4. Data acquisition settings

In a crash test, the difference between a clean injury metric and a misleading one is often decided before the vehicle ever moves. Sampling rate, amplitude resolution, and dynamic range are the three acquisition parameters that govern whether the recorded signal preserves the physical event faithfully. They interact: raising the sampling rate without adequate resolution produces precise timing on a coarse amplitude grid, while maximizing resolution without sufficient bandwidth captures a smooth but time-distorted waveform. This chapter explains how each parameter affects injury criteria such as HIC, chest deflection, and neck force, and how to select settings that keep measurement uncertainty well below the thresholds that matter.

Sampling Rate and Anti-Alias Protection

The Nyquist–Shannon theorem states that a signal must be sampled at more than twice its highest frequency component to be reconstructed without aliasing. Crash pulses, however, are not band-limited by nature. Structural ringing, metal-to-metal contact, and sensor resonance can inject energy far above the frequencies that dominate occupant injury. SAE J211 therefore defines channel classes by their cutoff frequency f_c (the −3 dB point of the required low-pass filter) and mandates a minimum sample rate of 10·f_c for the class. The commonly used classes are:

The critical engineering point is that the anti-alias filter must be applied before the analog-to-digital converter, not as a post-processing step. Once aliased, high-frequency energy folds back into the passband and cannot be removed. A 20 kHz structural ring sampled at 10 kHz without a pre-filter appears as a false 10 kHz component that corrupts the pulse shape and inflates or deflates HIC depending on phase. In practice, most data acquisition systems use a fixed high-rate front end (e.g., 20 kHz or 100 kHz) with a digital anti-alias filter, then decimate to the target rate. This preserves flexibility while guaranteeing alias rejection.

Resolution, Quantization, and Effective Bits

Resolution describes the smallest amplitude change an ADC can distinguish, expressed as LSB = FSR / 2^N, where FSR is the full-scale range and N is the bit depth. A 16-bit converter over a ±500 g range has an LSB of about 0.015 g; a 12-bit converter over the same range has an LSB of about 0.24 g. For HIC, which integrates acceleration over time, quantization noise acts as a random walk that can shift the integral by several percent if the signal spends most of its life near a few LSBs.

Effective number of bits (ENOB) is the real metric, not the datasheet bit count. Noise, distortion, and clock jitter reduce ENOB. A converter specified at 16 bits may deliver only 13.5 ENOB at the target sample rate. The practical consequence: choose a range that places the expected peak at 50–80 % of full scale. Too low a range clips the peak and destroys the metric; too high a range buries the signal in quantization noise. For a head acceleration channel expected to peak near 80 g, a ±250 g range with a 16-bit ADC gives an LSB of ~0.0076 g and comfortable headroom for the CFC 60 filter overshoot.

Dynamic Range and Sensor Matching

Dynamic range is the ratio between the largest and smallest signals a channel can handle simultaneously, usually expressed in dB. It is bounded by the sensor, the cabling, the signal conditioner, and the ADC. A piezoresistive accelerometer with a 500 g range and 1 % nonlinearity cannot deliver more than about 40 dB of usable dynamic range regardless of the ADC. Matching the sensor range to the ADC range is therefore essential: a ±500 g sensor feeding a ±5 V ADC with a gain of 10 will clip at 50 g, wasting 90 % of the sensor's capability.

For injury metrics, the relevant dynamic range is not the full crash pulse but the ratio between the peak and the pre-impact noise floor. HIC computation begins at the onset of the pulse, so a noisy baseline can trigger the integration window early and inflate the result. A channel with 80 dB dynamic range and a 0.01 g noise floor supports a 100 g peak with a 10 000:1 ratio, which is more than adequate. A channel with 60 dB dynamic range and a 0.1 g noise floor may not be.

Putting It Together: A Worked Configuration

Consider a frontal sled test measuring head acceleration for HIC. The expected peak is 80 g with a CFC 60 requirement. The sensor is a piezoresistive accelerometer with a ±250 g range and 1 % nonlinearity. The acquisition system offers a 16-bit ADC with a ±5 V input and a selectable front-end sample rate of 20 kHz.

1. Sample rate: CFC 60 requires f_c = 100 Hz, so the minimum is 1 kHz. Choose 10 kHz to provide margin for the digital filter transition band and to allow later re-filtering at CFC 180 if needed. The 20 kHz front end decimates to 10 kHz after the anti-alias filter.
2. Range and gain: Set the signal conditioner gain so that 80 g produces approximately 3.5 V at the ADC input. With a ±250 g sensor and 5 V full scale, the gain is 5 V / (250 g × sensitivity). If sensitivity is 2 mV/g, the sensor output at 80 g is 0.16 V; a gain of 22 brings it to 3.52 V, or 70 % of full scale. This leaves 30 % headroom for filter overshoot and pulse variability.
3. Resolution check: LSB = 10 V / 65536 = 0.153 mV. Referred to the input, this is 0.153 mV / (2 mV/g × 22) = 0.0035 g. The quantization noise is far below the 0.05 g noise floor of the sensor, so resolution is not the limiting factor.
4. Dynamic range check: The sensor noise floor is 0.05 g RMS. The peak is 80 g. The ratio is 1600:1, or about 64 dB. The ADC's theoretical dynamic range is 96 dB, so the sensor dominates. This is acceptable for HIC because the integration window is triggered by a threshold well above the noise floor.

After acquisition, apply the CFC 60 digital filter and compute HIC. Verify that the filtered peak is within 5 % of the unfiltered peak and that the pre-impact baseline does not exceed 0.5 g. If the baseline is noisy, the HIC integration window may start early; if the peak is clipped, the HIC will be underestimated. Both are detectable by inspecting the raw and filtered traces side by side.

Verification and Documentation

Every acquisition setting should be recorded in the test log alongside the channel name, sensor model, sensitivity, range, gain, filter class, and sample rate. During pre-test verification, inject a known calibration signal (e.g., 50 g peak half-sine) and confirm that the recorded peak is within 2 % of the reference and that the pulse duration is within one sample period. After the test, check for clipping, baseline drift, and aliasing artifacts by comparing the raw and filtered signals. If any channel shows a flat-topped peak or a baseline that wanders more than 1 % of full scale, the data should be flagged before injury metrics are computed.

The settings chosen here are not universal. A rear-impact test with a shorter, sharper pulse may require CFC 180 and a 3 kHz minimum sample rate. A pedestrian headform test may use CFC 1000. The principle remains the same: match the acquisition bandwidth to the physics, match the range to the expected peak, and verify that resolution and dynamic range keep quantization and noise well below the uncertainty that would change the injury metric's pass/fail outcome.