2. Accelerometer and pressure sensor fundamentals for crash detection

Crash detection depends on two complementary sensing channels: a MEMS accelerometer that measures the vehicle body's deceleration, and a pressure sensor that measures the rate of pressure rise inside a sealed door cavity during a side impact. Understanding how these devices convert physical motion into an electrical signal — and where their limits lie — is essential before any calibration or validation work begins.

MEMS Accelerometer Operating Principle

A capacitive MEMS accelerometer is a micromachined proof mass suspended by silicon springs between fixed electrode plates. When the vehicle decelerates, inertia displaces the proof mass relative to the frame. This displacement changes the differential capacitance between the moving mass and the fixed plates. An application-specific integrated circuit (ASIC) converts that capacitance change into a voltage or digital output proportional to acceleration.

The mechanical transfer function is a second-order system characterized by its resonant frequency fn and quality factor Q. For automotive crash sensors, fn typically falls between 1.5 kHz and 5 kHz, well above the crash signal bandwidth of interest (DC to roughly 400–500 Hz). This separation ensures the sensor operates in its flat, mass-controlled region where output tracks acceleration linearly. The damping ratio is deliberately set near critical (ζ ≈ 0.7) to suppress ringing after a shock event.

Two sensing axes matter in practice. A low-g accelerometer (±2 g to ±10 g range) captures the early, lower-amplitude phase of a crash and is used for safing logic. A high-g accelerometer (±50 g to ±250 g) captures the violent peak deceleration and triggers deployment. Many airbag ECU designs integrate both channels in a single package or place them on separate dies to avoid cross-coupling.

Frequency Response and Dynamic Range

The frequency response defines how faithfully the sensor reproduces crash pulses of different durations. A frontal barrier impact produces a deceleration pulse lasting 80–150 ms with significant energy below 200 Hz. A pole or narrow-object impact generates a shorter, sharper pulse with energy extending to 400–600 Hz. If the sensor's −3 dB bandwidth is too low, the measured peak will be attenuated and the ECU may under-integrate the crash severity.

Dynamic range is the ratio between the largest measurable acceleration and the noise floor, expressed in dB or simply as the usable g span. A sensor with ±100 g range and 10 mg RMS noise has a dynamic range of roughly 80 dB. In practice, the lower bound is set not by the sensor alone but by the ECU's signal conditioning — anti-alias filter noise, ADC quantization, and PCB vibration coupling all raise the effective floor.

Example — bandwidth check for a 50 g frontal pulse:
Assume a half-sine pulse of amplitude 50 g and duration 100 ms. Its fundamental frequency is 1/(2×0.1) = 5 Hz, with significant harmonics up to about 5× the fundamental (25 Hz) for a smooth pulse, but a real crash pulse with a sharp rise time of 5 ms has energy content up to 1/(π×0.005) ≈ 64 Hz. A sensor with a −3 dB point at 400 Hz passes this without measurable attenuation. If the same sensor were specified at 150 Hz, the 64 Hz component would still pass, but a 2 ms rise-time event (≈160 Hz content) would begin to roll off.

Saturation Behavior

Saturation occurs when the proof mass displacement exceeds the linear travel range and the output clips at the supply rail or a defined clamp level. This is not a gradual compression — it is a hard nonlinearity. Once saturated, the sensor output no longer carries amplitude information, and any integration performed by the ECU will underestimate velocity change (Δv).

Two saturation mechanisms exist. Mechanical saturation happens when the proof mass hits the mechanical stop; recovery may take several milliseconds and can produce a ringing artifact. Electrical saturation occurs when the ASIC output stage clips before the mass reaches its stop. Electrical saturation is preferable because recovery is immediate and predictable.

For a side-impact pressure sensor, the analogous limit is the maximum measurable pressure rate. A typical side-pressure sensor has a full-scale range of 300–600 kPa with a specified response time under 1 ms. If the pressure rise exceeds the full-scale rate, the output clips and the ECU loses the ability to distinguish a severe impact from a very severe one.

Critical validation point: Never assume that a sensor rated at ±250 g will never saturate in a real crash. Offset, temperature drift, and mounting resonance can shift the effective range. Always verify saturation margin against the worst-case crash pulse from the vehicle's crash pulse library, including tolerance stack-up.

Signal Chain from Sensor to ECU Decision

The diagram below shows the complete signal path for both sensing channels, from the physical event through the sensor element, signal conditioning, and into the deployment decision logic. Note where saturation and bandwidth limits enter the chain.

Crash Detection Signal Chain — Accelerometer and Pressure Channels Physical Event Impact / Deceleration MEMS Accelerometer Proof mass + capacitive bridge → ASIC fn = 1.5–5 kHz, ζ ≈ 0.7 Signal Conditioning Anti-alias LPF ADC (12–16 bit) fc = 400–500 Hz ECU Decision Logic Safing + deployment algorithm Δv integration, threshold Side Impact Door cavity compression Pressure Sensor Piezoresistive / capacitive diaphragm FS = 300–600 kPa, t < 1 ms Signal Conditioning High-pass / rate differentiation dP/dt threshold ECU Decision Logic Pressure + acceleration fusion Side airbag trigger ⚠ Saturation Risk Points: Accelerometer mechanical/electrical clip → Δv underestimation | Pressure sensor full-scale rate limit → loss of severity discrimination Bandwidth Check fc > 2× max signal Accelerometer path Pressure path Signal conditioning ECU decision

Practical Calibration and Validation Implications

When calibrating a crash detection system, the sensor's frequency response and saturation behavior directly determine the valid operating envelope. A typical calibration procedure involves:

  1. Bandwidth verification: Apply a swept-sine or impulse excitation to the sensor mounted on a shaker and measure the −3 dB point. Confirm it exceeds the highest significant frequency in the crash pulse library by at least a factor of two.
  2. Saturation margin test: Apply a calibrated shock pulse at 120% of the expected maximum crash amplitude. Verify the output does not clip and that the recovery time is within specification (typically < 2 ms).
  3. Cross-axis sensitivity check: Apply acceleration perpendicular to the primary sensing axis and confirm the output remains below the specified cross-axis sensitivity (usually < 5% of full scale).
  4. Pressure rate validation: For side sensors, use a pressure pulse generator to apply a known dP/dt ramp and verify the output tracks linearly up to the full-scale rate.

In a representative validation dataset, a ±100 g accelerometer with a 2 kHz resonant frequency and 400 Hz anti-alias filter was tested against a 50 g, 100 ms half-sine pulse. The measured peak was within 1.5% of the reference, and the integrated Δv error was under 3%. When the same sensor was subjected to a 180 g, 5 ms pulse, the output saturated for approximately 1.2 ms, causing a 12% Δv underestimation. This illustrates why saturation margin must be validated against the worst-case pulse, not the nominal one.

Calibration rule of thumb: Set the sensor full-scale range to at least 1.5× the maximum expected crash acceleration, and verify that the −3 dB bandwidth is at least 2× the highest frequency component of interest. For pressure sensors, ensure the full-scale pressure rate exceeds the maximum dP/dt by a factor of 1.3.

Understanding these fundamentals — MEMS transduction, frequency response, dynamic range, and saturation — provides the technical basis for every subsequent calibration decision. The next step is to apply these principles to specific sensor mounting locations and ECU algorithm tuning.

← Previous 📖 Contents Next →