3. Signal acquisition chain for crash tests
A crash test produces its measurement data through a chain of analog and digital stages that sit between the sensor and the recorded sample. Understanding this chain is essential for validation work, because many "sensor faults" diagnosed later in the workflow are actually artifacts introduced by excitation, gain, filtering, or sampling decisions made long before the data reached the analysis tool. This section walks through the chain stage by stage — transducer excitation, amplification, anti-alias filtering, sampling, and the data acquisition system (DAS) specifications that tie them together — and shows how each stage constrains the others.
Transducer Excitation
Most crash-test transducers are passive or semi-active devices that require an excitation source to produce an output. Strain-gauge accelerometers and load cells are typically excited with a constant DC voltage, commonly 5 V or 10 V, delivered by the DAS bridge supply. Piezoresistive accelerometers follow the same bridge-excitation model. IEPE (Integrated Electronics Piezoelectric) sensors instead receive a constant-current excitation, typically 2–4 mA, superimposed on the signal line, with the sensor's internal electronics biasing the output to a mid-supply quiescent voltage.
Excitation is not a background detail. Bridge output is ratiometric: the sensitivity in mV/V scales directly with the applied excitation, so a 10 V bridge excited at 5 V produces half the nominal full-scale output. If the DAS excitation drifts or a channel is configured for the wrong bridge voltage, the recorded sensitivity will be wrong even though the sensor itself is healthy. For IEPE channels, an open circuit or a shorted cable collapses the bias voltage, which is one of the fastest ways to distinguish a cabling fault from a genuine sensor failure. Always verify the excitation configuration against the sensor datasheet before attributing an amplitude discrepancy to the transducer.
Amplification and Signal Conditioning
Bridge outputs are small — a 2 mV/V sensor at 10 V excitation yields 20 mV at full scale — so a differential instrumentation amplifier boosts the signal to a range the ADC can resolve. The gain setting determines the channel's full-scale input range. A typical crash DAS offers programmable gains mapping to ranges such as ±20 mV, ±100 mV, ±1 V, and ±10 V. Selecting too high a gain clips the signal during the high-g impact phase; selecting too low a gain wastes resolution and pushes small pre-impact signals into the noise floor.
Two practical rules govern gain selection. First, estimate the expected peak from the sensor's rated range and the test severity, then choose the next range above that peak with at least 20–30% headroom. Second, confirm the range against the sensor's full-scale output, not just its rated g-level, because sensitivity tolerance can shift the actual output by several percent. A 500 g accelerometer with 0.2 mV/g/V sensitivity at 10 V excitation produces 100 mV at 500 g; a ±100 mV range leaves no headroom, so ±200 mV or ±500 mV is the safer choice.
Anti-Alias Filtering
Every sampled signal must be band-limited before the ADC to prevent aliasing. Crash signals contain high-frequency content from metal-to-metal contact, structure-borne ringing, and cable microphonics that can fold back into the low-frequency band of interest. The anti-alias filter is a low-pass filter placed immediately ahead of the sampler, and its corner frequency must sit below the Nyquist frequency — half the sampling rate — with enough transition band to attenuate content above Nyquist to below the ADC's noise floor.
In practice, crash DAS units use steep analog filters, often 4th- to 8th-order Butterworth or elliptic designs, with a cutoff set by the selected sample rate. A common configuration places the −3 dB corner at roughly 0.4 times the sample rate, so a 20 kHz sampling rate pairs with an 8 kHz anti-alias corner. Some systems also apply a digital decimation filter after the ADC, which further suppresses out-of-band content but cannot undo aliasing that already occurred in the analog domain. When validating a channel, check that the configured filter matches the sample rate; a mismatched filter setting is a frequent cause of unexplained high-frequency noise or attenuated impact peaks.
Sampling Rate and Resolution
Sampling rate is chosen from the highest frequency the test must resolve, not from convenience. For occupant and vehicle crash signals, SAE J211 and ISO 6487 define channel filter classes (CFC 60, CFC 180, CFC 600, CFC 1000) that specify the low-pass characteristic applied to the data. The sample rate must be high enough that the digital record preserves the CFC-filtered signal; a widely used rule is to sample at least ten times the CFC corner frequency, so a CFC 1000 channel is commonly sampled at 10–20 kHz. Sampling at exactly twice the corner frequency satisfies Nyquist for the analog filter but leaves almost no margin for the digital filter's transition band.
Resolution interacts with gain. A 16-bit ADC over a ±100 mV range gives a quantization step of about 3 µV, which is adequate for most bridge channels. A 24-bit ADC over the same range gives sub-microvolt resolution but is only useful if the amplifier noise and the sensor's own noise floor are below that level. Higher bit depth does not rescue a channel whose gain is set too low; it only digitizes the noise more finely.
DAS Specifications and Channel Budget
The DAS ties these stages together, and its specifications must be read as a system rather than as isolated numbers. Key parameters include the number of channels, maximum sample rate per channel, ADC resolution, programmable gain and excitation options, anti-alias filter characteristics, channel-to-channel skew, and onboard memory or streaming capacity. Channel-to-channel skew matters in crash testing because accelerometers and load cells are compared across locations; a skew of even a few microseconds can distort a derived quantity such as force versus acceleration timing.
The following table summarizes a representative mid-range crash DAS configuration for illustration; the values are typical teaching examples, not a specific product datasheet.
| Parameter | Typical Setting | Validation Check |
|---|---|---|
| Bridge excitation | 5 V or 10 V DC, per channel | Matches sensor rated excitation; ratiometric sensitivity correct |
| IEPE excitation | 2–4 mA constant current | Bias voltage within sensor spec; open/short detectable |
| Gain / input range | ±20 mV to ±10 V, programmable | Peak within 70–80% of range; no clipping |
| Anti-alias filter | 8th-order low-pass, corner ≈ 0.4 × fs | Corner consistent with sample rate and CFC class |
| Sample rate | 10–20 kHz for CFC 1000 channels | ≥ 10 × CFC corner; no aliased content |
| ADC resolution | 16-bit standard, 24-bit optional | Quantization step below sensor noise floor |
| Channel skew | < 1 µs between channels | Cross-channel timing consistent for derived quantities |
Reading the Chain as a Whole
The stages are not independent. Raising the sample rate forces a higher anti-alias corner, which admits more high-frequency content and may require a steeper filter; increasing gain narrows the input range and raises clipping risk; changing excitation shifts the effective sensitivity. A validation engineer should therefore record the full chain configuration for every channel — excitation, gain, filter, sample rate, and resolution — and compare it against the test plan before the run. When a channel later shows an anomaly, the first question is not "is the sensor bad?" but "does the recorded configuration explain the symptom?" A clipped peak points to gain, a rolled-off impact edge points to the anti-alias filter, a scaled amplitude points to excitation, and high-frequency hash points to a filter/sample-rate mismatch. Only after those are excluded does the transducer itself become the prime suspect.