Re: Simulating very stiff suspension-system correctly ?

Chris Hecker has a ton of great kinematics and dynamics resources on his web site: http://www.d6.com/users/checker/dynamics.htm You may need to do some numerical analysis or switch to a multistep integrator... "Gregor Veble" <[email hidden]> wrote in message news:[email hidden]... > Paal K Holmberg wrote: > > > > I'm working on a hobby-program to simulate a car, but I am havingproblems > > getting the suspensionsystem to stablize itself. Does anyone have anylinks

Quoted text (17 lines)

> > to resources on simulating very stiff suspensionsystems with discrete > > timesteps ? > > Thanks in advance. > > > > P�l K Holmberg > > I don't have any links, but I might be able to help. What parts of > suspension do you model? It is a bad idea to simulate the mass of the > wheels with very stiff springs, for example, as the oscilation period of > the wheels would be shorter than the timestep available during the > simulation with very stiff suspension and the effects of wheel mass > would not even be noticable in the simulation. > > Tell more about exactly how you model the suspension and the car and I > might be able to help out. > > -Gregor

Hey, Paal, I'm doing the same thing! In stabalizing the suspension, I'm assuming you ran into the same problem I did? You get the spring stuff in the prog, and the car bounces around, or rolls back and forth erratically and super quickly with high spring rates when you enter a corner? If so, you need to model the dampers as well (gotta do it!). How? Easy. (I hope I'm right on your problem. Otherwise, forgive me for an awful, condecending reply here!!) Dampers are velocity sensitive. So the force you calculate for them has nothing to do with the suspension components' compression. If the velocity is 1 foot per second over your time step, you apply 100 pounds of force up on the chassis and down on the tire (or whatever your shock/damper compression/rebound rate is.) My prog works great as it is now, even with wickedly high spring rates, as long as the damper rates are high as well. The car is always stable with high enough damper rates. Also, I had a problem with the inside front/rear wheel leaving the ground if that particular end was sprung too stiffly and there was a lot of grip with a high center of gravity, so if you're already doing the damper stuff, you may want to check and see if that may be happening. If I'm totally WAY off on your problem here and not understanding your question correctly (I do that sometimes), let me know! Would love to chat with you about this. Todd Wasson ---Drag Racing Performance Prediction Software http://PerformanceSimulations.Com

Paal K Holmberg wrote:

Quoted text (8 lines)

> > I'm working on a hobby-program to simulate a car, but I am having problems > getting the suspensionsystem to stablize itself. Does anyone have any links > to resources on simulating very stiff suspensionsystems with discrete > timesteps ? > Thanks in advance. > > P�l K Holmberg

I don't have any links, but I might be able to help. What parts of suspension do you model? It is a bad idea to simulate the mass of the wheels with very stiff springs, for example, as the oscilation period of the wheels would be shorter than the timestep available during the simulation with very stiff suspension and the effects of wheel mass would not even be noticable in the simulation. Tell more about exactly how you model the suspension and the car and I might be able to help out. -Gregor

I'm working on a hobby-program to simulate a car, but I am having problems getting the suspensionsystem to stablize itself. Does anyone have any links to resources on simulating very stiff suspensionsystems with discrete timesteps ? Thanks in advance. P�l K Holmberg

I dunno if you have already got this or if it is exactly what you want, but it seems pretty good. I would love to start writing a game engine (particularly a RACING game engine), but I am stuck in my previous project (race car timing).....and I really want to make it into a proper system. http://autopedia.com/stuttgart-west/StuttPhysics.html Ashley http://www.race-timing.com Paal K Holmberg <[email hidden]> wrote in message news:nCq%4.6390$[email hidden]... | I'm working on a hobby-program to simulate a car, but I am having problems | getting the suspensionsystem to stablize itself. Does anyone have anylinks

Quoted text (8 lines)

| to resources on simulating very stiff suspensionsystems with discrete | timesteps ? | Thanks in advance. | | P�l K Holmberg | | |

Gregor Veble <[email hidden]> wrote in message news:[email hidden]... > Paal K Holmberg wrote: > > > > I'm working on a hobby-program to simulate a car, but I am havingproblems > > getting the suspensionsystem to stablize itself. Does anyone have anylinks

Quoted text (17 lines)

> > to resources on simulating very stiff suspensionsystems with discrete > > timesteps ? > > Thanks in advance. > > > > P�l K Holmberg > > I don't have any links, but I might be able to help. What parts of > suspension do you model? It is a bad idea to simulate the mass of the > wheels with very stiff springs, for example, as the oscilation period of > the wheels would be shorter than the timestep available during the > simulation with very stiff suspension and the effects of wheel mass > would not even be noticable in the simulation. > > Tell more about exactly how you model the suspension and the car and I > might be able to help out. > > -Gregor

Thanks for the reply. At the moment, I am just modelling a simple suspensionsystem where the tyres moves straight up and down, with SuspensionForce = MinSuspensionForce + CompressionFraction * (MaxSuspensionForce - MinSuspensionForce). (If suspension is fully uncompressed (ie. CompressionFraction == 0.0), the SuspensionForce is set to 0.0). I have got decent results but still with some small vibrations by using the following approach: 1. Add gravity and Suspensionforces 2. Update state with current timestep 3. Add damping forces based on new angular-velocities and velocites at suspensionconnection-points.(if in contact with ground) 4. Update state with current timestep (damping forces only) 5. Update position and orientation for timestep PKH

You might play with what is sometimes called a "Quarter Car" model: just one suspension spring and damper suspending one quarter of the cars "sprung" mass. You will get a lot of the bugs worked out of the numerical basics that way without all the headaches of working with more of the car. If you review some basic texts on vehicle dynamics like the two below, you'll find out nifty things like normal ranges for the characteristics of the suspension springs, and potential problems like settings that can give pitch modes at higher frequencies than you might expect. You can also add a spring to model the tire, once you feel more adventurous. ;) The texts will give you examples and formulas to check your sim work. That is, your car model with particular springs and dampers should have certain resulting pitch and heave frequencies when you stimulate it, and the texts will tell you what should happen. One important point others have mentioned is the stepsize you chose for your numerical integration technique. For simplistic integrators you might try without much study, you can use the rule of thumb of integrating 100 times faster than the frequencies of the dynamics you are simulating. (One hundred, that is not a typo ;) For example, if you have a sporty car with a 3 cycles per second suspension, you should integrate at a step size of no more than 0.00333 sec. (300 cycles per second.) You can push it down quite a bit from that rule-of-thumb of a factor of 100, but you will need your quarter car model to make sure it is still stable. Fancier integration routines will both be more accurate and stabler at lower integration frequencies, but may require significantly more computational effort. (If you are concentrating on pushing the frequency down to the limit.) Depending on how complicated your model is, one method might match your exact problem better than another, even if it isn't the trendiest algorithm. For a real time application, you need to be concerned with both accuracy and frequency response. Most numerical analysis texts concentrate on accuracy, and don't discuss the other sufficiently, in my opinion. Have fun, - Matt A couple of good text books: The basics: �Fundamentals of Vehicle Dynamics,� Thomas Gillespie, SAE, Danvers, MA, 1992. ISBN 1-56091-199-9 You will find all the simple formulas you need to estimate the effects of various springs and damper characteristics. Then you can test them in your simulation ;) Larger: "Race Car Vehicle Dynamics," W.F. Milliken and D.L. Milliken, SAE, 1995. ISBN-1-56091-526-9 Both are published by the US Society of Automotive Engineers http:/www.sae.org They have an excellent online bookstore Paal K Holmberg wrote:

Quoted text (9 lines)

> > Paal K Holmberg wrote: > > > > > > I'm working on a hobby-program to simulate a car, but I am having > problems > > > getting the suspensionsystem to stablize itself. Does anyone have any > links > > > to resources on simulating very stiff suspensionsystems with discrete > > > timesteps ? > > > Thanks in advance.