Tuning the diff from driver feedback & data
Compare the measured rear wheel-speed difference with the kinematic one (track × yaw rate) to see whether the diff is free, locked or letting the inside wheel spin — then link it to the driver’s phase-specific comments.
Tuning the diff from driver feedback & data: L1 Foundation explanation
Driver feedback tells you where and what ("lazy at the apex of T4", "inside rear spins out of the hairpin"); data tells you what the diff is actually doing. You need both.
The key channel: rear wheel-speed difference.
compare with what pure kinematics would demand:
- : diff free (like an open diff), no spin.
- : diff locked — the wheels are forced to the same speed (yaw resistance).
- (inside faster than outside): inside wheel spin.
Example: rear track 1.6 m, yaw rate 30 °/s (0.52 rad/s) → ≈ 0.84 m/s ≈ 3 km/h. If the logged difference on exit swings from +3 km/h to −5 km/h, the inside wheel is spinning: the diff is not holding it, or the driver is overpowering both wheels.
Segers (§4.2) lists logging both driven wheels as the way to evaluate differential work, and at La Source he shows separate left/right rear slip channels: the lightly loaded inside rear came close to locking up under braking. Left/right slip traces like these show which wheel is at its limit, and so what the diff is being asked to do.
A simple procedure:
- Write down the driver's complaint per corner and phase (entry / mid / exit).
- Overlay ΔV_meas and ΔV_kin, slip ratios per wheel, throttle, brake, gear, steering and TC activity for that corner.
- Identify the diff state in that phase (free, locked, inside spin).
- Choose the diff parameter that acts in that phase (coast ramp, preload, drive ramp/faces).
- Change one thing, re-run, compare the same corner — and check the other corners did not get worse.
Tuning the diff from driver feedback & data: L2 Practitioner explanation
A locking index channel. Normalise the measured difference by the kinematic one:
| L | Meaning |
|---|---|
| ≈ 0 | diff free (open-like) — wheels follow kinematics |
| 0–1 | partially locked/slipping clutch |
| ≈ 1 | locked (spool-like) |
| > 1 | inside wheel faster than outside — inside spin |
| < 0 | outside wheel faster than kinematics — outside spin or inside lock (braking) |
Gate it: only where > 0.5 g and V > 40 km/h (otherwise ΔV_kin is tiny and L explodes). Use the average of both fronts for vehicle speed, the yaw rate from the IMU (or ), and the correct effective rear track (dynamic, ≈ static ± camber/scrub effects). Use the IMU yaw rate rather than on entry and exit: , so equals the yaw rate only in steady state. Wheel speeds are angular speed × an assumed radius, and the loaded outside tyre rolls on a slightly smaller effective radius than the unloaded inside one; a 0.5–1 % radius error is a large fraction of a 3 % kinematic difference. Calibrate per-wheel radius on straights and either correct with a rolling-radius-vs-load map or calibrate L on a constant-radius run with the diff known to be free.
Reading phases:
- Entry: L near 1 under engine braking = coast lock holding; spikes of L < 0 (inside rear dropping below kinematics) or both rear slips going strongly negative on downshifts = blip/coast interaction.
- Mid: a passive clutch can hold the axle stuck only if , with the effective longitudinal slip stiffness of the rear pair (defined so that , ≈ 40–60 kN per unit slip for a GT3). With ≈ 0.34 m and = 1.6 m, ≈ 0.7–2 kNm in 15–30 m corners — far above typical bias preload — so the diff slips: , typically 0.1–0.4, scaling with . In fast corners (R > ~120 m) falls to a few hundred Nm and preload can stick the axle (L → 1). Expect L to rise with corner radius at the same preload (combined slip lowers and raises L, so treat these as orders of magnitude). A "lazy apex" in a hairpin is the slipping preload's anti-yaw moment , not a stuck axle: reducing preload lowers both that moment and L.
- Exit: watch the time from throttle application to L crossing 1 (inside spin onset) and how high it goes. Rapid jump to L ≫ 1 with TC cutting = not enough drive locking (or too stiff rear roll); L stuck at ~1 with steering still increasing = locked diff pushing (power understeer).
Driver ↔ data cross-check: drivers often describe the result (understeer) not the cause. "Exit understeer" may be diff locking (L ≈ 1, rising steering) or simply front saturation from aero balance at speed (L normal). Confirm with the data before reaching for the diff.
Practical constraints:
- Diff changes usually take 30–90 min (gearbox rear cover/diff removal): plan them on the run plan, not between quick runs.
- Record breakaway torque before and after; plates bed in during the first runs.
- Pair diff changes with an assessment of TC activity and rear tyre temperatures (inside-rear overheating from spin, outside-rear from too much lock on exits).
Tuning the diff from driver feedback & data: L3 Advanced explanation
Separating diff behaviour from tyre slip. The measured wheel speeds include longitudinal slip from drive torque: . With different torques on each side, even with a perfectly locked diff — and differs by the kinematics. So
A locked diff () therefore implies — the kinematic mismatch is absorbed as tyre slip. In a 20 m radius = 8 %, which the two tyres share. The lightly loaded inside tyre has the lower slip stiffness and takes most of it (≈ +6 % vs ≈ −2 % outside at neutral throttle with a 1 : 3 load ratio). Combined with its lateral load, that pushes it past its peak: spin under power, scrub and heat at neutral throttle. It is also why a passive diff rarely stays stuck in a hairpin: holding that slip needs of the order of 1–2 kNm of locking (see the Mid bullet above). So in slow corners, a high L at low throttle points to very high preload or coast locking, and means tyre scrub = mid-corner understeer and rear tyre heat.
Estimating the actual torque split from data. Without torque sensors, estimate per-wheel longitudinal force from slip ratios and a tyre longitudinal stiffness estimate (from straight-line acceleration where both sides should match): in the linear range. Then . Compare its magnitude with the yaw-moment balance from the bicycle model () to quantify how much of the observed under/oversteer the diff explains.
Experimental design. Diff effects are phase- and corner-specific and noisy. Use:
- a corner-by-corner metric table (min speed, time to full throttle, L at apex, peak inside slip, TC integral) for baseline and change;
- at least 3 comparable laps per configuration (tyre state matched);
- a two-direction test (preload up and down) when the first change gives an ambiguous result — the non-monotonic state-transition behaviour means "more" and "less" can both feel better in different corners.
The final choice is a lap-time-weighted compromise across corners; document per-corner effects so that you can predict the right preload at the next circuit from its corner speed distribution.
Key formulas
Trackside takeaways
Compare V_out − V_in with t·yaw rate: free ≈ equal, locked ≈ 0, inside spin < 0.
Locking index L = 1 − ΔV_meas/ΔV_kin: 0 free, 1 locked, > 1 inside spin. Gate on lat g and speed.
Map each complaint to a corner phase, then to the diff parameter acting in that phase.
In hairpins a locked diff forces t/R of slip into the tyres — scrub, heat and understeer.
Diff changes take time: plan them in the run plan and record breakaway torque.
Confirm the diff is the cause before changing it — exit understeer can be aero or front saturation.
On your car
GT3 loggers normally record all four wheel speeds (ABS/TC sensors) — the locking index is easy to build. Preload is the main lever; ramps where homologation allows.
Check yourself
GT3 at neutral throttle in a 30 m radius corner: rear track 1.6 m, tyre radius 0.34 m, bias preload 200 Nm (the only locking at this torque). The rear pair has an effective longitudinal stiffness of 50 kN per unit slip, i.e. SR_in − SR_out = ΔF/50 000. Estimate the locking index L = 1 − ΔV_meas/ΔV_kin.
- Segers – Analysis Techniques for Racecar Data Acquisition — §4.2 Traction and Longitudinal Slip · left/right driven-wheel slip channels, rear wheel-speed difference, La Source example
- Milliken & Milliken – Race Car Vehicle Dynamics — Ch. 20 Driving and Braking
- Milliken & Milliken – Race Car Vehicle Dynamics — Ch. 11 Testing and Development · test design and one-change-at-a-time