4. Target pulse acquisition and conditioning

Before a sled can reproduce a crash event, the laboratory must first obtain a trustworthy target pulse — the acceleration–time history that the sled is expected to match. That pulse rarely arrives ready to use. It typically originates from a full-scale crash test recorded by accelerometers mounted on the vehicle body, and it must be filtered, resampled, and formatted before it can serve as a replication reference. This chapter covers where the data comes from, how SAE J211 filtering is applied, why resampling matters, and how to prepare a clean reference pulse for the sled controller.

Data Sources from Full-Scale Crash Tests

The primary source of target pulses is instrumented full-scale crash testing. Accelerometers are mounted on the vehicle structure — commonly at the B-pillar base, the rocker panel, the rear seat cross-member, or the tunnel — and their signals are recorded through a data acquisition system at a defined sample rate. The location of the accelerometer matters enormously: a pulse measured at the B-pillar will differ from one measured at the engine cradle, because each location experiences different structural filtering and local vibration.

When selecting a data source, confirm the following attributes before proceeding:

• Sensor location and axis orientation — the pulse must correspond to the direction of interest (longitudinal, lateral, or vertical), and the sign convention must be documented.
• Sample rate — full-scale crash data is often recorded at 10 kHz or higher, which is far above what a sled controller needs.
• Channel metadata — sensitivity, full-scale range, and any hardware anti-alias filter settings applied during recording.
• Test conditions — impact speed, barrier type, vehicle mass, and occupant configuration, so the pulse can be matched to the intended sled test.

In some programs, the target pulse is not taken from a single test but is a composite or corridor derived from multiple tests, or it may come from a regulatory or consumer-information protocol (for example, a US NCAP or Euro NCAP frontal pulse). In these cases, the source document defines the pulse numerically, and the same conditioning steps still apply.

Filtering per SAE J211

Raw accelerometer data contains high-frequency noise and structural ringing that are not physically meaningful for occupant response. SAE J211 defines the standard channel filter classes used across the automotive crash safety community. The filter is a low-pass, phaseless (zero phase-shift) Butterworth filter applied digitally, and the class number indicates the cutoff frequency at −3 dB:

For a sled target pulse representing vehicle body deceleration, CFC 60 is the usual choice. The filter must be applied in both forward and reverse directions to eliminate phase shift, and the data must be padded at both ends to avoid transient artifacts at the boundaries. A common implementation is a second-order Butterworth section applied forward and backward, yielding a fourth-order effective response.

Resampling and Time-Base Alignment

Full-scale data at 10 kHz is unnecessarily dense for a sled controller, which typically operates at 1 kHz or lower. Resampling reduces the data volume and aligns the pulse to the controller's time base. The correct order is: filter first, then resample. Resampling before filtering can alias high-frequency content back into the passband.

A practical resampling workflow:

1. Apply the SAE J211 CFC 60 filter to the raw 10 kHz signal.
2. Determine the target sample interval — for example, 1 ms (1 kHz) for a typical sled controller.
3. Interpolate onto the new time base using linear or cubic interpolation.
4. Verify that the resampled pulse preserves the peak acceleration and the pulse duration within acceptable tolerance.

Time-base alignment is equally important. The pulse must be referenced to a consistent t = 0, usually the moment of impact or the start of the deceleration event. If the source data uses a different trigger point, shift the time vector so that the pulse onset aligns with the sled controller's expected start.

Preparing the Reference Pulse for Replication

Once filtered and resampled, the pulse is formatted for the sled controller. Most controllers accept a time–acceleration table, often as a CSV or a proprietary format. The preparation step includes:

• Sign convention — confirm whether the controller expects deceleration as positive or negative. Document the convention and keep it consistent throughout the program.
• Units — convert to the controller's expected unit (typically g or m/s²).
• Endpoint handling — ensure the pulse starts and ends at zero acceleration, or at a defined baseline, to avoid step inputs.
• Duration — trim or extend the pulse to cover the full event, including the rebound phase if the controller requires it.

A useful sanity check is to numerically integrate the conditioned pulse twice to obtain velocity and displacement. The velocity change (delta-v) should match the known impact delta-v of the source test within a few percent. If it does not, the filtering or sign convention is likely wrong.

Worked Example: Conditioning a Frontal Pulse

Assume a full-scale frontal crash test recorded a B-pillar longitudinal acceleration at 10 kHz. The raw peak is 42 g, and the impact delta-v is 56 km/h. The conditioning steps are:

1. Load raw channel: 10 kHz, 42 g peak, 120 ms duration
2. Apply SAE J211 CFC 60 filter (100 Hz cutoff, zero phase)
3. Resample to 1 kHz using linear interpolation
4. Verify peak after filtering: ~38.5 g (expected reduction)
5. Integrate to velocity: delta-v = 55.4 km/h (within 1.1% of target)
6. Format as time (ms) vs acceleration (g), deceleration positive
7. Export CSV for sled controller

The reduction from 42 g to approximately 38.5 g after filtering is normal — the CFC 60 filter removes high-frequency spikes that inflate the raw peak. The delta-v check confirms that the conditioning preserved the physical event. If the integrated delta-v had come out at, say, 48 km/h, the likely causes would be an incorrect sign convention, a missing filter, or a truncated time window.

The pipeline above summarizes the sequence: raw acquisition, SAE J211 filtering, resampling, formatting, and verification. Each stage must be completed in order, and the verification step is not optional — it is the only way to confirm that the conditioned pulse still represents the physical crash event. A pulse that passes the delta-v check and preserves the peak shape is ready to be loaded into the sled controller as the replication target.