1. Crash signal acquisition fundamentals

Every injury metric you will later compute — HIC, chest deflection, femur load, Nij — inherits its validity from the acquisition chain that produced it. A crash pulse sampled at the wrong rate, or passed through a filter whose corner frequency was set carelessly, can still look plausible on a plot while being numerically wrong. This chapter establishes the physical and electronic foundations: which transducers are used in vehicle crash tests, how sampling rate and anti-aliasing interact, and how a data logger is configured so that the recorded channel is a faithful, traceable representation of the physical event.

Transducers in the Crash Environment

Crash test instrumentation divides broadly into three families. Accelerometers dominate: piezoresistive MEMS devices for the vehicle structure and anthropomorphic test device (ATD) segments, and piezoelectric types where very high frequency content or high g-ranges are expected. A typical full-frontal barrier test uses accelerometers on the vehicle tunnel, B-pillar base, engine block, and each ATD head, chest, and pelvis. Load cells measure force directly — femur load cells in the ATD legs, seat-belt load cells on the webbing, and barrier load cells in the face of the fixed rigid barrier. Displacement and rotation sensors include string potentiometers for steering column intrusion, rotary potentiometers for ATD joint angles, and optical or magnetostrictive sensors for occupant excursion.

Each family imposes its own conditioning requirement. Piezoresistive accelerometers need a stable bridge excitation and are sensitive to DC offset drift; piezoelectric devices cannot measure static or slowly varying loads and require charge amplifiers with a defined low-frequency corner. Load cells are strain-gauge bridges and share the accelerometer's bridge-excitation needs. Getting the transducer type right matters because the subsequent filter and range settings are chosen around its bandwidth and full-scale output.

Sampling Rate and the Nyquist Constraint

The sampling theorem requires that a signal be sampled at more than twice its highest significant frequency component. Crash signals are not band-limited by nature: a barrier impact produces a sharp acceleration step whose spectrum extends far beyond the range of engineering interest. The practical rule in crash testing is to sample well above the Nyquist rate for the highest frequency you intend to analyze, not merely above the highest frequency present.

SAE J211/1 recommends a channel frequency class (CFC) for each measurement, and the sampling rate must comfortably exceed the CFC corner. For CFC 1000 (used for head and chest acceleration), a corner frequency of 1650 Hz implies a minimum sample rate of roughly 10 kHz to preserve the filter's response; most labs run ATD accelerometers at 10–20 kHz. Vehicle structure channels at CFC 60 or CFC 180 are commonly sampled at 2–10 kHz. A useful working rule: sample at least five to ten times the CFC corner frequency, and never below 2.5 times the highest frequency you will filter for.

Anti-Aliasing Filters

Anti-aliasing filters are analog low-pass filters placed between the transducer conditioning amplifier and the analog-to-digital converter. Their job is to attenuate everything above the Nyquist frequency before sampling. A practical anti-aliasing filter is not a brick wall: a second-order Butterworth rolls off at 12 dB/octave, a fourth-order at 24 dB/octave. The design question is how much attenuation is needed at the Nyquist frequency to keep folded energy below the resolution of interest.

Consider a channel sampled at 20 kHz, so Nyquist is 10 kHz. If the anti-aliasing filter is fourth-order with a corner at 4 kHz, its attenuation at 10 kHz is approximately 24 × log₂(10/4) ≈ 31.7 dB, a factor of about 38. A 5 g out-of-band component at 10 kHz would fold to roughly 0.13 g — acceptable for most ATD channels. If the same filter were second-order, attenuation would be only about 15.9 dB (factor of 6.2), leaving 0.8 g of aliased content, which is not acceptable. This is why crash data loggers typically offer selectable anti-aliasing filter orders and why the filter corner must be chosen together with the sample rate, not independently.

Note the distinction between the anti-aliasing filter (analog, before the ADC, protects the sampled data) and the CFC filter (digital, applied after acquisition, defines the analysis bandwidth per SAE J211). The CFC filter does not protect against aliasing; it only shapes the spectrum for reporting and injury calculation.

Data Logger Configuration

A modern crash data logger (for example, a Kistler, HBM, or Dewesoft system) is configured channel by channel. The essential parameters are: transducer sensitivity and excitation, amplifier gain and range, anti-aliasing filter corner and order, sample rate, and channel name/units. The following is a representative configuration block for a single ATD head accelerometer channel, expressed as a structured record rather than a vendor-specific file:

channel: ATD_HEAD_X
transducer_type: piezoresistive_accelerometer
sensitivity: 0.500 mV/g
excitation: 5.0 VDC
full_scale_range: +/- 500 g
amplifier_gain: 100
anti_alias_filter: 4th_order_butterworth, corner 4000 Hz
sample_rate: 20000 Hz
cfc_class: CFC_1000
units: g
zero_offset_check: pre-test, 2 s static window

Two configuration habits prevent most acquisition failures. First, set the full-scale range so the expected peak occupies 50–80% of the range: too wide and you lose resolution, too narrow and the channel clips. A head accelerometer in a 50 km/h frontal test may peak near 150–250 g, so a ±500 g range is reasonable. Second, always record a pre-trigger window (typically 100–500 ms) and a post-trigger window long enough to capture the full rebound and settling — 200–300 ms after the primary pulse is common. The trigger itself is usually a threshold on a dedicated trigger channel, set above the noise floor but below the expected onset.

Signal Chain Overview

The figure below traces a single measurement channel from the physical event to the stored sample, showing where each configuration decision acts and where aliasing can enter if the anti-aliasing filter is inadequate.

Verification Before the Test

Configuration is only trustworthy after verification. Three checks are standard. A shunt calibration injects a known resistance across a bridge channel to simulate a defined output and confirms the gain path end to end. A zero-offset check records a static window before the event and confirms each channel sits near zero within its noise band; a drifting offset on a piezoresistive channel usually indicates a conditioning or excitation problem. A tap test on accelerometer mounting confirms the sensor is mechanically coupled and the channel responds. If a channel reads a constant value, clips early, or shows excessive noise, the fault is almost always in the transducer connection, the excitation setting, or the range — not in the software.

With the chain verified, the recorded samples are a defensible representation of the physical event. The next step is to understand how those samples are filtered and processed into the injury metrics that follow.