4. Data acquisition

Every number that appears on a Global NCAP injury scorecard — the chest deflection, the neck tension, the femur load — originates as an electrical signal from a transducer mounted inside or on a crash test dummy. Before any injury criterion can be computed, the test house must acquire that signal cleanly, filter it according to the correct channel frequency class, and time-align it with the vehicle's crash pulse. This chapter explains how the measurement chain is built, what each sensor actually measures, and how raw accelerometer data becomes the input to the injury criteria used in star rating calculations.

The Anthropomorphic Test Device as a Measurement Platform

Global NCAP frontal offset tests are conducted with the Hybrid III 50th percentile male ATD in the driver position, with a 5th percentile female ATD increasingly used in passenger positions for comparative assessment. The dummy is not a passive mannequin: it is a calibrated instrument platform whose mass distribution, joint stiffness, and geometry are controlled so that its response to a given crash pulse is repeatable across laboratories. Each ATD carries a serial number and a calibration certificate; the calibration interval is typically one year or a defined number of tests, and the certificate records the static and dynamic checks performed on the dummy.

The measurement locations that matter for Global NCAP scoring are concentrated in four body regions:

• Head — a triaxial accelerometer block at the head centre of gravity, plus a uniaxial accelerometer for head angular acceleration in some configurations.
• Neck — a six-axis upper-neck load cell measuring three forces (Fx, Fy, Fz) and three moments (Mx, My, Mz).
• Chest — a chest deflection potentiometer (or IR-TRACC in newer ATDs) measuring sternum-to-spine compression, plus triaxial accelerometers on the chest.
• Pelvis and legs — a pelvic triaxial accelerometer, femur load cells, and knee slider displacement sensors.

Each of these channels has a defined measurement range, sensitivity, and frequency response. A head accelerometer block is typically rated to ±2000 g with a resonance above 5 kHz; a femur load cell might be rated to 10 kN. Exceeding the range produces clipping, which silently corrupts the peak value used in the criterion — a failure mode that must be caught during data review, not after the report is issued.

Signal Chain: From Transducer to Scored Channel

The acquisition chain has five stages, and each one can degrade the data:

1. Transducer — converts mechanical load or acceleration into an electrical signal (typically a strain-gauge bridge or piezoelectric element).
2. On-board dummy data acquisition system (DAS) — a compact unit mounted in the dummy's torso or on the seat, which amplifies, anti-alias filters, and digitises each channel. Modern DAS units sample at 10 kHz or 20 kHz per channel with 16-bit resolution.
3. Umbilical or telemetry link — carries the digitised data from the moving dummy to the fixed acquisition rack. Cable umbilicals are common in sled and full-scale barrier tests; telemetry avoids cable drag but introduces its own dropout risk.
4. Central acquisition system — records all channels plus the vehicle's own accelerometers and the barrier/impact trigger, all on a common time base.
5. Post-processing — applies the channel frequency class filter, performs zero-offset correction, and computes the injury criteria.

Channel Frequency Class Filtering

Raw accelerometer data contains high-frequency content from structural ringing, sensor resonance, and cable vibration that is not physiologically relevant. SAE J211 (and its ISO 6487 equivalent) defines Channel Frequency Class (CFC) filters that must be applied before computing injury criteria. The filter class depends on the body region:

The filter is applied in both forward and reverse directions (zero-phase filtering) so that the peak is not shifted in time. Applying the wrong CFC is a common error: filtering head acceleration at CFC 180 will attenuate the sharp peak and under-report HIC, while filtering chest deflection at CFC 1000 will pass noise that inflates the deflection reading. The CFC used must be stated in the test report and must match the criterion definition.

From Accelerometer Data to Injury Criteria

The head acceleration signal is the input to the Head Injury Criterion (HIC). The resultant acceleration a(t) is formed from the three orthogonal components, and HIC is the maximum over all time windows of:

HIC = max [ (1/(t2 - t1)) * ∫ a(t) dt ]^2.5 * (t2 - t1)
      t1,t2
subject to: t2 - t1 ≤ 15 ms

In practice the integration is performed numerically on the CFC 1000-filtered resultant, and the window search is done over the full duration of the impact. A typical passenger-car frontal offset test produces HIC values in the range of 300–700; the Global NCAP scoring thresholds treat HIC below 650 as good and above 1000 as a likely zero-score region, depending on the protocol version.

Chest injury is assessed through chest deflection (the peak compression in millimetres) and the Viscous Criterion (VC), which combines deflection and deflection rate:

VC = max [ (dD(t)/dt) * D(t) ]   over the impact duration
where D(t) is chest deflection in metres

Neck injury uses Nij (the neck injury criterion), a normalised combination of axial force and bending moment referenced to tolerance values that differ for tension/extension, tension/flexion, compression/extension, and compression/flexion. Femur load is compared directly against a tolerance limit. Each of these criteria consumes a specific filtered channel, which is why the acquisition configuration must be defined before the test, not reconstructed afterwards.

Practical Data Review: What to Check Before Scoring

A disciplined review of the acquired data catches most problems before they reach the scorecard. The checks below are representative of what an experienced test engineer performs on every channel:

• Pre-trigger baseline — the signal should sit at zero (within the sensor's noise floor) for the pre-trigger window. A non-zero offset indicates a sensor zero error or a bridge imbalance.
• Clipping — inspect the peak against the channel range. A flat-topped peak means the true value is unknown and the channel is invalid.
• Dropout or spike — single-sample spikes usually indicate a telemetry glitch or a loose connector; they must be distinguished from real events by checking the raw (unfiltered) trace.
• Time alignment — verify that the vehicle's longitudinal deceleration begins at the expected trigger offset and that the dummy channels show the expected sequence (head acceleration after chest, chest after pelvis).
• Filter verification — confirm the CFC applied matches the criterion and that the filter was zero-phase.

Worked Example: Reconstructing a Head Resultant

Consider a representative (illustrative, not measured) dataset from a frontal offset test. The head accelerometer block provides three channels sampled at 20 kHz, already CFC 1000-filtered by the DAS. The engineer forms the resultant at each sample:

a_resultant[i] = sqrt(ax[i]^2 + ay[i]^2 + az[i]^2)

Suppose the resultant trace peaks at 85 g at t = 62 ms and the HIC window search returns t1 = 48 ms, t2 = 61 ms with an integrated mean acceleration of 52 g over 13 ms. The HIC computes to approximately 520. Cross-checking against the chest deflection channel (peak 32 mm at t = 58 ms) and the neck load cell (peak tension 2.1 kN at t = 70 ms) gives a coherent sequence: chest compresses before the head reaches peak deceleration, and the neck load peaks as the head rebounds. If the head peak appeared before the chest peak, the engineer would suspect a time-alignment error rather than a genuine physical response.

This kind of cross-channel consistency check is the practical bridge between raw acquisition and the injury scorecard. The numbers on the scorecard are only as trustworthy as the acquisition chain that produced them, and the review discipline described here is what separates a defensible rating from a plausible-looking but unreliable one.

The diagram above summarises the full chain: the ATD's regional sensors feed the on-board DAS, the digitised stream crosses the umbilical or telemetry link to the central DAQ where it shares a common time-zero with the vehicle and barrier channels, and the CFC filter stage then produces the specific filtered channels that each injury criterion consumes. The validity gate on the right is not optional — it is the checkpoint that determines whether the resulting numbers are fit to enter the Global NCAP scorecard at all.