Front Roll Center Migration

[ddd4114]

It's probably a long shot that other people have tried to measure their roll centers and are willing to share, but I'd appreciate any tidbits of knowledge that anybody would like to post.

To provide some background, I have a 2011 GT with basically stock suspension. The only change I've made is camber plates to bring the angle out to a whopping -1 deg. The springs, dampers, bushings, etc. are exactly as they were from the factory. I've been doing HPDE events for a few years, and I've gotten to the point where I'd like to do some suspension work for two reasons. The first (and most obvious) reason is to improve grip and transient response, and the second is to stop shredding the outside edges of my expensive tires and lengthen their lives a bit. I'd like to start doing NASA TT this summer (hopefully in TTB), so I'm also looking for a solution that's cheap in terms of points and also for my wallet.

I could do the usual "replace the springs and shocks and see what happens" approach, but the nerdy engineer in me wanted to have a better understanding of the effects. Therefore, I decided to do some "quick" calculations that have left me scratching my head.

For now, I'm just looking at the car in steady-state cornering. I know it's a very narrow view of the whole picture, but it's a good and simple starting point. I spent a couple hours mapping out my suspension points to approximate my roll centers, and since I've never done this on a car with struts, I wasn't sure what to expect. Obviously my measurements can only be so good with a plumb-bob and a ruler, and since I'm neglecting bushing compliance, I'm not expecting a ton of accuracy. However, I've played around with my model enough that unless I made a significant measurement error, I think it's still in the ballpark.

Without any body roll, the roll center heights are pretty easy to find:

I expected the front roll center to move around a bit, but I didn't realize how significant it was. With my car at its current (stock) height, the roll center is about 3" off the ground, but if I lower it by only 1", it goes right down to ground level.

Once I looked at the effect of roll, strange things happened. With the car at stock height, I got a pretty clear trend:

However, once I reduced the chassis height by 1", the roll center was basically impossible to find with any roll angle more than ~1 deg. Basically, the lines connecting the instant centers to the contact patches became almost parallel, so the roll center quickly moved off the screen. I bet if I reduced the chassis height by another inch, I could move the roll center to the moon.

The overall result is that the spring rates are just about meaningless in roll. Since the front roll center drops so much as the car is lowered, you have to run crazy high rates to actually increase front roll stiffness appreciably, and that isn't such a great idea for mechanical grip. Furthermore, since the roll axis becomes steeper, the lateral load transfer distribution will immediately shift to the rear by ~3%. That might not be such a bad thing considering the car pushes pretty badly, but if I ever want to shift the distribution back toward the front, it's basically impossible to do it with springs.

The anti-roll bars are already fairly stiff from the factory, and it does seem possible to effectively reduce body roll and manipulate LLTD by playing with their diameters. However, they would have to be a LOT stiffer to make a significant improvement by themselves - probably at least 1/8" larger OD, and I'm sure bushing compliance and even chassis strength might become a concern at that point. Stiffer bars also won't address other performance issues with the car; for example, they won't prevent it from standing on its nose under braking.

Has anybody else tried measuring this stuff? Do the trends I've posted seem reasonable? I've checked my work, but if I made a huge mistake somewhere, I'd like to avoid chasing my tail too much. I know that many people on here have modified their suspensions, but since it's easy to fall for the placebo effect with different shock tuning and natural frequencies, I'm hoping to get some objective information.

It seems like if I want to run the car in TTB and have a decent suspension setup, I might be looking for a unicorn. Whatever the case, I'll try to find some kind of acceptable compromise and see what's possible. I'm just hoping that a little measurement work up front will save me a lot of guesswork and (expensive) track time down the road.

Thanks.

 

[kcbrown]

Paging Norm Peterson...

I'm greatly interested in the answer to this question myself, because it'll have a major impact on my decisions about ride height. Which is to say, right now, my plan is to lower both the front and rear by at most 1/2", and to run stiffer springs in the rear than in the front (less conventional, but as long as the roll rate ratio remains about the same, I'd expect to maintain the same handling characteristics). But if the front roll center really behaves as described here, then it follows that I'm almost certainly better off putting greater stiffness in front than I'd intended while dropping the front ride height by at most 1/4".

 

[ddd4114]

Well, kcbrown, I guess we're on our own!

 

[kcbrown]

However, once I reduced the chassis height by 1", the roll center was basically impossible to find with any roll angle more than ~1 deg. Basically, the lines connecting the instant centers to the contact patches became almost parallel, so the roll center quickly moved off the screen. I bet if I reduced the chassis height by another inch, I could move the roll center to the moon.

I'm beginning to wonder if there's an error in the model here.

For a strut suspension, Puhn seems to locate the roll center at the intersection between the horizontal center of the vehicle and the line drawn between the instant center and the tire contact patch. His discussion of roll center location as it varies with roll seems to omit the possible case when the two swing arm lines are parallel.

However, in Suspension Analysis and Computational Geometry, the author talks about multiple types of roll centers, of which the geometric roll center (which, I believe, is what you're computing) is but one.

If one roll center type, such as the one you're computing, isn't useful because it's at infinity (or nearly so), then that suggests to me a limitation of the model itself, a region in which the model itself is invalid.


That book, Suspension Analysis and Computational Geometry, looks fascinating, and I went ahead and ordered it despite its $100 price tag. It's available on Amazon, and there are pages from it available on Google Books.

 

[kcbrown]

Also, suppose the roll center really is at infinity, or at least an order of magnitude or so more distant from the car's center of gravity than is the ground. Would that not basically mean that the car cannot roll any more than it has at that point? Transition through infinity is an impossibility, no?

If actual observation of a lowered car shows that the car is rolling beyond those infinity points, then it clearly illustrates that the geometric roll center model must be in error at that point.

 

[ddd4114]

I normally use Milliken's Race Car Vehicle Dynamics as my reference, and his book (as well as other references) define the geometric roll center as the intersection of the lines connecting the instant centers with the center of the tire contact patches.

If you think about it, the roll center cannot be defined using the centerline of the vehicle because it will move laterally as the car rolls.

Milliken's book also describes a force-based roll center solution, but I haven't read much about it. If I have time this weekend, I'll compare the two.

Generally, if you have an "invalid" solution, such as having parallel links, the instant centers are far enough away from the car (basically at infinity) that you can locate the roll center at the ground. In my case, the roll centers do move far away from the car, but there is still a significant vertical component.

There are a few inherent problems with my model that might be compromising its accuracy. For example, I'm completely neglecting bushing compliance, and I imagine that's very significant near the edges of the traction circle. Also, I'm defining an arbitrary point in the model about which the chassis rotates, and that's definitely not correct once the roll center moves laterally. I've experimented with the location of the point a bit, and it doesn't have a large influence, but it still has some effect. The real solution is actually an iterative approach, but that's just about impossible to do in Solidworks (as I'm doing now), so I'm just looking for something reasonably close.

 

[ddd4114]

Yes, if the roll center is located at infinity, it basically suggests that the suspension is either over-constrained, or the roll center is at some location that significantly limits chassis roll.

You are correct that if the roll center is at infinity in the model, then the car will basically just experience significant jacking forces and minimal chassis roll. In that case, I agree that it suggests an error in the model. I just can't find anything that would change it significantly. I feel like I have to be overlooking something.

 

[csamsh]

I think you guys worry too much. Several guys who race these cars and go fast say "eh" to this question. One frequent NASA time-trialer in particular on this board says don't worry unless you're dropped 2+"

 

[ddd4114]

Maybe so, but I'd still like to understand why. At the least, I think it's a good educational exercise, and it's why we have technical forums like this. If nobody knows the answer or cares about details like this, then that's fine.

Do you know how that particular time-trialer arrived at 2 inches? What happens when you drop below that?

 

[csamsh]

Maybe so, but I'd still like to understand why. At the least, I think it's a good educational exercise, and it's why we have technical forums like this. If nobody knows the answer or cares about details like this, then that's fine.

Do you know how that particular time-trialer arrived at 2 inches? What happens when you drop below that?

You run out of suspension travel, even with a coilover, unless you have crazy spring rates.

 

[Sky Render]

You run out of suspension travel, even with a coilover, unless you have crazy spring rates.

This. And you're bump steer gets frakked up.

Sent from my toilet using Tapatalk

 

[NoTicket]

After looking at your data, and having no other information to go on regarding the model you used I just have a few questions.

Are you sure you got the angle of the strut mount correct to begin with?

Did you account for changes in strut mount angle that must accompany changes in chassis roll angle?

Is your model set up in a manner in which it essentially ignores the forces exerted by the anti-roll bars (limiting suspension extension on the inner wheel will also limit instant center changes in the links on that wheel)?

Does your model account for changes in location of the tire contact patch center under chassis roll?

In general I am curious about how you calculated the angles of each component when chassis roll angle changed. Changes in roll angle occur very differently in MacPherson strut suspensions compared to an SLA/Multilink/Double wishbone suspension.

 

[ddd4114]

I know it looks like a mess without providing much explanation, but here is my model for the front view:

Basically, the strut length is allowed to change with chassis movement, but the LCA length is fixed. All of the inboard points are fixed to the chassis, and the chassis rotates about a point I define on the vertical centerline. For the outboard points, I'm fixing the location of the wheels, but the hub angle can change.

To answer your questions:

1. I'm confident that the strut angle is pretty reasonable. I checked it with an angle finder, and I also found that in my model, the strut inclination angle doesn't change all that much with suspension travel.
2. Yes
3. My model completely ignores the ARB. Could you explain what you mean by that? Based on droplink location, I'm not understanding why it acts like anything more than a third spring.
4. No, I have a fixed point for the tire contact patch. Do you think that's significant? The LCA is pretty close to parallel at stock height, and at full droop, it looks like the track width will be reduced by ~1" (~2%).

Thanks.

 

[Houstonnw]

I am curious, you have 1 deg negative camber, you are chewing up the outside of your tires, and you are worried about the roll center?

However, I would be very interested if the numbers confirm that you can't lower an S197 too much.

 

[Whiskey11]

I know it looks like a mess without providing much explanation, but here is my model for the front view:

Basically, the strut length is allowed to change with chassis movement, but the LCA length is fixed. All of the inboard points are fixed to the chassis, and the chassis rotates about a point I define on the vertical centerline. For the outboard points, I'm fixing the location of the wheels, but the hub angle can change.

To answer your questions:

1. I'm confident that the strut angle is pretty reasonable. I checked it with an angle finder, and I also found that in my model, the strut inclination angle doesn't change all that much with suspension travel.
2. Yes
3. My model completely ignores the ARB. Could you explain what you mean by that? Based on droplink location, I'm not understanding why it acts like anything more than a third spring.
4. No, I have a fixed point for the tire contact patch. Do you think that's significant? The LCA is pretty close to parallel at stock height, and at full droop, it looks like the track width will be reduced by ~1" (~2%).

Thanks.

That's a pretty big cluster fuck of points but I have a pertinent question:

Does your model take into account the lateral displacement of the front RC during body roll? If it doesn't, and it requires the RC to be "in the middle" then whenever the control arm angles don't intersect at the same point you would get a non-finite answer.

 

[ddd4114]

I am curious, you have 1 deg negative camber, you are chewing up the outside of your tires, and you are worried about the roll center?

The reason I'm concerned about the roll center is because if the roll couple increases (distance between the roll center and the CG), then chassis roll will increase using the same spring rates. If roll increases, so will the dynamic camber.

That's a pretty big cluster fuck of points but I have a pertinent question:

Does your model take into account the lateral displacement of the front RC during body roll? If it doesn't, and it requires the RC to be "in the middle" then whenever the control arm angles don't intersect at the same point you would get a non-finite answer.

It looks like a cluster, but it's the only way to fully constrain the model.

No, the roll center isn't fixed to the centerline. In roll, it moves laterally. Otherwise, the roll center height would be different between left and right turns.

 

[NoTicket]

I know it looks like a mess without providing much explanation, but here is my model for the front view:

Basically, the strut length is allowed to change with chassis movement, but the LCA length is fixed. All of the inboard points are fixed to the chassis, and the chassis rotates about a point I define on the vertical centerline. For the outboard points, I'm fixing the location of the wheels, but the hub angle can change.

To answer your questions:

1. I'm confident that the strut angle is pretty reasonable. I checked it with an angle finder, and I also found that in my model, the strut inclination angle doesn't change all that much with suspension travel.
2. Yes
3. My model completely ignores the ARB. Could you explain what you mean by that? Based on droplink location, I'm not understanding why it acts like anything more than a third spring.
4. No, I have a fixed point for the tire contact patch. Do you think that's significant? The LCA is pretty close to parallel at stock height, and at full droop, it looks like the track width will be reduced by ~1" (~2%).

Thanks.

An ARB exerts compression forces on the wheel that would otherwise experience extension travel. The ARB will deflect some amount, but the result is that in body roll the LCA angle change on the compressed wheel can corelate to a smaller magnitude change on the extending wheel.

Your model looks pretty good from a static point of view. However, since your outcomes are experimentally verifiable as incorrect regarding the effect of lowering on roll stiffness, there is something wrong. I suspect that it is the way the body is undergoing roll.

I believe your first graph is likely accurate. You would essentially lower roll center 3" for every 1" drop in CG. I don't believe your second set of data should be trusted. If it were accurate then you would see lowered cars that could not make any but the slowest turns without slamming down hard on their bump stops and that just is not the case. I am not sure of this is due to your model or if it is a shortcoming of the roll center concept all together (is the same method of calculating roll center at static angles useful when the car is already experiencing roll? Or is it only useful when each side is a near mirror of the other?)

 

[ddd4114]

Thanks for the feedback.

Sorry, I'm still not fully understanding the effect of the ARB as it applies to finding roll centers. It will of course increase roll stiffness for the reason you describe, but until you lift a wheel, I'm not seeing how it changes the geometry of the model. Are you saying that higher relative ARB stiffness will result in more lateral chassis movement for the same roll angle?

I did have the same thought as you - that many people with lowered cars would have reported severe predictability problems if this model is even remotely correct. I do think a good portion of the discrepancy is due to the limitation of using a kinematic roll center, but after reading about finding force-based roll centers, I'm not convinced that would be much more accurate given my limited resources. There will still be a ton of compliance that I can't accurately quantify - especially with the OEM suspension bushings that have the stiffness of jello at the track.

I would still like to experiment a little with how I constrain the chassis. As you suggested, track width is not really a constant in the real world, so it's not really correct that my model fixes the contact patches. However, if I took off that constraint, I would no longer have a single solution for the roll center. For a given chassis roll angle, the chassis would be able to travel on an arc, and each location would have its own solution for the kinematic roll center. Unfortunately, I don't know how to proceed from there. A reasonable approximation might come from comparing the wheel rate due to the springs to the wheel rate due to the ARB, which might be what you were implying earlier. I'm guessing it will lead to finding some sort of weird force-kinematic hybrid roll center with unknown usefulness.

Whatever the case, I only intended to use the model to find a ballpark reasonable suspension design. For my purposes, it really isn't worth my time to make a model that's super accurate. My main goal was to see which suspension parameters had large influences and what sort of effect a change would have on the system. However, I am glad that this thread stirred the pot a bit. I think it's good to question things we take for granted, even if it sometimes risks reinventing the wheel (for better or for worse).

 

[NoTicket]

All good info. With respect to the ARB, you will see angle restriction before the tire leaves the ground. The outside tire will be lifted some amount even of it is not enough to leave the ground. I'm not sure how much tire deflection is involved at certain roll angles.

Here is a picture of my car from last weekend. It has stock springs and sway bars.

As you can see there is a significant difference in sidewall height as well as contact patch. Much of this is due to the very large stock front ARB exerting upward forces on the outside wheel. I expect this would yield different LCA angles than are showing up in the existing model. I am unsure how to factor it in correctly...

On a side note, any way you can see what happens when the LCAs are lowered 3/4" on the wheel side after the car is lowered 2"? This is the equivalent of using the boss 302s/r extended ball joint.

 

[Vorshlag-Fair]

I think you guys worry too much. Several guys who race these cars and go fast say "eh" to this question. One frequent NASA time-trialer in particular on this board says don't worry unless you're dropped 2+"

Yep, I agree with Mark here. We've lowered hundreds of S197 Mustangs by 2 and even 3" from stock ride heights, and we don't ever use bumpsteer correction kits. Why? Because they don't need it. They drive just fine with the stock tie rods - and don't tend to have the unusual failures that bumpsteer kit equipped cars do. Same goes for aftermarket front control arms and a whole bunch of other parts made for the S197.

Remember: a good percentage of parts made for these cars are simply part of the array of "stuff" for you to buy. Available because someone makes it. Because people will buy it. Bolt-on braces, shiny doo-dads, whistle tips, and needlessly adjustable things to mess with, break and fail. They lighten your wallet but don't necessarily do anything worthwhile. This segment of the aftermarket is full of these things. If you add in all of the "style" crap made for the S197 that percentage of worthless available parts is approaching 90%.

There is an internet mass hysteria surrounding certain voodoo modifications that simply cannot be explained with calculations, formulas or measurements. People spend countless hours debating roll center migration, roll couples and use terms like "kinematics" and other pseudo scientific jargon... but from what I've seen in the automotive forums for the past 2 decades, the more incomprehensible the terms used the more full of crap the writer usually is.

Look - certain geometries and constraints are just part of a given chassis, and changing them to some new ideal could take tens of thousands of dollars worth in custom fabrication and parts - and usually have some serious side effects or entail other significant changes. The S197 chassis is light years better than the Fox/SN95 chassis that proceeded it (so much so that I always BEG customers that come to us with Fox/SN95 race car wishes to look at the S197 first), but it still isn't perfect. You can make significant track performance enhancements in TIRES, spring rates, dampers, antiroll bars, a few bushings, and rear axle location devices. But at the end of the day it is a big, heavy McPherson strut car with a solid rear axle. At a certain point, after you hit the big stuff mentioned above, you are just polishing a turd.

Don't get sucked into these internet arguments that have almost no real solutions. This chassis has been around for a while and a lot of smart people have raced in them - pretty much the major development aspects that matter have been tackled, until there is some new unforeseen technological breakthrough. Ask the faster drivers in the same cars what they are doing, and why. Not all of them will know why, either - there is a lot of "follow the leader" mentality in racing. But for the most part, you will see a pattern of the Parts That Matter.

Yes, think outside the box. Strive to learn, and read the Carrol Smith books on racing set-up, the ThinkFAST books and the like. But don't get bogged down in trying to make your car into something it is not. Don't try to reinvent the wheel, when the one you have is perfectly round. I see people fixate on these intangible or unrealistic goals, yet they tend to be are the ones that are still running around on skinny street tires or have some other massive handicap they refuse to address. Get that low hanging fruit first then worry about the minutia improvements later.

Honestly, if you want ideal suspension geometry, better street ride, and faster track performance... just go buy a C5/C6/C7 Corvette. That one "change" will elevate your goals far beyond you could ever achieve in an S197 chassis. Trust me, those GM engineers spent BILLIONS to develop those three Corvette chassis, and you ain't going to come up with a better suspension on a Mustang than that. Ever. Not trying to discourage you free thinkers and engineers, just keep that in mind - there is always a better chassis to start with. All of us here have Mustangs for various reasons, and I love this model, but it isn't a Corvette or a 911 or a Ferrari.

I love this chassis, and feel that it is the new car performance bargain for the past decade, bar none. Mostly due to the solid chassis and powerful 5.0L engine. It pains me to see people waste money on worthless parts trying to make these cars into something they are not. We could sell a LOT more "stuff" for these cars, and make more money doing it, but we won't do that. We talk people out of upgrades all the time, like when it doesn't meet the needs of their uses, or their experience/skill level. If you want honest advice on what works and what is a waste, just ask a shop you trust.

Anyway... that's just my two "grumpy old man" cents. Get off my lawn! I'm not trying to poke fun at anyone or any company,just stating my long held opinion that most aftermarket parts sold are worthless crap, and most internet arguments are a waste of bandwidth.

Cheers....