# Anti-roll bars and roll stiffness distribution

> Anti-roll bars add roll stiffness without adding heave stiffness. The front/rear split of total roll stiffness sets most of the lateral load transfer distribution (LLTD) and is the primary mechanical balance tool.

Area: Springs, Anti-roll Bars & Heave · Starting level: L1 Foundation · Updated 2026-11-17
Source page: https://race-engineer-academy.vercel.app/lesson/springs-anti-roll-bars-roll-stiffness

## L1 · Foundation

### What an ARB does
An anti-roll bar is a torsion spring connecting left and right wheels of one axle. When both wheels move together (heave) the bar just rotates in its bearings – no force. When they move in opposite directions (roll), the bar twists and resists: it pushes the outside wheel down (more load on the outside tyre) and pulls the inside wheel up relative to the body. So the bar adds **roll stiffness only**.

**See it on the bench.** Below, a torsion-bar ARB sits in two bush mounts; its levers are linked to two wheel pads. Move the pads together (heave) or opposite (roll) and watch the bar's twist and the force in each drop link.

**Try it:**
- Move only **Heave** to +30 mm: both pads rise, the whole bar turns in its bushes, and twist and both link forces stay at **0**. The bar adds nothing to ride stiffness.
- Move only **Roll** to **+10 mm** (left pad up, right pad down): the bar twists **7.6°** and each link carries **4.7 kN**, equal and opposite.
- Combine heave and roll: the forces follow the roll part only.

**Takeaway:** an ARB is a spring that works only in roll. That is why it is the first tool to tune roll stiffness and LLTD (balance) without changing ride height control or ride frequency. This bar (28 × 3 mm steel tube, 560 mm long, 150 mm levers) gives **470 N/mm per link in roll** and zero in heave.

Roll stiffness $K_\phi$ (Nm/deg) of one axle from its springs:

$$K_{\phi,springs} = \frac{K_w\,t^2}{2}\cdot\frac{\pi}{180}$$

with $K_w$ in N/m and track $t$ in m. The bar adds its own term in parallel.

### Why the distribution matters
In a corner the total lateral load transfer is fixed by $m\,a_y\,h/t$ – no bar can change it. What bars change is **how it is split between front and rear axles**. Because tyre grip rises less than proportionally with load (load sensitivity), the axle that takes more load transfer loses grip:

- **Stiffer front bar** → more front load transfer → less front grip → **more understeer** (but sharper response).
- **Stiffer rear bar** → more rear load transfer → **more oversteer**.

The front share of roll stiffness, $K_{\phi f}/(K_{\phi f}+K_{\phi r})$, is the *roll stiffness distribution*; together with roll-centre and unsprung effects it gives the *LLTD*.

**Typical:** a neutral-ish car runs front LLTD a few % above its static front weight distribution (Milliken's rule of thumb is about +5 %). A 47 % front-weight car would start around 50–54 % front LLTD and be trimmed from there.

**Common misconception:** "a stiffer front bar gives more front grip because it holds the car flatter". It reduces front grip in steady state; flatter roll helps only through camber/aero, which is usually second order.

## L2 · Practitioner

### Bar rate, blades and adjustment trackside
**Bar hardware.** Most race cars use a torsion bar with blade or drop-link levers. Adjustments come from:
- **Bar diameter / wall thickness** (big steps; bars are often swapped between sessions).
- **Lever length** (multiple holes – stiffness ∝ 1/r²).
- **Blade angle**: a flat blade rotated from flat (soft) to edge-on (stiff). Its bending stiffness scales with $b h^3$ flat vs $h b^3$ edge-on – a 25×5 mm blade is 25× stiffer edge-on. Because the blade is **in series** with the torsion bar, the system range is much smaller, typically 1.5–3× from softest to stiffest. Mid positions are non-linear; always use the manufacturer's measured bar-rate table, not "click counting".
- In-cockpit adjusters (LMP2, hypercar, some GT3s) rotate a blade.

**Bar rate at the wheel.** For a torsional rate $K_\theta = GJ/L$ (Nm/rad), lever arm $r$, and bar motion ratio $\mathrm{MR}_{arb}$ (link travel / wheel travel), the bar's roll stiffness is

$$K_{\phi,ARB} = K_\theta\left(\frac{\mathrm{MR}_{arb}\,t}{r}\right)^2$$

Bar torsional rate goes with $D^4$: +10 % diameter ≈ +46 % rate.

**Checking on the car.** Disconnect one drop link: with the car on the scales, corner weights must not change when you reconnect (otherwise the bar is **preloaded** – adjust the link length). A preloaded bar creates cross-weight and asymmetric balance left vs right.

**Using it.** ARB is the quickest mechanical balance tool for **mid-corner (steady-state)** balance – entry and exit are partially driven by dampers, diff and brakes. Change one end at a time; on a GT3, one blade step typically moves front LLTD by 0.5–2 %. Expect: more front bar → less turn-in grip mid-corner, better response, more inside-rear traction on exit in RWD cars (rear axle carries less LLT).

**Aero cars** run high total roll stiffness to stop roll changing the floor and wing attitude; balance is then set by the *ratio*.

## L3 · Advanced

### From roll stiffness to LLTD – the full picture
The front-axle lateral load transfer (RCVD ch.16/18 form, small angles, rigid chassis):

$$\Delta F_{zf} = \frac{a_y}{t_f}\left[ m_s h_\phi \frac{K_{\phi f}}{K_{\phi f}+K_{\phi r}} + m_{sf}\,z_{rf} + m_{uf}\,h_{uf} \right]$$

with $h_\phi$ the sprung CG height above the roll axis, $m_{sf}$ the front share of sprung mass, $z_{rf}$ the front roll-centre height and $m_{uf}h_{uf}$ the front unsprung term. Three paths:
1. **Elastic** (through springs/bars) – distributed by roll stiffness ratio. The only one bars control.
2. **Geometric** (through links, via roll centres) – instantaneous, not through springs or dampers.
3. **Unsprung** – fixed by unsprung mass and height.

Because only path 1 is affected, a bar change moves LLTD *less* than it moves roll-stiffness distribution. On a GT car with low roll centres, path 1 dominates (often 75–85 %). On a formula car it can be < 70 %: the unsprung mass is a large fraction of the total (≈15–20 %) with its CG almost as high as the very low sprung CG, and the roll centres – small in mm, the rear often higher than the front – are a bigger fraction of that low CG height.

**Tyres in series.** Each axle's roll stiffness is suspension in series with tyres: $K_{\phi,axle} = (K_{\phi,susp}K_{\phi,tyre})/(K_{\phi,susp}+K_{\phi,tyre})$, where $K_{\phi,tyre} = K_t t^2/2$. With very stiff bars the tyre becomes the limiting element and bar steps stop doing much – the "diminishing returns" drivers feel at the stiff end.

**Chassis torsion.** A finite torsional stiffness $K_{ch}$ sits between the axles, in series with them. It makes LLTD *less sensitive* to the front/rear stiffness split: with $K_{ch}$ only 3–5× the total roll stiffness (some older GT/production-based cars), bar changes visibly lose effect. Modern formula/prototype tubs are 10–30× stiffer; GT3 roll cages are typically adequate but not perfect.

**Trade-offs of high roll stiffness:** less roll → better aero attitude and camber control, faster transient response; but more single-wheel stiffness through the bar (≈ half its roll rate per wheel), worse kerb compliance, more load fluctuation in roll (tyre-dominated roll mode with little damping) and less driver "feel" of the limit.

## L4 · Expert

### Edge cases and model limits
- **Load-sensitivity is the mechanism**, so the effectiveness of an LLTD change depends on tyre state. Near the peak of the $F_y$–$F_z$ curve's curvature (high loads, aero), a 1 % LLTD shift can be worth more than at low load. On tyres with flatter load sensitivity (some Pirelli/Michelin GT tyres when hot) bar changes feel weak – drivers report "the car doesn't respond to the bar".
- **Camber and aero couple in.** Roll changes camber (geometry lesson) and floor/wing attitude. A softer bar may *add* front grip via load transfer and *remove* it via camber/aero. On high-downforce cars the aero term can dominate, so measured balance can move "the wrong way".
- **Bump-stop engagement in roll**: if the outside front touches its bump rubber in a fast corner, front roll stiffness steps up mid-corner and LLTD jumps forward → speed-dependent understeer. The bar-only model misses this; plot damper position vs bump-stop gap.
- **Steady-state only.** LLTD from stiffness applies once roll has stopped. During roll velocity, dampers add their own "roll stiffness" – see transient balance.
- **Lateral CG shift and jacking**: at large roll the CG moves outward and jacking forces from roll centres lift/lower the body – second-order on stiff race cars but measurable on GT4/TCR.
- **Diff and drive forces** (locked diff transfers torque that loads one rear tyre) can swamp a bar change on exit; ensure the balance problem is steady-state before reaching for bars.

## Key formulas

**Axle roll stiffness from springs**

$$K_{\phi,springs} = \frac{K_w\,t^2}{2}\;[\mathrm{Nm/rad}]$$

- K_w: wheel rate per side (N/m)
- t: track (m)

**ARB roll stiffness**

$$K_{\phi,ARB} = K_\theta\left(\frac{\mathrm{MR}_{arb}\,t}{r}\right)^2,\quad K_\theta = \frac{GJ}{L},\quad J = \frac{\pi(D^4-d^4)}{32}$$

- K_\theta: bar torsional rate (Nm/rad)
- r: lever arm (m)
- MR_{arb}: link travel / wheel travel
- G: shear modulus (~79 GPa steel)
- D, d: outer / inner diameter
- L: active bar length

**Roll stiffness distribution**

$$\mathrm{RSD}_f = \frac{K_{\phi f}}{K_{\phi f} + K_{\phi r}}$$


**Front lateral load transfer**

$$\Delta F_{zf} = \frac{a_y}{t_f}\left[m_s h_\phi \frac{K_{\phi f}}{K_{\phi f}+K_{\phi r}} + m_{sf} z_{rf} + m_{uf} h_{uf}\right]$$

- h_\phi: sprung CG height above roll axis (m)
- m_{sf}: front sprung mass (kg)
- z_{rf}: front roll centre height (m)
- m_{uf}, h_{uf}: front unsprung mass (kg), its CG height (m)

**Tyre roll stiffness in series**

$$K_{\phi,axle} = \frac{K_{\phi,susp} K_{\phi,tyre}}{K_{\phi,susp} + K_{\phi,tyre}},\quad K_{\phi,tyre} = \frac{K_t t^2}{2}$$


## Trackside takeaways

- ARBs add roll stiffness only (no heave); they redistribute, never reduce, total lateral load transfer.
- Stiffer front bar = more understeer in steady state; stiffer rear bar = more oversteer.
- Neutral starting point: front LLTD ≈ front weight distribution + ~5 %.
- Blade/lever adjusters are non-linear and in series with the bar – use measured rate tables.
- Check for ARB preload on the scales: reconnecting a drop link must not change corner weights.
- Bars only act on the elastic path – roll-centre (geometric) transfer and unsprung transfer are untouched.

## By car category

### GT3

Several bar options plus blade/hole positions front and rear. Front-engine cars (AMG, BMW M4) run front LLTD ~52–60 %; rear-engine 992 GT3 R needs a relatively stiff front to tame its rear weight and protect rear traction. Rear bar often softest-possible for traction on bumpy tracks.

### One-make Cup

ARB positions (e.g. multi-hole blade/lever settings front and rear on Porsche Cup cars) are one of the few free tools – the primary balance adjuster during a session.

### GT4

Limited adjustment (few holes per bar). Big roll angles (0.6–1.0 °/g) mean camber effects of roll are significant.

### TCR

FWD: run a very stiff rear bar (front LLTD at or below static front weight ~60 %) to help rotation and keep the inside front loaded for traction. Inside rear wheel lifting is normal.

### LMP2

In-cockpit front/rear bar adjusters let the driver trim balance with fuel load and tyre wear. High total roll stiffness for aero attitude.

### Hypercar (LMH/LMDh)

Driver-adjustable bars common; balance through a stint is managed with bars and brake/hybrid settings.

### Formula 4

Front and rear bars with a few positions; low aero so bar changes are very effective and clearly felt.

### FIA F3

Roll stiffness from bars and corner springs; heave element independent. A large unsprung fraction (CG near the sprung CG height) and roll centres that are significant relative to the low CG mean a smaller fraction of load transfer passes through the springs and bars.

### IndyCar

Driver-adjustable rear (and on some configurations front) bar; drivers trim through a stint and in traffic.

## References

- Milliken & Milliken, Race Car Vehicle Dynamics (SAE 1995): Ch. 16 Ride and Roll Rates
- Milliken & Milliken, Race Car Vehicle Dynamics (SAE 1995): Ch. 18 Wheel Loads (lateral load transfer distribution)
- Milliken & Milliken, Race Car Vehicle Dynamics (SAE 1995): Ch. 21 Suspension Springs (anti-roll bar design)
- Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE 2014): §9.2 The Roll Gradient (tyre, spring and ARB roll stiffness contributions (Table 9.1))
- Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE 2014): §12.5 Modal Analysis (springs and ARBs in parallel, tyres in series)
