Re: Suspension correction

----- Bernhard Wymann wrote:
> Hi David
> 
> > A real car has the springs preloaded (lest you have loose suspension).
> 
> Sure, but this just shifts the point of eqilibrium, the spring is still 
> a spring (at least for the linear springs we model).
> 
> > I've been following the commits to TORCS and heard about
> > forces due to inertia and mass acting on the axle.
> > If TORCS' formulas are calculated in a proper and generalized
> > way, the spring may have tensile forces (negative values) at times.
> 
> The "if" statement which your patch removes does exactly deal with that. 
> The spring can just generate forces in one direction, if it is fully 
> extended it is just loose. Depending on the ride height/suspension 
> course setup this does not happen, but if it happens we are prepared.
> 
Good.

> > This patch does the bounds checking for negative values where
> > the tire contacts the road, and not where damping may act
> > against the spring.
> >
> > The car is more controllable on dirt tracks, and has no new
> 
> This is no argument, we could just replace it all with a constant force 
> and handling would be great, but the model would be dead wrong anyway.
> 
> The second part of your change is identical (semantically) to the 
> previous code, think about it again:
> 
> 1.3.8-test1:
> 	tdble internalForce = springForce(susp) + damperForce(susp);
> 	if (internalForce <= 0.0f) {
> 		susp->force = 0.0f;
> 	} else {
> 		susp->force = internalForce * susp->spring.bellcrank;
> 	}
> 
> Yours:
> susp->force = (springForce(susp) + damperForce(susp)) * 
> susp->spring.bellcrank;
> 	if (susp->force < 0.0f)
> 		susp->force = 0.0f;
> 
> So whatever you have felt must be something not contained in this patch. 
> Agreed?
It appears that keeping negative spring values in the first part
of this patch is what I felt.

> 
> > problems that I could find. Thus, I figured this is an improvement.
> >
> > On a real car, the spring may be preloaded so that it never gets
> > a tensile force on it (no need for negative bounds checking).
> 
> Agreed, but if somebody gets the parameters of the cars wrong I do not 
> want the simulation to "explode", otherwise you have to implement 
> validations in some other places.
> 
> Kind regards
> 
> Bernhard

Thank you for the explanations.

I just conducted an experiment on Dirt-2 and car1-trb1 with
a printf() added in torcs/src/modules/simu/simuv2/susp.cpp

/*
 * get spring force
 */
static tdble springForce(tSuspension *susp)
{
	tSpring *spring = &(susp->spring);
	tdble f;
	
	/* K is < 0 */
	f = spring->K * (susp->x - spring->x0) + spring->F0;
	if (f < 0.0f) {
/***/		printf("%f",f);
		// Compression spring, so the force can never change the sign.
		f = 0.0f;
	}
	
	return f;
}

Numbers in the range of -7000 are printed when the car is jumped in the air.
Please check to see if this is reasonable.
Does car1-trb1 have loose suspension, or is something else happening?
People that make torcs cars need to know if their setup is what they
intended.

Sincerely,
David Savinkoff

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot