on other sites More sharing options...
edward Posted June 10, 2004 Share Posted June 10, 2004 Sure, but I would probably never need it. Something for people who want to dabble with the HDK. The closest thing right now to be more mathematical is that it sports some fourier transform ability I think. Quote Link to comment Share on other sites More sharing options...
TheDunadan Posted June 10, 2004 Author Share Posted June 10, 2004 I don't think it would be that useless really at all. Assume you had a suspension system for some car animation: you could animate the whole thing in a straight forward manner and by specifying the transfer function of your physical system you'd only need to input your animation into your chop network and you'd get a realistic animation out of it. Or for instance you'd want to animate some robot arm: again you could simply animate the arm in the final resting positions and by specifying the transfer function of your "control path" you'd get a fairly realistic animation motion. Quote Link to comment Share on other sites More sharing options...
malexander Posted June 13, 2004 Share Posted June 13, 2004 Having gone through engineering, I honestly never thought that I'd hear about the s-domain ever again once I started into 3D graphics. I gave it a bit of a thought while writing CHOPs, but I figured no user would actuallly ever such a CHOP (guess I was wrong ) . It's useful if you understand the s-domain, but unfortunately most people don't. If you're interested in doing differential equations, I recommend using the Slope and Area CHOPs to determine the instantaneous derivatives and integrals of channels, and combining them in various combinations using Math CHOPs into the differential equations. The Slope CHOP can evaluate the 1st, 2nd and 3rd derivatives, while the Area CHOP can evaluate the 1st, 2nd and 3rd integrals. Quote Link to comment Share on other sites More sharing options...
edward Posted June 13, 2004 Share Posted June 13, 2004 How would such a CHOP work? Looking into it some more, wouldn't you actually do a Z transform CHOP? Quote Link to comment Share on other sites More sharing options...
TheDunadan Posted June 14, 2004 Author Share Posted June 14, 2004 A bit more detailed about what I intended: In control loop engineering we usually study the behaviour of physical systems like this. We try to find a function that describes our physical system well (representing the underlying differential equations) and this transfer function will allow us to simulate and study the behaviour of our physical system. My idea on how we could make good use of this in Houdini was the following: We have some animation like a walk cycle of a character, but unfortunally our mocap actor was pretty much of a feather weight and the animation looks terribly wrong on our oger. No matter how hard our oger would try, he could never move as lightly as our actor or the actor move like an oger. This is due to their physical restrictions and mathematically represented in their transfer function. I think it might be possible to filter this animation with such a transfer function so our walkcycle would get this "oger-feeling" (some closed loop controler is required as well so the motions somehow match, but can be again modulated the same way). Such filtering of mocap data will likely take a fair bit of research, as we have jointed differential equations here, but at least simple tests I made showed enough potenial I'll spent some more time on it. Using the existing Math CHOPs works but it gets tedious really to modulate something such as a PT2-behaviour, that's why I'd prefer some s-domain CHOP. Luckily the animators won't have to see this mess of CHOPs later Has anyone else thought about some similar approach yet ?! Since we're talking about discrete data, it's a z-transformation. But anyhow the idea is still the same How i'd like it to work: The ability to directly put the function in there, e.g. for a integral element I'd have a "1/s". Quote Link to comment Share on other sites More sharing options...
ds604 Posted January 9, 2009 Share Posted January 9, 2009 Hello, I came across this older thread, and now I'm curious, is anyone using Houdini alongside Matlab or Octave? Or, I suppose, R as well. This thread I think is from before the Houdini 9/Python change, so it seems like that might have opened up possibilities. -David Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.