0.28 Discussion Thread

Hey r3eckon, thanks for spotting this error. I took care of fixing it. The fix will be there for the next update

 

The reason for the deepcopies in wheels.lua is that all the different options are used at the same time while constructing the wheel below, and they all overwrite pretty much the same attributes (beamSpring, etc). Deepcopying them, makes each copy independent from the rest, so that the attributes can be changed without affecting the rest of the copies.

In the addBeamWithOptions case the "b" variable, in which the deepcopy is assigned to, ends up being used to store the beam's attributes. In this case we want all of the different beams to be independent from each other, so we need to use deepcopy.

Thanks for looking and commenting on the code

 

Bug found by Akimbo Beta on Discord. The Covet springs are unrealistically soft. It uses the same suspension geometry as the Honda Civic, and the car is 15% heavier than the Civic, so the springs were made also 15% stiffer. (Covet springs: 31200/19500 F/R N/m. 1996 Honda Civic springs: 28900/14000 F/R N/m.) However Jbeam wise the springs are also around 2 times longer than the Civic ones due to various Jbeam limitations (despite being the same length in the model), and this difference has not been taken into account. In order to behave as they should, the springs have to be roughly 2 times stiffer than they are currently.

Spring length in model vs in Jbeam: 9.5" vs 18.9"


Attached videos of driving on stock springs vs roughly 2x stiffer ones to properly replicate the Civic suspension, you can see a massive difference. Videos recorded by Akimbo Beta who found this bug. Additional proof that it's a thing: closing the hatch causes the whole rear end to bounce, which should not happen if the springs have correct stiffness.

 

Unit of spring stiffness in jbeam is N/m, that is Newtons per meter, so no matter the spring lenght it takes set amount of newtorns to compress the spring a meter.
here's a really quick and dirty demonstration

This would be the initial setup:

Test consistes of a box of nodes that crete rails for springtest2x nodes to slide on freely with no resistance
springtext2x nodes are suspended on a single beam each from springtestx nodes.
Nodes "springtest2x" are 40kg each
Both tests have the same "beamSpring" set at 1000
Once physics gets unpaused this is the result:



As you can see, even with test on the right having over double the initial lenght of the beam, final movement of the nodes in the test is the same, prooving that lenght of the beam does not matter for this.
I hope this all makes sense, I'm awful at describing stuff

 

That makes sense, but then it doesn't explain why the Covet springs are too soft, when the weight increment from the Civic matches up. Maybe it's the angle of the spring beams being drastically different from the angle on the model?

 

in the rear it matches really closely, in the front If I recall correctly jbeam was closer to actual civic geometry, which the mesh doesn't quite follow perfectly, as it would require more work remaster wise, for not much more gain

 

I was under the impression that the model was based on laser scanned Civic suspension. But apparently the Jbeam is closer to it? Now I have no idea what causes it to be so soft, or was it intentionally made this way? If yes then that's a weird choice in my opinion.