# Understeer angle from data

> Subtract the kinematic (Ackermann) steer angle the car would need at zero slip from the actual road-wheel angle: what remains is the understeer angle – a math channel and a lap statistic for tracking balance over runs, setups and stints.

Area: Steering & Ackermann · Starting level: L2 Practitioner · Updated 2026-11-16
Source page: https://race-engineer-academy.vercel.app/lesson/st-understeer-angle

## L1 · Foundation

**The idea.** At very low speed, a car following radius $R$ needs a road-wheel angle of only

$$\delta_{kin} = \frac{L}{R}$$

(the **kinematic** or **Ackermann steering angle** – not to be confused with Ackermann *geometry* between the two front wheels). At speed, the front and rear tyres run slip angles. If the car needs **more** steer than $L/R$, it is understeering; less, it is oversteering:

$$\delta_u = \delta - \frac{L}{R} = \alpha_f - \alpha_r$$

**From data.** $R$ is not measured, but in steady cornering $R = V^2/a_y$:

$$\delta_{kin} = \frac{L\,a_y}{V^2}$$

so:

$$\delta_u = |\delta_{road}| - \left|\frac{L\,a_y}{V^2}\right|\cdot 57.3$$

with $\delta_{road}$ in degrees, $L$ in m, $a_y$ in m/s², $V$ in m/s. Positive = understeer, negative = oversteer. Absolute values remove the left/right sign convention.

**Worked example:** GT3, L = 2.7 m, 150 km/h (41.7 m/s), 1.6 g (15.7 m/s²): $\delta_{kin} = 2.7 \times 15.7 / 1736 = 0.0244$ rad = 1.40°. Logged steering 30° with ratio 13:1 → road wheel 2.31°. Understeer angle ≈ +0.9°.

How big is it? Segers' GT1 examples range from slightly negative to about 1.2° lap average, with race-stint values often only 0.2–0.5°. The value depends heavily on gating, ratio calibration and compliance, so the *trend* matters more than the absolute value.

## L2 · Practitioner

**Building the channel (Segers' method)**
1. **Road-wheel angle:** measure the steering-wheel → outside-wheel map on turn plates and fit it (linear or polynomial). Math channel: $\delta_{road} = f(\delta_{SW})$.
2. **Kinematic angle:** with speed in km/h and lateral acceleration in g:
   $\delta_{kin} = \frac{G_{lat}\cdot 9.81}{(V/3.6)^2}\cdot L \cdot 57.3$
3. **Understeer angle:** $\delta_u = |\delta_{road}| - |\delta_{kin}|$.
4. **Gate it:** only valid in quasi-steady cornering. Typical gates: V > 50–60 km/h (the $1/V^2$ term explodes at low speed), $|a_y|$ > 0.4–0.5 g, low steering velocity, and optionally brake < threshold / throttle within a band to split entry, mid, exit.

**Using it**
- **Track map overlay:** colour the understeer angle on the track map to spot which corners/phases are understeer- or oversteer-limited.
- **Lap statistics:** average understeer angle per lap, per stint. Segers shows a run where average understeer fell lap by lap (balance drifting towards oversteer as the tyres aged) and another where a driver's best laps came at low average understeer while his team-mate's came at higher values – evidence for different setup needs.
- **Setup comparison:** average understeer angle in mid-corner phase before/after a change, at similar track conditions.
- **Phase gating:** entry (brake on, steering increasing), mid (throttle low, |steer rate| small), exit (throttle rising) – reveals "entry understeer, exit oversteer" patterns.

**Pitfalls**
- **Sensor offsets:** a 0.02 g lateral offset or 1° steering zero error biases the channel; calibrate on straights (mean of both should be ~0).
- **Banking and camber of the track:** the accelerometer measures a component of gravity on banked corners; $a_y$ no longer equals $V^2/R$. Prefer the yaw-rate version (Level 3) there.
- **Compliance:** the effective ratio is higher at load, so the channel reads high in high-load corners. Comparisons between setups remain valid; absolute values between cars do not.
- **Non-linear range:** the SAE-style definition of understeer/oversteer is strictly valid in the linear tyre range and steady state. Race cars live in the non-linear range – use it as a *comparative* tool.

## L3 · Advanced

**Yaw-rate formulation.** In steady state $a_y = V\,r$ and $R = V/r$, so the kinematic angle can also be written $\delta_{kin} = L\,r/V$. With a yaw-rate gyro:

$$\delta_u = |\delta_{road}| - \left|\frac{L\,r}{V}\right|$$

Advantages: far less affected by banking and accelerometer mounting angles – the body-fixed gyro reads $(V/R)\cos\theta_{bank}$, about 1 % low at 9° (Indianapolis) and ≈9 % low at 24° (Texas turns 3–4; turns 1–2 are 20° since 2017), against a gravity bias of $g\sin\theta_{bank}$ (≈0.4 g at 24°) in the accelerometer version; correct with a bank-angle map if you need absolute values. Less sensitive at low speed ($1/V$ instead of $1/V^2$). Disadvantages: in transients yaw rate lags steering, so the channel spikes positive on turn-in and negative on exits even for a neutral car – gate on steering velocity or compare the two versions.

**Relation to understeer gradient.** In the linear range:

$$\delta = \frac{L}{R} + K\,a_y \quad\Rightarrow\quad \delta_u = K\,a_y$$

so a scatter plot of $\delta_u$ vs $a_y$ (gated, steady state) has slope $K$ (deg/g at the road wheel) – the understeer gradient. For a race car the plot is linear at low g and curves upward (front-limited) or downward (rear-limited) near the limit. Comparing these curves between setups is more informative than a single average: two setups with the same lap-average understeer can have very different limit behaviour.

**Separating contributions.** $\delta_u = \alpha_f - \alpha_r$ only if steering compliance, roll steer and compliance steer are zero. In reality:

$$\delta_u^{meas} = (\alpha_f - \alpha_r) + \delta_{compl}(M_z, F_y) + (\varepsilon_f - \varepsilon_r)\phi$$

Use the steering-torque channel ($\Psi$, see power-steering lesson) or a sideslip sensor (optical / GPS dual-antenna) to separate tyre saturation from compliance and geometry. With a sideslip measurement $\beta$ at the CG, you can compute front and rear axle slip angles directly. In the convention used above (slip positive in the direction of the turn): $\alpha_f \approx \delta - \beta - a\,r/V$ and $\alpha_r \approx b\,r/V - \beta$, so $\alpha_f - \alpha_r = \delta - L\,r/V = \delta_u$. Check your logger's sign conventions for $\beta$ and $r$ before trusting the result.

**Statistics done right.** Use distance- or time-weighted averages within gates, not raw sample means; track condition (grip factor), fuel and tyre age covary with balance – plot understeer vs tyre age across stints, and compare setups at matched tyre age. A 0.1° shift in average understeer angle is meaningful only if run-to-run noise (same setup, same tyres) is smaller – estimate it from repeated laps first.

## L4 · Expert

**Limits of the method.** The kinematic reference $L/R$ assumes a bicycle model in steady state. In the non-linear range, an "understeer angle" of 1° could arise from a front axle at its peak and a rear axle well below it, or from both axles at large slip angles with a 1° difference – very different handling states with the same channel value. The channel says nothing about *margin*.

Better approaches for expert analysis:
- **Axle slip angles** from a sideslip sensor compared with tyre-model peak slip angles: tells you which axle is at the limit.
- **Yaw-moment method** (RCVD Ch. 8): compute the yaw moment needed (I_z·dr/dt) vs available; requires good mass/inertia and tyre data but separates control from stability.
- **Lap-sim correlation:** run the same metric on simulated laps; differences between real and simulated understeer angle per corner identify where the vehicle model (compliance, tyre, aero map) is wrong.

**Driver influence.** The understeer angle depends on the line the driver chooses: a driver who turns in early and carries more steering mid-corner shows higher values without the car being different. Normalise by comparing the same driver, or use segments where lines converge. Segers' driver-versus-understeer plot is a reminder: the "best" balance is a property of the car–driver system, especially in endurance racing where a compromise must work for several drivers.

## Key formulas

**Kinematic (Ackermann) steer angle**

$$\delta_{kin} = \frac{L}{R} = \frac{L\,a_y}{V^2}$$

- L: wheelbase (m)
- a_y: lateral acceleration (m/s²)
- V: speed (m/s)

**Understeer angle (logger units)**

$$\delta_u = |\delta_{road}| - \left|\frac{G_{lat}\cdot 9.81}{(V/3.6)^2}\,L\right|\cdot 57.3$$

- G_lat: lateral acceleration (g)
- V: speed (km/h)
- δ_road: outside road-wheel angle (deg)

**Yaw-rate version**

$$\delta_u = |\delta_{road}| - \left|\frac{L\,r}{V}\right|\cdot 57.3$$

- r: yaw rate (rad/s)

**Bicycle-model identity**

$$\delta - \frac{L}{R} = \alpha_f - \alpha_r = K\,a_y$$


## Trackside takeaways

- Understeer angle = road-wheel angle − L·ay/V²; positive = understeer.
- Needs a measured steering-wheel → outside-wheel map; calibrate sensor zeros on straights.
- Gate on speed, lateral g and steady state; split by corner phase.
- Use it comparatively (setups, stints, drivers), not as an absolute number between cars.
- Yaw-rate version is better on banked tracks; cross-check saturation with steering torque or sideslip.

## By car category

### GT3

Segers' examples are GT cars: average understeer-angle trends over a stint show the balance moving to oversteer as tyres age (often rear degradation in RWD cars). Use it to compare drivers in endurance line-ups.

### One-make Cup

Same car for everyone: lap-average understeer angle vs lap time across drivers in the team is a quick way to see who needs which balance.

### TCR

FWD cars show high understeer angles under power on exits; gate by throttle to separate entry/mid balance from traction-limited exit.

### LMP2

Telemetry bandwidth limits: send the lap-statistic (average understeer angle per phase) rather than full channels for live monitoring.

### FIA F3

Unassisted steering means column compliance is significant at high load – calibrate the ratio and compare runs rather than absolute values.

### IndyCar

On ovals, banking makes the accelerometer-based version unreliable; use the yaw-rate version.

## References

- Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE 2014): §7.5 The Understeer Angle (kinematic steer angle, math channel, lap statistics)
- Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE 2014): §7.6 Vehicle Balance with a Yaw Rate Sensor
- Milliken & Milliken, Race Car Vehicle Dynamics (SAE 1995): Ch. 5 Simplified Steady-State Stability and Control (understeer gradient)
- Milliken & Milliken, Race Car Vehicle Dynamics (SAE 1995): Ch. 19 Steering Systems
