Developing a Proper Suspension Model

[Thorpedo]

Originally Posted by fika84

I doubt a manufacturer is going to give you a recommended camber value. It's going to change depending on what goals you have. When I was working as a vehicle dynamics engineer we would go to Calspan and test the tires we wanted to model on a tire testing machine, fit a Pacejka model and use it in simulations to determine optimum setups per track. Tire manufacturers do this testing in house typically keep it hush hush.

Actually many of them do give a rough value for camber. Especially in competition oriented tires.

[Amaya]

Originally Posted by RBbugBITme

Yeah, tire data is as rare as a unicorn. You literally have to buy tons of sets yourself and run the same track over and over at different settings with shock pots and g-meters to get any real understanding of whats going on. That's on top of having the suspension modeled accurately.

Just a thought, if you had a test vehicle with a rigid "suspension" that allowed you to adjust camber (and maybe other settings)? I'm thinking drive it in circles and track lateral g's reached with different camber settings. I would think that optimum camber would change based on lateral acceleration and having a vehicle where the camber didn't change under lateral g's would yield more precise results.

[fika84]

Originally Posted by Thorpedo

Actually many of them do give a rough value for camber. Especially in competition oriented tires.

Yea youre right, it just depends on the series. IRL actually supplied the tire data. Brazilian stock car didnt allow testing. V8 Supercar made you test your own tires. All different.

[Shankenstein]

@RBbugBITme ,

Thanks for helping out the thread with your mad WinGeo skills. I've added your Caster, Roll Center, and Motion Ratio data to the OP. If we can support your efforts/project, don't hesitate to ask.

I was curious about how you pulled the points for your model. It sounds pretty precise. Is the Toyota 3D model an assembly file or one giant file? What kind of access are you given?

For my current employer, the OEMs are pretty forward with the level of detail in their models (once you sign the NDA)... but it's usually just assemblies of STEP files (dumb solids). We use them to check clearances, mounting locations, air/water paths, etc. I was curious the level of play that Toyota is giving the aftermarket and enthusiast community.

If you can't say, that's OK too. It's great to have another suspension modeling guy on the 86 forums.

[RBbugBITme]

Originally Posted by Amaya

Just a thought, if you had a test vehicle with a rigid "suspension" that allowed you to adjust camber (and maybe other settings)? I'm thinking drive it in circles and track lateral g's reached with different camber settings. I would think that optimum camber would change based on lateral acceleration and having a vehicle where the camber didn't change under lateral g's would yield more precise results.

eeehhhhhhhhhhhhhhhhhhhhhh having to think to much, too many non-linear relationships with tires.

Originally Posted by Shankenstein

@RBbugBITme ,

Thanks for helping out the thread with your mad WinGeo skills. I've added your Caster, Roll Center, and Motion Ratio data to the OP. If we can support your efforts/project, don't hesitate to ask.

I was curious about how you pulled the points for your model. It sounds pretty precise. Is the Toyota 3D model an assembly file or one giant file? What kind of access are you given?

I was provided with separate 3D models of certain suspension and chassis components. All of them had the exact same origin so if I created an assembly myself and mated the origins of all the individual files I would eventually have an entire car model with all sheet metal, suspension, etc. perfectly placed in space. I then place axes and points in the center of each suspension pick up point or in the center of a ball joint. Then create a vehicle centerline by placing a plane at midpoint between any left and right side suspension point. Follow that up with a sketch of an OEM tire cross section to create an realistic ground plane and then measure fore/aft distances from the front hub centerline because that is how WinGeo requires the measurements in that direction (negative is forward of front axle centerline).

I'm also provided with assemblies of just the suspension in 3 different positions. I wasn't quite sure what this was but I found it the middle position was ride height and the other two were *designed* full droop/compression.

[RBbugBITme]

Alright, here are the camber curves and FAP's at two different ride heights.

What I've done is select a static steer angle to account for the effect of caster, and then iterated chassis roll angle from -3 degrees to 0 degrees, and ride height from -1.5" to 0" in the first one and -2.5" to 1.5" in the second image.

In other words, this is a right hand turn with variable chassis roll and variable ride height with the same steer angle which would mean we're looking at the same steering wheel input at varying track speeds (chassis roll angle varies with speed/G's acting on your FAP's/RC/moment arm).

The solid black vertical line in both images is at -1.5 degrees of chassis roll at -1" and -2" ride height. CamberL is the outside front tire which is the most heavily loaded tire in this scenario. Each sawtooth section is the change in ride height by .100" and ride height is increasing from left to right.


Left most sawtooth section is -1.5" ride height with roll angle varying from -3 deg to 0 deg. Right most saw tooth is OEM ride height.


Same thing as above but the left most sawtooth is at -2.5" ride height with roll angle varying from -3 deg to 0 deg. Right most sawtooth is -1.5" ride height.

[ZDan]

The problem I have is with the RC height curve shown here (with context):

Originally Posted by RBbugBITme

This one is particularly nasty. This is a plot of -3 to 3 deg of roll at a ride height of -3 on the far left to +2 on the far right in 0.10" steps. The vertical black line is at -1" ride height which is right around where the roll center goes crazy shooting off to infinity and switching back and forth above and below the ground plane. I wouldn't personally ever set a BRZ to a 1" drop after looking at this without roll center correction.

The roll center height plot is saying that:
1. At -1.05" ride height, roll center starts at zero (ground level) at zero roll, and rapidly goes CRAZY HIGH (way above ground) with any roll in either direction.
2. At -0.95" ride height, roll center starts at zero (ground level) at zero roll and rapidly goes CRAZY NEGATIVE (way below ground) with any roll in either direction.

Even with perfectly rigid spherical bearings all around, no way that's what happens. Such a car would exhibit zero roll in cornering at -1.05" ride height, but would immediately flop all the way to the bump stops in corners at -0.95" ride height.

I still strongly believe the math model is breaking down with roll applied when static RC height is at ground level and giving unrealistic RC height change with roll.

The roll center height calculation is based on an intersection point of a line that is tending out towards infinity. In that region, you get to a point where you can't calculate RC height with enough precision to give reasonable results.

You can run the numbers and get results, but when basing RC height which is on the order of inches on a line that's going out billions of miles and beyond, the calculated RC height value becomes kinda meaningless...

I would bet that a physics-based vehicle simulation model would give different results, and I'm *certain* that an actual FR-S/BRZ won't behave like it had infinite roll stiffness at -1.05" ride height and zero roll stiffness at -0.95" ride height even with spherical bearings all around.

[fika84]

Originally Posted by ZDan

The problem I have is with the RC height curve shown here (with context):



The roll center height plot is saying that:
1. At -1.05" ride height, roll center starts at zero (ground level) at zero roll, and rapidly goes CRAZY HIGH (way above ground) with any roll in either direction.
2. At -0.95" ride height, roll center starts at zero (ground level) at zero roll and rapidly goes CRAZY NEGATIVE (way below ground) with any roll in either direction.

Even with perfectly rigid spherical bearings all around, no way that's what happens. Such a car would exhibit zero roll in cornering at -1.05" ride height, but would immediately flop all the way to the bump stops in corners at -0.95" ride height.

I still strongly believe the math model is breaking down with roll applied when static RC height is at ground level and giving unrealistic RC height change with roll.

The roll center height calculation is based on an intersection point of a line that is tending out towards infinity. In that region, you get to a point where you can't calculate RC height with enough precision to give reasonable results.

You can run the numbers and get results, but when basing RC height which is on the order of inches on a line that's going out billions of miles and beyond, the calculated RC height value becomes kinda meaningless...

I would bet that a physics-based vehicle simulation model would give different results, and I'm *certain* that an actual FR-S/BRZ won't behave like it had infinite roll stiffness at -1.05" ride height and zero roll stiffness at -0.95" ride height even with spherical bearings all around.

I second this motion . These are limitations to the geometric method. That's why I would like to see a force based simulation, or track data with wheel force transducers (at $50k/wheel I don't think anyone has this...) so that I could calculate the force based RC myself. Or as @Amaya was hinting at (even though he didn't know it..) was skid pad data. They don't put rigid suspension in for this since they want the car to behave as it's supposed to.. the hardest part about skid pad testing is getting a driver to be able to maintain constant lateral g (and you need lots of sensors). They DO however put solid suspension rods in when doing K&C testing (which can also give you force based roll centers!), which is an entirely other animal used to test the compliance in a vehicle (how much does it move when forced through metal bending and bushings compressing).

The geometric RC has some merit (it's better than nothing and is typically all that is used for FSAE teams to design a little racecar) and is much easier to attain, but it doesn't tell the entire story and it's important to understand where it does and doesn't work (like all simulation analysis).

[RBbugBITme]

ZDan, I understand any RC related equations become useless when it is very near the ground plane. That doesn't change the fact that when it (or the FAP if you prefer) passes back and forth above and below the ground plane your jacking forces will be constantly reversing direction. That isn't going to be good for stability at the limit of tire adhesion unless you have minuscule jacking forces to begin with. I don't know what typical jacking forces are for this car.

[ZDan]

Originally Posted by RBbugBITme

ZDan, I understand any RC related equations become useless when it is very near the ground plane.

So why plot them and suggest they indicate "bad things"?

That doesn't change the fact that when it (or the FAP if you prefer) passes back and forth above and below the ground plane your jacking forces will be constantly reversing direction. That isn't going to be good for stability at the limit of tire adhesion unless you have minuscule jacking forces to begin with. I don't know what typical jacking forces are for this car.

Jacking forces are small when roll center height is near zero. Going from RC height +0.05" to -0.05" isn't going to give any bigger or more profound change in jacking force than going from RC height +0.15" to +0.05".

The fact that it goes from slightly anti-roll to slightly pro-roll is not in itself a huge deal. It's not like the forces go crazy near zero.

[RBbugBITme]

Originally Posted by ZDan

So why plot them and suggest they indicate "bad things"?


Jacking forces are small when roll center height is near zero. Going from RC height +0.05" to -0.05" isn't going to give any bigger or more profound change in jacking force than going from RC height +0.15" to +0.05".

The fact that it goes from slightly anti-roll to slightly pro-roll is not in itself a huge deal. It's not like the forces go crazy near zero.

You're right but the roll center for this car isn't that controlled. Depending on your ride height you can see RC heights during roll from -1" to +10" or +1" to -10". That is a big swing worth noting. Its also worth noting that RC movement is fairly well controlled at stock ride height so I would guess a properly designed roll center correction kit would be worth while.

[Wepeel]

Originally Posted by RBbugBITme

These are based on Toyota/Scion 3D models that have been paid for so I can't share the data points.



I have all the FAP data as well and will try to apply some FAP theory but I'm limited on time, I have a side business developing suspension parts for other cars and I was at this until I think 3am last night.

FAP data indeed...

Thanks for posting and doing all of the work. I haven't looked at this thread in a while... time to read up.

[Racecomp Engineering]

First a good link on FAPs (Force Application Points) vs. Kinematic roll centers that might help everyone that's interested in this discussion:

https://drive.google.com/file/d/0B3p...ew?usp=sharing

It's worth reading...and rereading a couple more times.

- Andy

[Racecomp Engineering]

Second, a quote from a discussion I had with Bryan at JRZ USA a while ago:

Originally Posted by Bryan at JRZUSA

Old theories and descriptions of the roll center show that the RC should never pass through the ground. Thats incorrect. Using force based roll centers or what I call, Newtonian physics, the RC passing through the ground doesn't have the "over center" effects of the RC going very far out sideways. It't not necessarily a bad thing, I'd focus more on the mechanical balance of the car and the cambers than exactly where the RC is relative to the ground.

One reason someone would put the RC close to the CG is to reduce the roll moment carried by the springs. The roll moment not carried by the springs is then carried by the suspension links as manifested by the "jacking force." Note that there is no significant reduction in "weight transfer" by raising the RC. You are only balancing between which part the load transfers over. The only major determining factors for load transfer are the CG height and track width. Since the suspension arms are much stiffer than the spring/damper, the response of the car is altered as well. This can manifest in a quicker roll response at the chassis, or degraded dynamic grip since the vertical load might build up faster than the cornering load can. High RCs can also mean more track change with suspension movement and more tire scrub. Again, the scrub affects the response of the tire. There is a limit to effective roll center height.

Note that changing the RC height changes the camber curve as does the caster. All of these effects are interdependent which is what makes K&C machines popular. In my opinion, its more valuable than time at a shaker rig given that you don't have the kinematic data from the car.

- Andy