# Power steering & steering torque as a data channel

> Power assistance changes how much of the tyre's message reaches the driver. A steering-torque channel lets the engineer see the front limit, driver effort and system health objectively.

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

## L1 · Foundation

**Why race cars have power steering.** GT and prototype cars have heavy front axles, wide slicks, big caster and high downforce. Unassisted steering torque would exceed what a driver can sustain over a stint (tens of N·m at the rim, peaks higher). Assistance lowers effort while trying to keep the *shape* of the torque signal – the feel.

**Types**
- **Hydraulic (HPAS)** – engine- or electric-pump-driven; assist force from a pressure difference across a piston on the rack, modulated by a torsion bar valve in the column.
- **Electro-hydraulic (EHPAS)** – electric pump, hydraulic rack; independent of engine speed.
- **Electric (EPS)** – electric motor on the column or rack, controlled by a torque sensor; assist curve programmable.

**Who has it (generally):** GT3, GT4, TCR, LMP2, Hypercar and Super Formula (KYB electric power steering since the SF19) – assisted. F4, Formula Regional, F3, F2 and IndyCar – unassisted (IndyCar's steering effort on street courses is notorious).

**Steering torque as a channel.** Measured with a strain-gauged column or quick-release hub, or taken from the EPS torque sensor via CAN. It is the torque the **driver** applies – on an assisted car not the total rack force.

**What it tells you, basic level**
- Steering effort per lap: peak and RMS torque → driver fatigue, especially relevant on unassisted cars.
- Torque rising with lateral g, then **flattening or dropping** while steering angle keeps increasing → front is saturating.
- Spikes on kerbs → kickback: the impact force acts through the wheel centre on the kingpin offset at hub height, plus caster/KPI jacking from the vertical hit.

## L2 · Practitioner

**Assist and feel.** Total rack force is shared:

$$F_{rack} = \frac{2\pi}{c}\,\eta\,T_{driver} + F_{assist}$$

With hydraulic assist, $F_{assist} = \Delta p \cdot A_{piston}$; the valve opens more with torsion-bar twist, so assist rises with driver torque, often non-linearly (boost curve). Implications:
- **Assist curve slope matters for feel.** If assist grows steeply at high torque, the driver's torque saturates while rack force keeps rising – the SAT drop at the limit is compressed into a small change of driver torque: numb at the limit.
- **Too little assist** – heavy, tiring; the driver's precision falls over a stint.
- **EPS maps** (some cars/series allow selection): lower assist for fast tracks (more feel, effort tolerable), more for street tracks.

**Building useful math channels**
1. **Torque vs lateral g scatter** per corner type: initial slope = steering "weight"; the roll-over point = front limit. Compare setups or tyre sets at the same track.
2. **Steering torque gradient:** $\frac{dT_{SW}}{d\delta_{SW}}$ gated in quasi-steady cornering. A falling gradient while steering increases flags front saturation (understeer onset) objectively, independent of the driver's comment.
3. **Rack force estimate (hydraulic):** with assist pressure sensors on both rack chambers, rack force = driver term + Δp·A. Rack force × steering-arm length ÷ total trail ($t_m + t_p$) approximates front-axle lateral force – a powerful front-grip estimate (see Level 3).
4. **Health channels:** PAS fluid temperature and pressure, pump current, EPS motor temperature and current – detect overheating (assist derating) before the driver reports "steering got heavy".

**Pitfalls**
- Column torque includes steering-wheel inertia and friction: filter and gate on low steering velocity for steady-state analysis.
- Hydraulic assist has hysteresis; compare loading and unloading separately.
- Sensor offset drift: check the straight-line mean (should be near zero; a consistent offset can mean camber/caster asymmetry or crosswind).

## L3 · Advanced

**Front-limit detection, quantitatively.** Using the SAT physics: $M_{kp} \approx F_{y,f}(t_m + t_p(\alpha))$. In the linear region, $T_{SW}$ grows proportional to $a_y$. As the front approaches its limit, $t_p$ falls, so $T_{SW}/a_{y}$ falls. Define:

$$\Psi = \frac{T_{SW}}{a_y}\bigg/\left(\frac{T_{SW}}{a_y}\right)_{linear}$$

(normalised by the low-g reference slope from the same run). $\Psi$ ≈ 1 in the linear range, falling to 0.5–0.7 as the front saturates. Lap metrics such as "time with $\Psi$ < 0.8" or average $\Psi$ in mid-corner phases quantify front-limited running – complementary to the understeer angle, but *independent of the steering-ratio and Ackermann calibration*.

**Reading $\Psi$ with the understeer angle.** $\Psi$ only tells you whether $t_p$ is collapsing, i.e. how close the front tyres are to their $M_z$ peak.
- Positive understeer angle with $\Psi$ ≈ 1: ordinary linear understeer. In the bicycle model $\delta_u = K\,a_y$ with $K > 0$ while $F_y \propto \alpha$, so $\Psi$ stays at 1. This is the tyre slip balance (weight distribution, LLTD) plus any compliance and roll steer – not saturation.
- Understeer angle rising *and* $\Psi$ falling together: front saturation.

**Pitfalls**
- **Assisted cars:** compute $\Psi$ from rack force (driver term + $\Delta p\,A$, or EPS motor torque), never from column torque alone. With a rising boost curve, $T_{SW}/a_y$ falls as rack force rises even with perfectly linear tyres, and column-torque $\Psi$ would flag saturation that is not there.
- **Downforce:** vertical load lengthens the contact patch, so $t_p$ and SAT per g rise with speed. Build the linear reference slope in speed bands, not one slope for the whole lap.

**Assist removal for analysis.** For hydraulic systems with pressure sensors:

$$\hat F_{y,f} \approx \frac{F_{rack}\cdot a_{arm}}{t_m + t_p}$$

with $a_{arm}$ the steering-arm length (moment about the kingpin per unit rack force). Since $t_p$ is unknown and varies, use it for *relative* comparisons (setups, tyre sets, stints) rather than absolute force.

**Driver-effort engineering.** On unassisted cars (F3, F2, IndyCar), steering effort is a physical limit over a race: a driver's precision deteriorates late in the race if RMS torque is high. Track RMS/peak torque per lap and over a race; compare against driver fitness data. Setup choices that raise front aero load or caster increase effort – sometimes a slightly less "optimal" setup is faster over the full race because the driver can keep precision.

**Transient content.** Spectral analysis of steering torque reveals kickback frequencies (kerbs, wheel-hop), and hydraulic assist instability (chatter ~10–30 Hz). Persistent steering-torque oscillation with low steering-angle motion is a sign of assist instability or worn rack mounts.

## Key formulas

**Rack force with assist**

$$F_{rack} = \frac{2\pi}{c}\,\eta\,T_{SW} + \Delta p\,A_{piston}$$

- c: rack travel per rev (m/rev)
- T_SW: driver steering torque (N·m)
- Δp: assist pressure difference (Pa)
- A_piston: rack piston area (m²)

**Normalised torque ratio (front-limit indicator)**

$$\Psi = \frac{T_{SW}/a_y}{(T_{SW}/a_y)_{linear}}$$


**Front lateral force estimate**

$$\hat F_{y,f} \approx \frac{F_{rack}\,a_{arm}}{t_m + t_p}$$

- a_arm: steering-arm length (m)

## Trackside takeaways

- The column torque sensor measures driver torque; on assisted cars rack force needs assist pressure or motor current too.
- Steering torque vs lateral g rolls over as the front saturates – an objective understeer marker.
- Steep assist curves numb the limit feel; assist level is a feel setting, not just comfort.
- Monitor PAS temperatures and currents – assist derating looks like a sudden balance change to the driver.
- On unassisted cars, track RMS steering effort as a race-performance factor.

## By car category

### GT3

Hydraulic or electro-hydraulic assist on most cars; some allow assist maps. Overheated PAS fluid in hot races is a known failure mode – log fluid temperature.

### TCR

Road-derived EPS: torque sensor data is often available on CAN. Assist maps are generally homologated.

### LMP2

Assisted steering; steering torque logging is common and used as a front-limit check alongside understeer angle.

### Hypercar (LMH/LMDh)

Assisted; front e-motor torque (LMH) and high downforce change steering torques – compare torque traces between deploy and non-deploy phases.

### FIA F3

Unassisted: steering torque directly reflects front axle forces. Effort tracking matters for driver fitness and late-race precision.

### IndyCar

Unassisted: steering effort on street courses, with the aeroscreen's extra front mass, is a major fitness and setup constraint. Column torque is a direct front-force signal; caster and front-aero choices are partly effort choices.

### Super Formula

KYB EPS-assisted: the EPS torque sensor gives a clean driver-torque channel, but the assist map shapes the limit feel – use motor torque too when computing front-limit metrics.

## References

- Milliken & Milliken, Race Car Vehicle Dynamics (SAE 1995): Ch. 19 Steering Systems (power steering, steering torque and feel)
- Milliken & Milliken, Race Car Vehicle Dynamics (SAE 1995): Ch. 2 Tire Behavior (aligning torque vs slip angle)
- Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE 2014): §7.4 Driver Activities That Indicate Vehicle Balance (steering-angle-based balance indicators – compare with the torque-based ones here)
- Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE 2014): §14.6 Steering (driver feels front grip through aligning-torque changes at the wheel)
