4. Instrumentation planning

Instrumentation planning is the step that converts a crash test objective into a concrete, buildable measurement system. Before any sensor is bonded to a vehicle or any cable is routed, the test engineer must decide what to measure, with which transducer, over what range, at what sampling rate, and with what filtering. These five decisions are interdependent: a channel list drives sensor selection, sensor selection constrains range and bandwidth, and bandwidth dictates both sampling rate and anti-aliasing filter corner frequency. Getting this chain wrong produces data that looks plausible but is physically meaningless — clipped peaks, aliased high-frequency content, or phase-distorted pulses.

Building the Channel List

The channel list is the master document of the test. It should be derived directly from the test objective and the regulatory or consumer-protocol requirements, not from whatever sensors happen to be in the lab. A typical frontal 56 km/h rigid-barrier test for a consumer rating program might require on the order of 60–120 channels. Each entry in the list should carry, at minimum: channel name, measurement location, physical quantity, sensor type, full-scale range, sampling rate, filter class, and the dummy or vehicle subsystem it belongs to.

Channels fall into several families that are planned differently:

• Anthropomorphic test device (ATD) channels — head accelerations (triaxial), chest accelerations, chest deflection, pelvis acceleration, femur loads, neck forces and moments, and belt load cells. These are governed by the dummy's own instrumentation and by the rating protocol's channel requirements.
• Vehicle structure channels — accelerometers on the B-pillar, sill, floorpan, seat rails, and engine cradle, used for occupant-compartment intrusion and deceleration analysis.
• Restraint channels — seat-belt load cells, buckle load cells, pretensioner and airbag trigger signals, and steering-column displacement.
• Contact and timing channels — barrier contact switches, high-speed video sync, and event triggers.

A useful discipline is to assign every channel a purpose tag (rating, diagnostic, or research). Rating channels must meet the protocol's accuracy and bandwidth requirements; diagnostic channels can often use looser specifications. This prevents over-specifying the entire system and keeps the data-acquisition (DAQ) channel count and cost realistic.

Sensor Selection and Measurement Ranges

Sensor selection follows from the physical quantity and the expected signal amplitude. The critical rule is that the full-scale range must be large enough to avoid clipping the worst-case event, yet small enough to preserve resolution. A common mistake is to select a range based on the nominal pulse and then discover that a structural resonance or a secondary impact drives the signal far beyond it.

For ATD head accelerometers, a typical full-scale range is ±2000 g, because head resultant acceleration in a severe frontal impact can exceed 1000 g. Chest accelerometers are usually ±500 g. Femur load cells are rated around 10 kN, and neck load cells around 5 kN with moment ranges of a few hundred N·m. Vehicle body accelerometers are often ±500 g for the occupant compartment and ±2000 g for locations near the impact zone.

Range selection should also account for the DAQ input range and the sensor's own overload behavior. A piezoelectric accelerometer that saturates mechanically will produce a distorted output that no amount of post-processing can recover. Where the expected peak is uncertain, it is better to choose the next range up and accept slightly lower resolution than to risk clipping.

Sampling Rate and the Nyquist Criterion

Sampling rate is set by the highest frequency content that must be preserved. The Nyquist–Shannon theorem states that a signal must be sampled at more than twice its highest frequency component to be reconstructed without aliasing. In crash testing, the relevant bandwidth is defined by the measurement's purpose and by the filter class specified in the protocol.

ISO 6487 and SAE J211 define four commonly used filter classes for crash data:

The sampling rate must be high enough that the anti-aliasing filter's transition band is fully captured. A common rule is to sample at 5–10 times the filter corner frequency. For CFC 600 channels, a sampling rate of 10 kHz or 20 kHz is typical; for CFC 1000 channels, 20 kHz or higher is used. All channels in a test are usually sampled at a common rate (often 10 kHz or 20 kHz) for synchronization simplicity, even if some channels could tolerate a lower rate.

Anti-Aliasing Filter Requirements

An anti-aliasing filter is a low-pass filter placed before the analog-to-digital converter. Its job is to attenuate frequency content above the Nyquist frequency so that it cannot fold back into the measured band. Without it, a 12 kHz structural ringing component sampled at 10 kHz would appear as a false 2 kHz signal — indistinguishable from real data.

The filter must satisfy two conditions: its passband must be flat over the measurement bandwidth (so it does not distort the signal of interest), and its stopband must provide sufficient attenuation at the Nyquist frequency. A typical specification is a Butterworth or Bessel response with a corner frequency at or slightly above the CFC corner, and at least 40–60 dB attenuation at the Nyquist frequency. Bessel filters are often preferred for crash pulses because their maximally flat group delay preserves pulse shape.

Modern DAQ systems frequently implement the anti-aliasing filter in hardware per channel, followed by digital filtering to the exact CFC class. The hardware filter corner is set above the CFC corner to avoid double-filtering the passband, while still preventing aliasing. The digital filter then applies the precise ISO 6487 response for rating calculations.

Putting It Together: A Worked Planning Example

Consider a 64 km/h offset deformable barrier (ODB) frontal test for a consumer rating program. The objective requires head, chest, and femur data from the driver and passenger ATDs, plus vehicle B-pillar and sill accelerations. The planning sequence proceeds as follows.

1. Channel list: 3 head accelerometers per ATD (Ax, Ay, Az), 3 chest accelerometers, 2 femur load cells, 1 chest deflection, 1 neck load cell (6-axis), plus 4 vehicle accelerometers. Total ≈ 22 rating channels, plus diagnostic channels for belt loads and airbag triggers.
2. Sensor selection: Head accelerometers ±2000 g, chest ±500 g, femur 10 kN, neck 5 kN / 200 N·m, vehicle body ±500 g. All selected with a 40 % margin over the expected peak from a similar prior test.
3. Bandwidth: Head and chest channels are CFC 600 (corner 1000 Hz); femur and neck are CFC 180 (corner 300 Hz); vehicle body channels are CFC 60 (corner 100 Hz).
4. Sampling rate: The DAQ is configured for a common 20 kHz rate, which satisfies the ≥ 5× rule for the highest CFC 600 channel and simplifies synchronization.
5. Anti-aliasing: Hardware Bessel low-pass filters set at 4 kHz (well above the 1000 Hz CFC corner but below the 10 kHz Nyquist), providing ≥ 50 dB attenuation at Nyquist. Digital CFC filters are applied post-acquisition for rating calculations.

After the test, the engineer verifies that no channel clipped, that the anti-aliasing filter was active, and that the digital CFC filter response matches the ISO 6487 tolerance band. A common failure symptom is a channel that shows a suspiciously smooth or low-amplitude pulse — often the result of an anti-aliasing filter corner set too low, which attenuates the genuine high-frequency content of the crash pulse. Another is a channel with a high-frequency oscillation that does not appear in the video; this usually indicates aliasing from a disabled or misconfigured hardware filter.

Documentation and Traceability

The channel list should be finalized and frozen before the test, with each channel's range, sampling rate, and filter class recorded. After the test, the same document becomes the basis for data validation: every channel is checked against its planned range and filter setting. This traceability is what allows a rating body or a court to accept the data as valid. In practice, the channel list, the DAQ configuration file, and the post-test validation report form a single auditable package.

Instrumentation planning is therefore not a paperwork exercise — it is the engineering discipline that determines whether the measured data can answer the question the test was designed to ask. A well-planned channel list, correctly ranged sensors, adequate sampling, and properly specified anti-aliasing filters are the foundation on which every subsequent analysis depends.