4. Physical test instrumentation and data acquisition

Restraint system validation depends on the quality of the measurement chain, not just the crash test itself. A sled test or full-vehicle crash can be executed perfectly, yet produce unusable data if a load cell is mounted with the wrong axis orientation, if an accelerometer saturates, or if the channel is filtered incorrectly. This chapter covers the transducers most commonly used in restraint system testing — accelerometers, load cells, potentiometers, and high-speed cameras — and the data acquisition and filtering practices defined by SAE J211.

Transducer Selection and Mounting

Each transducer type answers a different physical question, and the choice is driven by what the restraint engineer needs to correlate against CAE output.

• Accelerometers measure acceleration at a point. Piezoresistive (MEMS) types are preferred for crash testing because they respond to DC (steady-state) acceleration, unlike piezoelectric types that cannot hold a static reference. Typical ranges are ±50 g to ±2000 g depending on location. A B-pillar accelerometer might use ±500 g, while a seat-track or occupant chest accelerometer may use ±50 g to ±250 g. Mounting must be rigid: a stud or adhesive block bonded directly to the structure, with the sensitive axis aligned to the vehicle coordinate system.
• Load cells measure force in a defined direction. In restraint testing, they appear in seat belt anchorages, buckle assemblies, pretensioner load paths, and steering column mounts. A load cell's rated capacity and axis must match the expected load direction; a belt load cell rated at 20 kN will not resolve the 200 N range needed for a pretensioner pull-in measurement.
• Potentiometers (string pots and rotary pots) measure displacement. They are used for belt payout, pretensioner stroke, seat track travel, and steering column collapse. A string pot with a 250 mm range and 0.1 mm resolution is typical for belt payout measurement.
• High-speed cameras provide the spatial and temporal context that point sensors cannot. They capture occupant kinematics, airbag deployment, belt routing, and component separation. Frame rates of 1000–2000 fps at 1280×720 or higher are standard for full-scale crash; 5000–10000 fps is used for airbag deployment studies.

SAE J211 Channel Classes and Filtering

SAE J211 defines four channel classes based on the frequency content of the signal. The class determines the low-pass filter corner frequency (the −3 dB point) that must be applied before the data is used for injury criteria or correlation.

The filter is a second-order Butterworth low-pass applied in both forward and reverse directions (zero-phase, or "phaseless") to avoid introducing a time shift. The standard specifies the filter's magnitude response tolerance: within ±0.5 dB of the ideal Butterworth response in the passband, and the −3 dB point must fall within a defined tolerance band. In practice, this means the data acquisition software applies the filter digitally after acquisition, not with an analog filter in the signal chain.

Data Acquisition Chain and Sampling

The acquisition chain runs from transducer to storage: transducer → signal conditioner (amplifier, bridge excitation, anti-alias filter) → digitizer (ADC) → data logger → post-processing. Each stage has requirements.

Sampling rate must be at least 10× the highest frequency of interest. For CFC 1000 (1650 Hz corner), a minimum sample rate of 20 kHz is typical; 50 kHz is common for airbag pressure and high-g structural channels. The anti-alias filter in the signal conditioner must be set below half the sample rate (Nyquist) to prevent aliasing. If the sample rate is 20 kHz, the anti-alias filter should be set at 8–10 kHz.

Bridge excitation for strain-gauge load cells is typically 2–10 V DC. The full-scale output is usually 1–3 mV/V, so a 10 V excitation gives 10–30 mV full-scale — requiring a low-noise amplifier with gain of 100–1000. Zero balance and shunt calibration are performed before each test: a known shunt resistor is placed across one bridge arm to simulate a known strain, verifying the entire channel gain.

High-Speed Camera Synchronization and Analysis

High-speed cameras must be synchronized to the data acquisition system and to each other. A common time base (IRIG-B or a shared trigger pulse) is used. The trigger is typically the sled or vehicle impact event, with pre-trigger recording enabled so the moments before impact are captured.

Camera placement follows the measurement need: a side view for occupant kinematics, a front view for airbag deployment, an overhead view for belt routing, and a close-up on the pretensioner or buckle. Lighting must be high-intensity and flicker-free — LED arrays synchronized to the camera frame rate are standard. Without adequate lighting, motion blur at 1000 fps makes displacement tracking impossible.

Analysis uses tracking software to digitize targets on the occupant, belt, or vehicle structure. The output is a time-history of 2D or 3D coordinates, which can be differentiated to velocity and acceleration. This is particularly valuable for validating CAE occupant kinematics where point sensors are not feasible.

Practical Calibration and Verification

Before a test series, each channel should be verified end-to-end. The procedure is straightforward but must be followed consistently:

1. Zero check: With the transducer installed and at rest, confirm the channel reads within ±0.5% of full scale of zero. If not, rebalance the bridge or check for preload.
2. Shunt calibration: Apply a known shunt resistor across the bridge. The output should match the calculated equivalent strain within 1%. This verifies gain, not just offset.
3. Polarity check: Tap the transducer in the positive axis direction and confirm the signal goes positive. A reversed polarity on a belt load cell will invert the entire force history and corrupt injury criteria.
4. Filter verification: Inject a known sine sweep or step input and confirm the −3 dB point matches the selected CFC class. This catches software filter misconfiguration before the test, not after.

Common Failure Modes and Troubleshooting

Three failure modes account for most unusable restraint test data:

• Accelerometer saturation: The signal clips at the rated range. Symptom: a flat-topped acceleration pulse. Fix: select a higher-range accelerometer or relocate to a lower-g location. Never "fix" this by filtering — the information is already lost.
• Load cell axis misalignment: The cell measures a component of the force, not the full vector. Symptom: belt forces that are consistently 10–20% lower than CAE prediction across all tests. Fix: verify mounting angle with a protractor or use a multi-axis load cell.
• Filter class mismatch: Test data filtered at CFC 180 compared to CAE at CFC 60. Symptom: test peaks are sharper and higher than CAE. Fix: re-filter both to the same class before comparison. This is a post-processing error, not a physical one.

High-speed camera issues are usually lighting or synchronization related. If targets are blurred, increase light intensity or reduce exposure time. If camera and data channels are offset in time, check the shared trigger and IRIG-B connection. A 1 ms synchronization error at 50 km/h corresponds to roughly 14 mm of vehicle travel — enough to shift an occupant kinematics comparison visibly.

The measurement chain is only as strong as its weakest link. A well-calibrated, correctly filtered, properly synchronized dataset is the foundation for meaningful test–CAE correlation. Without it, even the most sophisticated CAE model cannot be validated.