Car Physics: Roll centers question

sorry to keep you people on this list from more important stuff - each question i ask raises five new questions ;-) i kept hearing about roll centers a lot recently, and did some research throughout the net, but i don't quite grasp it. at the moment i calculate the torque vector a force applies on the car body as torque=(attackPoint-CG) X force . attackPoint is the point where the force is applied (usually road contact point of a wheels). X is just a cross-product. now it is dawning me that this is not correct, and that i should calculate torque seperatly for the x,y and z axis. instead of using the CG for calculating roll (z in my coord system) torque i should use a roll center. is this correct? but if this is so wouldn't there also have to be a pitch center and yaw center? why does one never hear about those. and also where does the CG come into the results, if i use roll centers instead of CG for torque calculation? or have i misunderstood it all? TIA jonas hmm.. has anyone ever thought about moving these discussions to a rec.autos.simulators.physics ng?

jonas echterhoff wrote:

Quoted text (8 lines)

> sorry to keep you people on this list from more important stuff - each > question i ask raises five new questions ;-) > > i kept hearing about roll centers a lot recently, and did some research > throughout the net, but i don't quite grasp it. at the moment i > calculate the torque vector a force applies on the car body as > > torque=(attackPoint-CG) X force .

If attackPoint , CG and force are in world coordinate this will give the angular torque in car coordinate. Then from the car orientation, the car angular momentum, the car local inertia tensor you can calculate the wolrd inertia tensor and the new car orientation. Imagine a longitudinal force. If you aplly it BELOW the CG you have the car dive correctly but if you apply it above the car will dive like you brake instead of when like you accelerate ! -- Sebastien TIXIER - Game Developer Dynamics and Car Physics http://www.eden-studios.fr GPLRank Normal:-42.98 Monster:-117.44

jonas echterhoff <[email hidden]> wrote in message news:060920011454181283%[email hidden]... > seperatly for the x,y and z axis. instead of using the CG for > calculating roll (z in my coord system) torque i should use a roll > center. is this correct? I believe a reasonable way to do it is to treat the sprung and unsprung masses separately. For the sprung mass, lateral forces generate a couple between the RC and the CG; for the unsprung mass, it's between the RC and the contact patch (ground level). In the steady state this only really affects sprung mass roll angle, but in transients an important difference is that some force is 'absorbed' in rolling the body and thus doesn't all appear at the tyres immediately. The other thing to note is that not all the weight transfer appears at the springs - some goes directly through the suspension linkage. > but if this is so wouldn't there also have to be a pitch center andyaw > center? why does one never hear about those. and also where does theCG The pitch centre does exist, although for horizontal / parallel wishbone hinges it's at ground level. It becomes more relevant with anti-squat/dive suspension geometry. Yaw centre I'm not sure about - I /think/ that for one to exist, the wheels would have to move significantly in the horizontal plane (relative to the chassis) during the application of suspension loads. > come into the results, if i use roll centers instead of CG fortorque > calculation? or have i misunderstood it all? See above... HTH. Jonny

In article <[email hidden]>, Sebastien Tixier <[email hidden]> wrote:

Quoted text (21 lines)

> jonas echterhoff wrote: > > > sorry to keep you people on this list from more important stuff - each > > question i ask raises five new questions ;-) > > > > i kept hearing about roll centers a lot recently, and did some research > > throughout the net, but i don't quite grasp it. at the moment i > > calculate the torque vector a force applies on the car body as > > > > torque=(attackPoint-CG) X force . > > If attackPoint , CG and force are in world coordinate > this will give the angular torque in car coordinate. > > Then from the car orientation, the car angular momentum, the car local > inertia tensor you > can calculate the wolrd inertia tensor and the new car orientation. > > Imagine a longitudinal force. If you aplly it BELOW the CG you have the car > dive correctly but if you apply it above the car will dive like you brake > instead of when like you accelerate !

sure - all this works for me as you described. the question was where do roll centers come into the equation. jonas

One quick trick I've used is to find the normal to the plane that contains the instant axis and the contact point, and use that normal to split the force vector into body and wheel components (If I recall correctly, the math is something like f_n = f dot n, f_wheel = n * f_n, f_body = f - f_wheel ) this can give you anti-dive and other jacking effects for independent suspensions... live-axle rears are more of a pain. you can start out by just pre-computing a fixed normal in body space, and later on start perturbing it with suspension travel. The rational behind this technique is that torques about the instant axis will only be reacted to by the springs, shocks, and ARBs. I tend to think of roll centers as being a secondary concept like engine horsepower, used for analysis, and instant axis as being something you simulate directly like engine torque. Unless, of course, you simulate the links directly, in which case roll centers are tertiary and instant axis is secondary (Note: this is all handwaving). I think every simulator seems to use a different approach here, many of which are equivalent--it depends on whether you model wheel hope and/or gyroscopic effects... -Dave P. MGI "jonas echterhoff" <[email hidden]> wrote in message news:060920011454181283%[email hidden]...

Quoted text (28 lines)

> sorry to keep you people on this list from more important stuff - each > question i ask raises five new questions ;-) > > i kept hearing about roll centers a lot recently, and did some research > throughout the net, but i don't quite grasp it. at the moment i > calculate the torque vector a force applies on the car body as > > torque=(attackPoint-CG) X force . > > attackPoint is the point where the force is applied (usually road > contact point of a wheels). X is just a cross-product. now it is > dawning me that this is not correct, and that i should calculate torque > seperatly for the x,y and z axis. instead of using the CG for > calculating roll (z in my coord system) torque i should use a roll > center. is this correct? > > but if this is so wouldn't there also have to be a pitch center and yaw > center? why does one never hear about those. and also where does the CG > come into the results, if i use roll centers instead of CG for torque > calculation? or have i misunderstood it all? > > TIA > > jonas > > > hmm.. has anyone ever thought about moving these discussions to a > rec.autos.simulators.physics ng?

hooo... sorry, i didn't understand correctly the question... well, i still don't understand your question , my english is not good enough :o( regards, -- Sebastien TIXIER - Game Developer Dynamics and Car Physics http://www.eden-studios.fr GPLRank Normal:-42.98 Monster:-117.44

Wow, Dave, I was about to write something of the kind myself, but you've worded it so nicely that I'll just add a 'what he said' :). I guess that separates you from us mere mortals... ;). -Gregor Dave Pollatsek wrote:

Quoted text (7 lines)

> > I tend to think of roll centers as being a secondary concept like engine > horsepower, used for analysis, and instant axis as being something you > simulate directly like engine torque. Unless, of course, you simulate the > links directly, in which case roll centers are tertiary and instant axis is > secondary (Note: this is all handwaving). >

Simply, Lateral force is applied through the CG. However, the car rolls about the roll centers - or roll axis . There is a torque arm that forms from the distance between the two (the CG always being above the RC). This torque arm has to be figured into your roll resistence formulas. This is why changing a trackbar height on the back of a NASCAR will change the over/understeer characteristics of the vehicle. It changes that ends resistence to roll buy changing the rear RC height, which in turn changes the "torque arm" thats resisting the roll - making that ends lateral force more or less effective on that axles spring package. As far as the formulas to do all that?............no clue :) -- Doug Arnao www.vehiclecraft.com "jonas echterhoff" <[email hidden]> wrote in message news:060920011454181283%[email hidden]... > sorry to keep you people on this list from more important stuff - each > question i ask raises five new questions ;-) > > i kept hearing about roll centers a lot recently, and did some research > throughout the net, but i don't quite grasp it. at the moment i > calculate the torque vector a force applies on the car body as > > torque=(attackPoint-CG) X force . > > attackPoint is the point where the force is applied (usually road > contact point of a wheels). X is just a cross-product. now it is > dawning me that this is not correct, and that i should calculate torque > seperatly for the x,y and z axis. instead of using the CG for > calculating roll (z in my coord system) torque i should use a roll > center. is this correct? > > but if this is so wouldn't there also have to be a pitch center and yaw > center? why does one never hear about those. and also where does the CG > come into the results, if i use roll centers instead of CG for torque > calculation? or have i misunderstood it all? > > TIA > > jonas > > > hmm.. has anyone ever thought about moving these discussions to a > rec.autos.simulators.physics ng?

As Dave already mentioned, while the way you describe it is very good in terms of understanding the car balance, it's not the best way to handle car physics in calculations. It's best to use instant axes for wheel travel and apply forces and torques per Physics 101, and the effects you mention should be taken care of automatticaly. There's usually a difference between how our mind percieves physics and what's the most natural way to describe the phenomena. On the other hand, describing phenomena in the way your mind isn't used to thinking in has the effect of adjusting the way you see the problem, which is half, wait, whom am I kidding, all the fun there is to physics ;). -Gregor Doug Arnao wrote:

Quoted text (17 lines)

> > Simply, Lateral force is applied through the CG. However, the car rolls > about the roll centers - or roll axis . There is a torque arm that forms > from the distance between the two (the CG always being above the RC). This > torque arm has to be figured into your roll resistence formulas. This is why > changing a trackbar height on the back of a NASCAR will change the > over/understeer characteristics of the vehicle. It changes that ends > resistence to roll buy changing the rear RC height, which in turn changes > the "torque arm" thats resisting the roll - making that ends lateral force > more or less effective on that axles spring package. > > As far as the formulas to do all that?............no clue :) > > -- > Doug Arnao > www.vehiclecraft.com >

From sebastians message, I got the impression he wasn't sure how the RC and CG interacted to make roll etc. I was just describing that process.....understanding the "phenomenon" I would think makes the coding process easier and more accurate? As far as the coding or claculations........again I have little clue <g>. -- Doug Arnao www.vehiclecraft.com 1-800-845-5948 "Gregor Veble" <[email hidden]> wrote in message news:[email hidden]... > As Dave already mentioned, while the way you describe it is very good in > terms of understanding the car balance, it's not the best way to handle > car physics in calculations. It's best to use instant axes for wheel > travel and apply forces and torques per Physics 101, and the effects you > mention should be taken care of automatticaly. > > There's usually a difference between how our mind percieves physics and > what's the most natural way to describe the phenomena. On the other > hand, describing phenomena in the way your mind isn't used to thinking > in has the effect of adjusting the way you see the problem, which is > half, wait, whom am I kidding, all the fun there is to physics ;). > > -Gregor >

Hi Doug, too much knowledge may in fact be detrimental <g>. The fact is, the basic system isn't all that difficult to implement into equations to be solved, but to analyze and understand the special solutions (e.g. steady state) of such a system, concepts like roll center etc. are very helpful and they follow from these basics, not the other way around. -Gregor Doug Arnao wrote:

Quoted text (8 lines)

> > From sebastians message, I got the impression he wasn't sure how the RC and > CG interacted to make roll etc. I was just describing that > process.....understanding the "phenomenon" I would think makes the coding > process easier and more accurate? > > As far as the coding or claculations........again I have little clue <g>. >

Well, my first problem is to translate r.a.s posts into french ;o) I "think" i understand that rolling center and cg are not the same. the simpliest exemple is a cube. if a cube is just touching the ground by 1 edge , this edge is the rolling axle of the cube. If the cube is touching by a corner, this corner is the rolling center. Meanwhile the CG of the cube has never changed ... the more the rolling center is far from the CG the easiest it will be to rotate the cube. the more the force is far from the RC the more torque you'll have. If what i'm saying isn't true then i REALLY didnt' understand what rolling center is ;o))) so just enlight me please ! regards, -- Sebastien TIXIER - Game Developer Dynamics and Car Physics http://www.eden-studios.fr GPLRank Normal:-42.98 Monster:-117.44 Doug Arnao wrote: > From sebastians message, I got the impression he wasn't sure how the RC and > CG interacted to make roll etc. I was just describing that > process.....understanding the "phenomenon" I would think makes the coding > process easier and more accurate? > > As far as the coding or claculations........again I have little clue <g>. > > -- > Doug Arnao > www.vehiclecraft.com > 1-800-845-5948 > "Gregor Veble" <[email hidden]> wrote in message > news:[email hidden]... > > As Dave already mentioned, while the way you describe it is very good in > > terms of understanding the car balance, it's not the best way to handle > > car physics in calculations. It's best to use instant axes for wheel > > travel and apply forces and torques per Physics 101, and the effects you > > mention should be taken care of automatticaly. > > > > There's usually a difference between how our mind percieves physics and > > what's the most natural way to describe the phenomena. On the other > > hand, describing phenomena in the way your mind isn't used to thinking > > in has the effect of adjusting the way you see the problem, which is > > half, wait, whom am I kidding, all the fun there is to physics ;). > > > > -Gregor > >

On Mon, 10 Sep 2001 15:47:37 +0200, in rec.autos.simulators Sebastien Tixier <[email hidden]> wrote:

Quoted text (10 lines)

>Well, my first problem is to translate r.a.s posts into french ;o) >I "think" i understand that rolling center and cg are not the same. the >simpliest exemple is a cube. if a cube is just touching the ground by 1 edge , >this edge is the rolling axle of the cube. If the cube is touching by a corner, >this corner is the rolling center. Meanwhile the CG of the cube has never >changed ... the more the rolling center is far from the CG the easiest it will >be to rotate the cube. the more the force is far from the RC the more torque >you'll have. > >If what i'm saying isn't true [...]

No problem, that's correct :-) Eric