2. Clipping and dropout definitions

In crash test instrumentation, two failure modes dominate the corruption of otherwise valid measurement data: clipping and dropout. They are frequently confused because both produce "flat" or "missing" traces on a plot, yet they originate from entirely different physical mechanisms and demand different diagnostic responses. Establishing precise definitions before any fault-isolation work is therefore not academic housekeeping — it determines whether an engineer replaces a sensor, re-ranges a channel, or hunts a loose connector.

Amplitude Saturation (Clipping)

Clipping occurs when a signal exceeds the linear input range of some element in the measurement chain and is limited to that element's maximum (or minimum) output. The result is a waveform whose peaks are truncated — flattened at a constant value — while the underlying physical event continues to evolve. Three distinct saturation mechanisms must be separated:

• Sensor saturation: the transducer itself reaches its full-scale limit. A piezoresistive accelerometer rated at ±500 g will not faithfully report a 700 g pulse; its output pins near full scale. Capacitive MEMS devices may additionally exhibit recovery behavior (a slow return toward zero) after overload.
• Amplifier/conditioner saturation: the signal conditioner's gain stage or anti-alias filter output rail-clips at its supply-derived limit (e.g., ±10 V for a ±10 V full-scale DAQ input). This can occur even when the sensor is well within range, if gain is set too high for the expected pulse.
• ADC/digital saturation: the digitizer's input range is exceeded, and samples are pinned at the code corresponding to positive or negative full scale (e.g., 2047 or −2048 for a 12-bit bipolar converter).

The diagnostic signature of clipping is a plateau: consecutive samples at exactly the same extreme code, with the plateau duration proportional to how far the true signal exceeded the limit. Crucially, clipping is not signal loss — the channel remains connected and continues to sample; only the amplitude information is destroyed. A clipped accelerometer pulse still yields a usable time-of-peak and often a usable pulse duration, but its peak magnitude and any derived quantity (peak g, severity index, injury criterion integrals) are invalid.

Signal Loss (Dropout)

Dropout is the absence or invalidity of signal over an interval. Unlike clipping, the channel does not report an extreme value; it reports either no data, a constant baseline (often zero), or noise unrelated to the physical event. Dropout subdivides into:

• Complete loss: the channel goes to zero or to a fixed offset for a contiguous span — the classic signature of an open circuit or a powered-down conditioner.
• Intermittent dropout: brief, repeated gaps or glitches, often single-sample or few-sample, caused by a connection that makes and breaks under vibration. These are the most treacherous because they can survive casual visual inspection and corrupt derived metrics subtly.
• Data-link dropout: samples are lost in transmission (e.g., a dropped packet on an Ethernet or CAN-based DAQ), producing gaps in the recorded stream even though the sensor and conditioner were healthy.

The diagnostic signature of dropout is a gap or baseline excursion that is inconsistent with the physical event: a zero-valued interval in the middle of a crash pulse, a step to a fixed offset, or a burst of physically implausible samples. Unlike clipping, dropout destroys time-continuity — the record no longer represents the event over that interval, and any integration across the gap is meaningless.

Physical Causes in the Crash Test Environment

Crash testing is an unusually hostile environment for signal integrity: high shock (hundreds to thousands of g), broadband vibration, rapid cable flexure, and tight packaging inside a vehicle that is being deliberately destroyed. The physical causes map onto the two failure modes as follows.

Two causes deserve emphasis because they are easily misattributed. First, connector fretting: under repeated vibration, mating contacts micro-move, wearing through plating and generating intermittent high-resistance contacts. The resulting dropout is often invisible in a static bench test but appears reliably during the crash event. Second, cable strain relief failure: a cable that is not properly anchored will flex at the connector backshell, fatiguing conductors until they break — producing a complete dropout that may only manifest at the moment of peak acceleration.

Distinguishing the Two in Practice

The decisive discriminator is what the channel reports during the anomalous interval. Clipping reports a valid but limited value at the extreme of the range; dropout reports an invalid value (zero, offset, or noise) that is inconsistent with the physical event. A second discriminator is continuity: clipping preserves sample-to-sample continuity up to the plateau; dropout breaks it. A third is correlation with the event: clipping typically coincides with the peak of the physical pulse, whereas dropout may occur at any phase and is often correlated with vibration or flexure rather than amplitude.

The following SVG summarizes the decision logic and the physical-cause mapping that an isolation engineer should apply when a suspect trace is first examined.

Practical Isolation Sequence

When a suspect channel is identified, the following sequence resolves the classification quickly and reproducibly. It assumes raw counts are available alongside the scaled engineering units — always retain raw data, because scaling can mask a full-scale plateau.

1. Inspect raw extremes. Compare the channel's maximum and minimum raw codes against the converter's full-scale codes. Exact equality with consecutive repeats indicates clipping; a shortfall indicates a sensor or conditioner limit.
2. Measure the anomalous interval. For clipping, the plateau duration should scale with the over-range magnitude. For dropout, the interval is a gap or baseline excursion whose duration is unrelated to pulse amplitude.
3. Check continuity and correlation. Plot the trace against a known-good reference channel (e.g., a redundant accelerometer or the vehicle's event trigger). Clipping aligns with the pulse peak; dropout often aligns with vibration or flexure phases and may appear at low-amplitude moments.
4. Bench-verify the suspect element. For suspected clipping, apply a known over-range signal and confirm the plateau level. For suspected dropout, perform a vibration or flexure test on the cable/connector while monitoring continuity — a marginal crimp will reveal itself as intermittent resistance spikes.
5. Correct and re-verify. Re-range the channel or reduce gain for clipping; repair or replace the connector/cable for dropout. Re-run the bench verification before the next test.

Recording these classifications in the test log — with the raw extreme values, plateau or gap duration, and the corrective action taken — builds the institutional memory needed to prevent repeat failures. In a crash test campaign, the same connector family or gain setting frequently recurs across builds; a well-kept classification log turns a one-off diagnosis into a reusable check.