1.9.9b Major Hinge Behavior Problem

[guyboy]

It seems as though the hinge physics keep getting worse with every release... Let me explain. I like to do unconventional things with hinges to test them. For example I created this here scene:

It's made of boxes, each connected by two hinges. This is because I want to pack these boxes together and have them explode:

When I run this on older versions of Algodoo and Phun, it works as expected. The boxes explode back into place.

In the Algodoo Beta 1.9.5, however, because of the new solver, the unfolding becomes extremely slow and uniform motion-y, and Raunit-like effects can be observed, probably due to the stretched hinges. I have constrained this in various situations in a test file downloadable below. When it hits a plane it shows extreme bounciness and detection failures. The test samples hinged to the background showed peculiar displacement when colliding with a plane.

In 1.9.9b, it's quite different. The samples seem to disappear into infinity when they collide with planes. This sometimes happens with moving geoms as well. The RAunit effect is gone, because it was specifically removed. But I suspect this to have been a quick & dirty hotfix and a mere treatment of symptoms.

The slow hinge conflict resolving issue is, IMO, a serious one, as it also affects chains when they are pulled abruptly. And is quite a bit more ugly than the old hinge stretching problem.

I hope there is a way to fix all of these problems so that we have perfectly well-behaved hinges in the future. Maybe speeding up the conflict resolution procedure? Making it more physics-y (dynamic (force-y)) like before? Just some thoughts.

Glitchy Hinge Physics

guyboy

Rating:
Filesize: 23.95 kB
Comments: 0
Ratings: 1

[Someone Else]

Try going int the Script Menu of the hinges and setting legacyMode to 1.

This tells Algodoo to use the old solver. It helps a lot in some cases.

And by RaUnit-like effects, do you mean Lucid-Device-Like effects? The Lucid device was in face eliminated in I think 1.9.8b, the only RaUnit I know of uses spinning air friction to try to point the device in a specified direction.
A bug, reported by Kilinich (of course) informed Emil of a problem with this in I think 1.9.7, which was fixed shortly thereafter.

[guyboy]

No, legacyMode = 0 makes it use the old solver. The raunit was a hinge-based stabilizer in a very early version of this solver. It was eliminated, but the new solver reintroduced it, then it was again eliminated, which introduced extremely bouncy problems caused by extreme forces.

My main point is that the new direct solver comes with an annoying drawback that I call "cramping" or "slow conflict resolution". It happens when you jerk on a chain, for example. You can notice the segments freezing in into place and only loosening slowly.

legacyMode = 1 is the new solver, the one that has the issues I pointed out (cramping).
legacyMode = 0 is the old one, with stretchy hinges.

Which doesn't really make sense??? Legacy mode is supposed to be for backwards compatibility, and legacyMode = 1 makes it sound like it's on, when actually it's off.

[Someone Else]

Yeah, I meant legacyMode=0. I typoed.

With chains (and only chains) this may actually be deliberate, the cramping thing.

LegacyMode actually does make sense.Take springs for example. 0 is Phun springs, 1 is primitive-beta-Algodoo springs, 2 is iterative Algodoo springs, 3 is direct Algodoo springs.

But still. There is some nonsense going on here.

[guyboy]

I think cramping is a nasty side-effect in the new direct solver. Check this out:

Chain Cramping

guyboy

Rating:
Filesize: 40.29 kB
Comments: 1
Ratings: 1

Drag the lower part of the chain. It goes completely crazy.

[Someone Else]

Well, whatever it is, it's not good.

[emilk]

The chain 'cramping' (known in script menu as autoBend) is to prevent explosions due to oscillations higher than the Nyquist frequency. I agree it is not perfect, but perfection is hard to achieve. I hope you still think the new chains are better than the ones in 1.8.5 and that the physics does not in fact get "worse and worse".

As for your scene: there was a bug in earlier betas which turned on this 'cramping' for all hinges, not just the ones in chains. This has since be fixed so if you would have created your scene in 1.9.9 it would have worked fine, or if you just open up the script menu and set autoBend to false for your hinges, it works fine again.

As for legacyMode: think of it as "version" - lower number is older version.