Introduction
D-dot sensors are essential instruments in electromagnetic pulse (EMP) research, valued for their ability to measure transient electric fields without physical contact. They offer high linearity and wide bandwidth, making them ideal for capturing fast-rising pulses. However, the accuracy of these measurements depends entirely on the "sensitivity coefficient" ($k_s$), which converts a measured voltage into a physical field value. Traditional calibration methods often struggle in noisy laboratory environments, leading to significant measurement errors.
The Calibration Challenge
Standard calibration techniques typically rely on comparing peak values in the time domain or utilizing Fast Fourier Transforms (FFT). In practice, laboratory environments introduce noise, reflections, and signal distortions such as "ringing" at waveform peaks. These factors can obscure the true physical characteristics of the sensor. To address this, researchers have proposed a more robust approach treating the sensor as a "black box" through system identification.
Methodology: System Identification
The core innovation of this research is the application of System Identification theory, modeling the D-dot sensor as a Linear Time-Invariant (LTI) system.
Modeling and Pre-processing
The researchers employed a discrete Output Error (OE) model to describe the relationship between the known excitation pulse (input) and the sensor's response (output). This model effectively accounts for stochastic white noise. To ensure data integrity, several pre-processing steps were implemented:
- Resampling: Data was standardized to a 1GHz sampling rate.
- Filtering: Savitzky-Golay filtering smoothed the signals without losing essential waveform characteristics.
- Time Alignment: Cross-correlation was used to eliminate time delays between input and output signals.
Parameter Identification
Using the Akaike Information Criterion (AIC), the optimal model complexity was determined to be a 3rd-order transfer function. Partial linear regression was then applied to the resulting transfer function curve to extract the sensitivity coefficient ($k_s$).
Experimental Setup
The calibration was performed using a controlled experimental rig. A PRIMA EFT61004B pulse generator served as the excitation source, feeding into a Transverse Electromagnetic (TEM) cell where the D-dot sensor was placed.
The calibration dataset consisted of 30 sets of data with input voltages ranging from 200 V to 2 kV. Validation was subsequently performed using a separate square wave generator and a GTEM cell.
Key Results
The system identification method demonstrated significant improvements over traditional techniques:
- Accuracy: The calculated sensitivity coefficient ($k_s$) was $2.70 \times 10^{11} (V/m/s)/V$, showing high agreement with the theoretical value of $2.74 \times 10^{11}$.
- Stability: Unlike FFT methods, which often produce jagged, unstable curves at high frequencies, the proposed method yielded a smooth parametric curve, making it easier to identify the sensor's linear operating range.
- Robustness: The model maintained an average fitting degree of over 85% across various voltage levels, proving its resilience to noise and distortion.
Conclusion
By shifting from simple peak-matching to a sophisticated system identification framework, this method provides a more accurate and reliable way to calibrate D-dot sensors. This approach is particularly valuable for measurements in extreme environments, such as Nuclear Electromagnetic Pulse (NEMP) testing, where traditional calibration methods often fail due to severe waveform distortion.
Summary updated