Model roll as a working degree of freedom

Generated from content/lms/vehicle-dynamics-ii-theory/05-roll-and-compliance-effects/01-roll-dynamics-and-load-transfer.md; edit the source file, not this page.

Source path: content/lms/vehicle-dynamics-ii-theory/05-roll-and-compliance-effects/01-roll-dynamics-and-load-transfer.md

Course: Read the forces that steer the car

Module: Add roll and compliance to the rigid model

Estimated duration: 50 minutes

Principle: roll is an active state, not a visual symptom.

When you add roll to a handling model, you are making a promise to yourself: the sprung mass is allowed to rotate about a fore-aft axis, and that rotation has consequences before you ever talk about driver feel or photos from the outside of the corner. Roll angle, roll velocity, and roll acceleration become part of the state of the vehicle. Lateral acceleration and roll acceleration create an overturning couple. The front and rear suspensions, anti-roll bars, tires, and geometry decide how that couple is resisted and how the vertical tire loads are redistributed.

That sounds abstract until you ask a practical setup question. If a car understeers in a sustained corner, did it do that because the front tires are overworked by lateral load transfer distribution, because the front geometry steered or cambered the tire in roll, because the rear roll stiffness is too low or too high, because roll center movement changed the moment path, or because the driver is simply asking the tire for too much slip angle? If roll is not a degree of freedom in your thinking, all of those get collapsed into one vague phrase: the car rolls too much. That phrase is not a model. It does not tell you where the force went, what to measure, or what to change.

Your first job is to separate body roll from suspension roll. Body roll is mainly suspension roll plus some axle roll from tire deflection. Suspension roll is the rotation of the sprung mass relative to the wheel-center line. Tire deflection adds another part of the apparent body angle, because the tire carcass and vertical compliance are also moving under lateral load. This matters because the sensor you choose determines what part of roll you actually measured. Suspension potentiometers can tell you suspension roll angle. They do not, by themselves, give you the tire roll angle caused by unsprung-mass lateral transfer and geometric transfer. If you treat a suspension-pot channel as total vehicle roll without that distinction, your model starts with a false input.

The useful mental model is a stack. At the bottom are the tire contact patches and their vertical loads. Above that are tire vertical compliance and carcass deflection. Above that are the front and rear suspension mechanisms, each with springs, dampers, anti-roll bars, roll centers, and compliance. Above that is the sprung mass with its center of gravity, roll angle, pitch, heave, yaw, and lateral motion. A simple model may keep only a few of those states. A more complete race-car model may include roll, pitch, and heave of the sprung mass plus lateral, longitudinal, and yaw velocities of the whole car. Another model may include roll, yaw, lateral motion, steering compliance, and tire vertical compliance. The exact list depends on the question, but the lesson here is that roll is not a garnish on top of yaw and lateral acceleration. It is one of the motions that can be solved.

A roll degree of freedom needs three kinds of information. First, it needs the forcing path: lateral acceleration, roll acceleration, and the overturning couple that tries to rotate the sprung mass. Second, it needs the resistance path: front and rear suspension roll stiffness, anti-roll-bar contribution, and roll damping. Third, it needs the consequence path: how the resulting roll angle and roll speed affect tire loads, tire slip angles, wheel steer angles, camber, and the balance between front and rear axles.

Do not confuse roll stiffness distribution with total roll stiffness. Vehicle roll stiffness is the sum of the separate suspension roll stiffnesses. Roll stiffness distribution is how that total is split between front and rear. If the total stiffness goes up equally at both ends, the body may roll less for a given lateral acceleration, but the front-to-rear balance may not move much. If the distribution moves, the load-transfer distribution between the axles moves, and the understeer or oversteer balance can move with it. The data-acquisition setup text is blunt about the practical goal: suspension tuning influences balance by adjusting roll stiffness distribution through springs, roll bars, damping, tire pressures, and related setup variables.

That does not mean roll stiffness is the whole load-transfer story. Tire lateral load transfer is the vertical load shifted from one tire on an axle to the other because of lateral acceleration, rotational effects, or inertial effects. Tire lateral load-transfer distribution is the share of the total that occurs at the front versus the rear. Roll stiffness distribution is one way to influence that distribution, but it is not identical to it. Roll center height, unsprung mass, tire spring rate, track width, and sprung-mass center of gravity height also belong in the story. The more precisely you want to diagnose balance, the less acceptable it becomes to say stiff front bar equals front load transfer and stop there.

Now put the anti-roll bar in the model correctly. An anti-roll bar is not just another vertical spring. In the student race car example, the bar can rock during parallel wheel travel, but resists opposite wheel travel when the body rolls. One pushrod moves as one side goes into bump, the other responds as the opposite side moves into rebound, and the torsional stiffness of the bar resists that roll motion. That is why anti-roll-bar stiffness must be converted through its geometry into an effective roll stiffness at the wheel or at the axle. Race-shop bar stiffness may be measured as force at the end of the bar arm per deflection between the arms. Your model still needs to translate that shop number through motion ratio, spring base, and linkage geometry into the roll moment equation.

This distinction is also why a car can be made stiffer in roll without becoming equally stiffer in heave. The anti-roll-bar mechanism may allow parallel wheel travel while resisting opposite wheel travel. Higher-formula three-spring systems take that idea further by separating roll and heave control when rules allow. For this lesson, the takeaway is simpler: before you say the front is stiffer or the rear is softer, ask whether you are talking about ride stiffness, heave stiffness, roll stiffness, or a linkage that couples them in a particular way.

Roll centers define another force path. A low roll center tends toward larger roll angles and uses suspension travel in roll, which can leave less travel for bumps. A high roll center reduces roll angle but can disturb the tire contact patches laterally and produce instability. Costin and Phipps describe the design compromise as relatively low roll centers used with anti-roll bars, not an extreme roll-center solution by itself. They also note the common design choice of a front roll center lower than the rear to build an initial understeer characteristic. You should not turn that into a universal setup recipe, but it is a useful warning: roll geometry and elastic roll stiffness are cooperating, and sometimes fighting, in the same balance problem.

The next trap is assuming the roll center stays where your static drawing says it is. Multibody analysis warns that roll centers migrate as the vehicle rolls, especially as the vehicle approaches limit conditions. The movement can be vertical and lateral. A simple equivalent roll stiffness model can still be accurate enough for many handling simulations, but near the limit the geometry is not frozen. If your model says one thing and the car says another, migrating roll centers are one of the reasons you check before blaming the driver or the tire.

Roll speed deserves its own line in your notebook. Roll is geometrically similar to bump on one side and droop on the other side relative to the body. As the body rolls, roll speed generally creates a scrub speed of the tires relative to the ground. That can temporarily change slip angles and tire forces. This is one of the reasons a car can feel different during the transient than it does after it takes a set. In a steady corner you may be mostly judging the final roll angle and load distribution. During turn-in, you are also judging how quickly the sprung mass moves, how the tire contact patches scrub, and how the suspension geometry steers or cambers the wheels as the roll builds.

That last point connects to the sibling lesson on hidden steer in roll. In a rolled position, suspension geometry can change wheel steer angles relative to the body. That is roll steer. It is equivalent to bump steer for independent suspension, though not for solid axles. This lesson does not try to quantify roll steer; that belongs to the sibling lesson. But you must leave space for it in the model. If the car takes a set and the driver reports that the rear seems to point itself into or away from the corner, do not diagnose only front and rear roll stiffness. Roll has moved the suspension through travel, and the geometry may have moved the steer and camber paths with it.

For an intermediate driver or engineer, the useful modeling workflow is five steps.

Step one: name the roll state. Decide what your roll angle means. Is it suspension roll from wheel travel sensors, body roll including tire deflection, or a sprung-mass angle from a model? If you mix those definitions, your comparisons will not survive the first setup change. A clean note might say suspension roll from front left and front right pots, front roll angle in degrees, rear roll angle in degrees, lateral acceleration from logger, ride-height lasers not available. That note is more valuable than a confident but undefined roll number.

Step two: split the front and rear resistance. Determine front and rear roll stiffness separately when you can. Include springs, anti-roll bars, motion ratios, and the damper positions if you are estimating equivalent linear roll damping. The multibody text describes constraining the vehicle body to rotate about an axis through the front and rear roll centers to determine front-end roll stiffness. You do not need to build that exact rig at the track, but you do need the same discipline: roll stiffness is an axle-level property built from geometry and components, not a single loose opinion about how flat the car looks.

Step three: keep lateral load transfer separate from roll angle. More roll angle does not automatically mean more total load transfer, and less roll angle does not automatically mean better grip. The load-transfer distribution is what matters for balance, and the tire loads matter because racing tires are nonlinear. A racing-car model cannot assume the gentle linear region used for passenger-car analysis. At the limit, lateral tire force depends on slip angle, camber angle, and vertical load. If a setup change makes the car flatter but transfers load in a way that worsens the tire pair you needed most, the stopwatch may punish the apparently cleaner body motion.

Step four: add the geometry warning. Roll center movement, scrub, roll steer, compliance camber, and tire vertical compliance can all modify what the tire actually sees. The sibling lessons cover some of those paths in more detail. Here, your rule is to avoid treating roll stiffness distribution as the only force path. A car with a clever bar change and a poor camber path can still lose the outside tire. A car with reduced roll angle and a bad contact-patch disturbance can still become nervous. A car that looks soft in photos can still be faster if it preserves tire load and wheel angle better.

Step five: validate against data or repeatable observation. Segers' data workflow starts by calculating suspension roll angle from suspension travel, then comparing it with speed and lateral acceleration. The example is a Formula One car at Hockenheim, but the habit scales down. You are looking for relationships: as lateral acceleration builds, do the front and rear suspension roll angles build cleanly? Does the front roll angle lag or spike compared with the rear? Does the car take a set in the same part of the corner each lap? Does a bar change alter the relationship you expected, or only make the driver adapt around the same balance problem?

For driving, the skill is to feel roll as a sequence, not a pose. In corner entry, the car is building lateral acceleration and rolling toward its outside tires. The rate of that build affects the tire scrub and temporary slip-angle behavior. At midcorner, the car may be closer to a steady rolled state, and the load-transfer distribution is a larger part of the balance story. At exit, if the corner is still loaded, roll is unwinding while longitudinal load transfer and throttle demand enter the tire budget. A driver who only says the car rolls too much loses those phases. A driver who says the car rolls onto the outside front quickly, waits, then the rear takes a set later has started to give the engineer a state-based report.

The driver also needs to avoid the comfort trap. A certain amount of roll can tell the driver how much cornering force is being developed. Excessive roll can be uncomfortable, can put the suspension near bump stops, can bottom the suspension over disturbances, and can push wheel angles into regions that require correction. But a flat car is not automatically a faster car. The useful question is whether the roll motion preserves suspension travel, tire contact, steer and camber behavior, and front-rear tire loading. Comfort matters, confidence matters, and driver feedback matters, but the model must still ask what happened at the tire.

The reason this lesson sits before the sibling four-DOF handling model is that roll changes what a reduced model is allowed to mean. A three-degree model that solves only the degrees of freedom it includes can be educational and useful. A multibody model may solve body heave and pitch relative to the ground even when an equivalent roll stiffness simplification is being used. The simplification is not wrong just because it is simple. It is wrong only when you forget what it left out. If your model has no roll state, then front and rear load-transfer distribution, roll-speed transient tire force changes, and geometry changes through bump and droop have nowhere honest to live.

By the end of this lesson, you should be able to look at a cornering problem and write a roll story in plain language. The forcing side: lateral acceleration creates an overturning demand. The resistance side: front and rear roll stiffness and damping oppose the motion, with springs and anti-roll bars translated through their geometry. The geometry side: roll centers and suspension travel decide how the body moves and how the tire contact patches are disturbed. The tire side: vertical load, camber, and slip angle decide whether the load distribution helped or hurt grip. The validation side: suspension travel, lateral acceleration, speed, and driver report must agree before you make a setup conclusion.

Keep the scope narrow. This lesson does not teach the full numerical roll equation. It does not prescribe front or rear bar changes. It does not quantify roll steer. It teaches the model discipline that makes those later calculations meaningful. Treat roll as a working degree of freedom, define what you measured, split the force path by axle, and refuse to diagnose balance from body attitude alone.

Worked example: Formula One roll channels at Hockenheim

Segers' Hockenheim example is useful because it shows what a roll model looks like when it is tied to data instead of paddock adjectives. The logged channels include vehicle speed, lateral acceleration, and front and rear suspension roll angle math channels from a Formula One car. The important lesson is not a particular number from the trace. The important lesson is the channel relationship.

Start with lateral acceleration. That is the forcing context for the roll state. Then look at front and rear suspension roll angle. If both ends build roll in proportion to lateral acceleration, the car may be moving in a coherent elastic path. If the front builds sharply while the rear lags, your model should ask whether the front roll stiffness, bar motion ratio, damper contribution, or geometry is controlling the transient. If the rear shows a different shape lap to lap, you should hesitate before making an axle-balance conclusion, because the driver, tire state, or measurement may be moving around.

Now apply the warning from the same data chapter: suspension travel measurement gives you suspension roll angle, not the entire body-and-tire roll story. The first two roll-angle components Segers names are tied to unsprung mass transfer and geometric transfer, and they combine into tire roll angle. To capture those directly, the suspension measurement needs ride-height measurement or known parameters. So the Hockenheim workflow teaches both confidence and humility. You can build a useful roll channel from suspension pots, but you must label it correctly. A suspension roll trace is a powerful witness, not the whole court case.

Worked example: the student race car anti-roll bar

The student race car anti-roll-bar mechanism is the cleanest mechanical example in the bond. During parallel wheel travel, the anti-roll bar can rock back and forth as the bell cranks rotate. During opposite wheel travel, which is what body roll creates, the bar resists rotation. One side of the suspension is moving in bump, the other side in rebound, and the torsional spring-damper representation of the bar resists that opposite motion.

This is exactly why a roll degree of freedom is not the same as a heave degree of freedom. If both wheels move upward together, the bar mechanism may mostly ride with the motion. If one wheel moves up and the other moves down, the bar is loaded in torsion. Your model therefore has to put the bar into the roll resistance path, not simply add it to ride rate as though every wheel movement used it the same way.

The shop implication is direct. A bar rated in force at the arm end has to be translated through the linkage into wheel rate effective in roll, and then into axle roll stiffness. The analysis implication is just as direct. If you change the anti-roll bar and the car's balance changes, you have not only changed how flat the car looks. You have changed how the front or rear axle participates in resisting the overturning couple, and therefore how the tire lateral load-transfer distribution may be shared.

Worked example: the six-DOF banked-track race-car model

The Carroll Smith-sourced model of steady cornering on a banked track includes sprung-mass roll, pitch, and heave, plus lateral, longitudinal, and yaw velocities of the complete car. That list is a good reality check for intermediate modeling. Roll does not live alone. It interacts with the other body motions and with the whole-car motion through the corner.

Imagine using that model to think about a car on a banked corner. The bank changes the cornering force environment. The car still has a sprung mass that can roll, pitch, and heave. The tires still see vertical load, camber, and slip-angle effects. If you left roll out, you would still have a yaw and lateral story, but you would be missing the axle-by-axle elastic response that helps determine how the vertical loads move across the tires.

The same source describes anti-roll bars for both front and rear suspensions, with bar stiffness specified in common race-shop terms and then converted into roll stiffness for the roll moment equations. That is the modeling move you should learn from the example: trackside component language must be translated into state-equation language before it can explain handling.

Drill: build the roll story from one corner

At your next event, choose one sustained corner where you can drive repeatably and one transition corner where the car clearly takes a set. Run this as a three-session drill. In session one, do not change the setup. For five laps, write only the sequence of roll feel after each lap: where the car begins to roll, whether the front or rear seems to take a set first, whether the balance changes during the roll build, and whether the car feels calmer once loaded. Keep the words phase-based, not judgment-based.

In session two, add the minimum data you have. If you have suspension potentiometers, create or review front and rear suspension roll channels with speed and lateral acceleration. If you do not have pots, use video, lateral acceleration if available, and consistent driver notes. Your success criterion is not a perfect number. It is a coherent statement that separates forcing, resistance, and consequence. For example: lateral acceleration rises before the front takes a set, rear roll follows more slowly, midcorner balance is front-limited, and exit improves only after steering is unwound.

In session three, make one small observational test, not necessarily a setup change. Compare two clean laps through the same corner. Ask whether the roll story repeated. If it did, you have a model candidate. If it did not, you do not yet have a setup conclusion. The drill is complete when you can state which part of roll you measured, which part you only inferred, and which sibling topic needs the next investigation: hidden steer in roll, force-path compliance, or the larger four-DOF handling model.

Common mistakes

Mistake one is treating flat as fast. A flatter car may feel tidier, but the bond does not support that as a universal performance rule. Excessive roll can consume suspension travel, approach bump stops, and worsen wheel-angle control. At the same time, some roll can inform the driver about cornering force, and roll reduction that harms contact-patch behavior or load-transfer distribution can make the car slower. Good looks like asking what happened to tire load, tire angle, and driver confidence, not what the outside photo looked like.

Mistake two is using suspension roll as total roll. Suspension potentiometers let you calculate suspension roll angle. They do not automatically include the tire roll components tied to unsprung-mass lateral transfer and geometric transfer. Good looks like labeling the channel as suspension roll and listing what additional measurement or calculation would be needed for total body-and-tire roll.

Mistake three is diagnosing balance from roll stiffness alone. Roll stiffness distribution is important because it influences front-rear load-transfer distribution and therefore balance, but roll centers, tire compliance, camber, steer changes, and nonlinear tire load sensitivity all matter. Good looks like saying the front roll resistance path appears dominant, then checking whether roll geometry or roll steer is also changing what the front tire sees.

Mistake four is forgetting the transient. Roll angle at midcorner is only part of the story. Roll velocity can create tire scrub speed and temporary slip-angle changes. Good looks like separating turn-in, set, and exit. If the car is sharp on initial input but steady once loaded, that is not the same diagnosis as a car that remains front-limited all through the corner.

Mistake five is trusting a static roll center too much. Roll centers can migrate as the vehicle rolls, especially near limit conditions. Good looks like treating static geometry as a starting point and using data, repeatable observation, and model humility near the limit.

Calibration cues

A better roll model shows up in the language you use. Early on, the report is vague: too much roll, too soft, too loose, too tight. As the model improves, the report becomes phased and sourced: the lateral load builds smoothly, the front suspension roll angle rises faster than the rear, the car takes a stable set, then the outside front becomes the limiting tire. That sentence is not automatically correct, but it is testable.

In data, improvement looks like cleaner channel relationships. Speed, lateral acceleration, and front and rear suspension roll angle should tell a story that repeats across similar laps. A setup change should move the expected part of that story. A front bar change that barely changes front suspension roll shape, for example, should make you inspect bar motion ratio, linkage, driver input, tire state, or whether the corner is dominated by another force path.

In driver feel, improvement feels like less surprise. You may still feel roll, but you know when it starts, when the car takes a set, and whether the balance change belongs to entry, midcorner, or exit. An instructor would hear fewer global complaints and more sequence: the car rolls onto the outside front quickly, waits, then pushes after the set; or the rear takes a set early and the car rotates before the driver can add throttle. Those are useful because they can be mapped to roll stiffness distribution, roll-speed effects, geometry through travel, and tire load sensitivity.

When to trust the simple model, and when to widen it

A simple equivalent roll stiffness model can be accurate enough for many handling simulations, and it has educational value because it forces you to state the front and rear roll resistance clearly. Use it when the question is mostly about how an axle-level stiffness or distribution change should affect a repeated cornering condition.

Widen the model when the evidence stops matching the simplification. If the car is near the limit, roll-center migration becomes harder to ignore. If the balance changes during turn-in but not at steady midcorner, roll velocity and tire scrub deserve attention. If the driver reports self-steer as the car takes a set, move to the sibling roll-steer lesson. If a bar change affects ride behavior in a way your model did not expect, revisit the linkage and whether roll and heave are really separated. If suspension travel is near bump stops, the linear stiffness picture is no longer enough.

The rule is not to worship complexity. The rule is to give each observed behavior a place to live. If a simple roll degree of freedom explains the repeated evidence, use it. If the evidence is coming from heave, pitch, compliance, tire vertical deflection, or moving geometry, widen the model before you blame the driver or chase setup at random.

Author Review

No quiz questions are attached to this lesson.

Sources

#	Document	Chunk	Pages	Score	Collection
1	Tires Suspension and Handling Second Edition Dixon John C	3eaba0c8-18ae-2e5c-7e02-d2531750b8ba	249	1	uio_books_raw_v1
2	Analysis Techniques for Racecar Data Acquisition (Jorge Sergers)	59fe115fc09f34b0eca5bfdc4d4b4f1a	12	1	uio_books_raw_v1
3	The Multibody Systems Approach to Vehicle Dynamics Michael Blundell Damian Harty	39c4b92a-d196-c932-a400-4bb5e68da569	368	1	uio_books_raw_v1
4	Racing and Sports Car Chassis Design Costin Micael Phipps David	fd5b4379-8f53-0ced-4b74-ec94618a37c7	82	1	uio_books_raw_v1
5	Racing Chassis and Suspension Design Carroll Smith	b32bdc44-76e9-a186-3d52-c34efa400aa6	150	1	uio_books_raw_v1
6	Fundamentals of vehicle dynamics Gillespie T. D. Thomas D.	807abedc-ca4f-0760-8fc6-bf1ff19f0a2c	253	1	uio_books_raw_v1
7	The Multibody Systems Approach to Vehicle Dynamics (Michael Blundell, Damian Harty)	288be60f04ba392f9d5cf8e6633ae30e	415	1	uio_books_raw_v1
8	Racing Chassis and Suspension Design Carroll Smith	dfdc3055-2e5f-bbd3-92ad-ae0b370072e4	131	1	uio_books_raw_v1