4. Instrumentation and data acquisition for knee mapping

Knee mapping is only as trustworthy as the measurement chain that feeds it. The impactor applies a controlled load to the knee of a seated or positioned occupant, and the resulting force–deflection characteristic is reconstructed from three synchronized sources: a load cell in the impactor, accelerometers on the impactor and on the vehicle or buck structure, and high-speed cameras that capture the physical interaction. Each source answers a different question, and the mapping result is only valid when all three agree in time and space. This section explains what each instrument contributes, how the channels are configured and synchronized, and how to recognize when the data cannot be trusted.

What each instrument actually measures

The load cell is the primary force transducer. In a typical knee-mapping impactor it is mounted between the impactor face and the moving mass, so it measures the force transmitted through the contact interface rather than the inertial force of the impactor body. This distinction matters: a load cell placed behind the accelerating mass will include the mass–acceleration product and will overstate the force applied to the knee. The cell should be selected so that the expected peak force falls between roughly 10% and 90% of its rated capacity — a 5 kN cell for a peak near 3 kN gives good resolution without risking overload during setup mishaps. Piezoresistive cells are common because they respond to static and quasi-static loads; piezoelectric cells offer higher stiffness and bandwidth but require charge conditioning and cannot hold a DC offset, which complicates the low-frequency portion of a knee map.

Accelerometers serve two purposes. An accelerometer on the impactor mass provides the inertial reference needed to interpret the load cell signal and to detect ringing or off-axis motion. Accelerometers on the buck or vehicle structure record the boundary condition — if the structure moves during the impact, the measured deflection is not purely the knee's response. Uniaxial IEPE accelerometers with a range of ±500 g and a mounted resonance above 30 kHz are typical for impactor use; structural channels may use ±50 g to ±200 g units. Mounting matters more than the datasheet: a stud-mounted accelerometer on a machined flat is reliable, while a wax or adhesive mount can shift or resonate and produce spurious high-frequency content.

High-speed imaging does not produce a numeric channel, but it is the reference that validates the other two. Cameras positioned to view the knee–impactor interface from the side and, where possible, from above allow the analyst to confirm contact location, impactor alignment, soft-tissue compression, and any rotation of the leg. A frame rate of 1000 fps is a practical minimum for knee mapping; 2000–5000 fps is preferable when the event is short and the contact patch is small. Resolution should be sufficient to resolve the impactor face edge to within about 1 mm at the subject distance, which usually means a narrow field of view and careful lighting rather than a high pixel count over a wide scene.

Configuring the acquisition chain

A workable starting configuration for a knee-mapping test looks like this. The load cell is sampled at 20 kHz or higher with a 5 kHz anti-alias filter, giving adequate margin over the contact event's frequency content while suppressing impactor ringing. Accelerometer channels use the same sample rate and a low-pass corner appropriate to the sensor's mounted resonance — typically 10 kHz for impactor channels. All channels share a common trigger, and the pre-trigger buffer should hold at least 100 ms so the zero-load baseline is captured before contact. Zero the load cell and accelerometers immediately before the run, with the impactor at rest and the leg in its final position; a drifting baseline is the most common cause of an apparently offset force curve.

Camera synchronization deserves explicit attention. Hardware triggering from the same event that starts the DAQ is preferable to software triggering, because software latency varies between runs. If the cameras cannot be hardware-triggered, record a visible sync pulse — an LED or a flash — in the camera field of view and on a DAQ channel, then align the two records in post-processing. A 1 ms misalignment is negligible for the force curve but can shift the apparent contact frame by two frames at 2000 fps, which is enough to misjudge the instant of first contact.

From raw channels to a usable knee map

The knee map is built from force and deflection, and deflection is where the instrumentation choices become visible in the result. Deflection is not measured directly by the load cell; it is derived either from impactor displacement (integrated from impactor accelerometer data, with drift correction) or from image tracking of the impactor face relative to a fixed reference on the leg. Integrating acceleration is sensitive to baseline offset and will drift over the event, so the integration should be high-pass filtered or anchored to the camera-derived displacement at the start and end of contact. When the two methods disagree by more than a few millimetres at peak deflection, the cause is usually structural motion at the accelerometer mount or a camera calibration error, and the run should be treated as suspect until resolved.

Force–deflection curves should be filtered consistently before comparison. A common practice is to apply a low-pass filter at 300–600 Hz to the force channel for mapping purposes, while retaining the unfiltered channel for peak-force reporting. The filter corner must be stated alongside the map, because a lower corner will reduce the apparent peak force and shift the curve's shape. The same applies to any smoothing applied to the camera-derived displacement.

Recognizing bad data

Several failure signatures are worth memorizing. A force trace that rises before the camera shows contact indicates a trigger offset or a pre-loaded load cell. A force trace with strong, sustained oscillation at a single frequency suggests impactor ringing or a loose mount rather than knee response. An accelerometer channel that saturates — a flat-topped peak — invalidates any integration that uses it. Camera frames that show the impactor face out of plane, or the leg rotating before contact, mean the boundary condition assumed by the map is not the one that was tested. In each case the correct response is to fix the instrumentation and repeat the run, not to post-process around the problem; a knee map is a physical measurement, and the acquisition chain is part of the measurement.

The instrumentation described here defines the resolution and credibility of everything that follows in the mapping workflow. A well-configured chain produces force, acceleration, and image records that agree on when contact began, how the load developed, and how far the knee deflected — and those three agreements are the practical definition of a valid knee-mapping dataset.