A Vectorial Current Density Imaging Method Based on Magnetic Gradient Tensor

Yangjing Wu, Mingji Zhang, Chung‐Kang Peng, Zehuang Zhang, Yichen He, Wenwei Zhang, Liang ChangSensors 2023 · DOI

TL;DR

This paper introduces a novel method for "seeing" electrical current flow within electronic devices, not just its strength but also its direction. By measuring subtle changes in magnetic fields (magnetic field gradients), they can reconstruct a detailed, non-invasive image of current density, overcoming limitations of previous techniques that only showed magnitude and were susceptible to noise.

What Problem Does This Paper Address?

Understanding how electrical current flows within electronic devices is crucial for diagnosing faults, ensuring reliability, and optimizing performance. Current methods often rely on invasive "lumped parameter" measurements (like I-V curves) that provide an overall picture but can't pinpoint the exact location or direction of issues. Existing non-invasive techniques, known as Magnetic Current Imaging (MCI), primarily visualize the magnitude of current density. However, they struggle with background noise and cannot reveal the direction of current flow, which is vital for understanding phenomena like current leakage or magnetic induction. This paper aims to bridge this gap by developing a method that can image both the magnitude and direction of current density, while also being more robust to noise.

Methods

The core innovation lies in using the magnetic field gradient tensor to reconstruct the current density. Instead of just measuring the magnetic field itself, the researchers measure how it changes in space. This is achieved by using a tri-axial fluxgate sensor and employing a "second-order differencing" technique.

Here's a breakdown:

  • Ampere's Law Foundation: The method is grounded in Ampere's Law, which relates magnetic fields to electric currents.
  • Magnetic Field Gradient Tensor: The key insight is that the spatial derivatives of the magnetic field (the gradient tensor) directly relate to the current density. The paper derives a closed-form analytical solution for inverting these gradients to obtain the vectorial current density (magnitude and direction).
  • Noise Rejection: A significant challenge in magnetic sensing is background noise. The paper addresses this by using a "reference raw dataset" to subtract common-mode noise from both near-field and far-field measurements. This is further enhanced by the second-order differencing, which inherently suppresses certain types of noise.
  • Experimental Setup: A custom system was built, including a non-magnetic substrate, a metrology grating for precise positioning, a step-motor control module, a DC current supply, a data acquisition device, and the tri-axial fluxgate sensor.
  • Data Processing: Raw magnetic field data is processed through several steps: noise rejection, second-order differencing to obtain the gradient tensor, and finally, the closed-form inversion to reconstruct the vectorial current density.

Datasets

The paper does not explicitly mention training datasets. Instead, it focuses on experimental measurements and simulations.

  • Experimental Data: Measurements were taken using a tri-axial fluxgate sensor over various DUTs (Device Under Test) shaped as triangle, square, circle, and snake patterns, made from 16AWG electrical wires. These wires were excited with a 5A DC current.
  • Simulation Data: Finite Element Analysis (FEA) models were created using COMSOL Multiphysics to simulate the magnetic field gradients and current density distributions for the same DUT shapes and conditions. This simulation data serves as a benchmark to validate the experimental results.

Key Insights and Ideas

  • Vectorial Imaging is Achievable: The most significant insight is the successful demonstration of a method to image not just the magnitude but also the direction of current density non-invasively. This opens up new possibilities for understanding complex current flow patterns.
  • Magnetic Field Gradients are Key: Moving beyond simple magnetic field measurements to their spatial derivatives (gradients) is crucial for accurate current density reconstruction and for mitigating noise.
  • Closed-Form Solution: The derivation of an analytical, closed-form solution for current density inversion simplifies the reconstruction process and provides a direct link between measured gradients and the desired current density image.
  • Noise Suppression Effectiveness: The combination of reference subtraction and second-order differencing proved effective in reducing common-mode background noise, leading to cleaner and more reliable current density images.
  • Experimental Validation: The strong agreement between experimental results and FEA simulations (with a relative error less than 3%) validates the proposed method's accuracy and reliability.

Key Results

  • Spatial Resolution: Achieved a spatial resolution of 0.69 mm.
  • Current Density Resolution: Achieved a current density resolution of 24.15 mA/mm².
  • Probe-to-Sample Separation: Demonstrated effective imaging with a separation of 2 mm.
  • Scanning Area: Capable of scanning a large area of 300 mm × 300 mm.
  • Accuracy: Experimental results showed a relative error of less than 3% compared to simulation results.
  • Noise Rejection: Demonstrated highly coherent magnetic field results with a relative error less than 1% after noise rejection.
  • Qualitative Results: The reconstructed vectorial current density images clearly showed the ingoing and outgoing current imprints, allowing for the distinction of different DUT shapes and current flow patterns.

Limitations and Future Work

  • Probe-to-Sample Distance: The paper acknowledges that the probe-to-sample distance (2 mm) leads to a "lift effect," causing magnetic field attenuation and diffusion in the current density images. While a correction factor was applied for magnitude, diffusion is best alleviated by reducing this distance.
  • Sensor Size and Sensitivity: The authors suggest that improving spatial resolution and current density resolution will require reducing the size of the sensor and increasing its sensitivity. This would enable imaging at finer scales, such as on printed circuit boards (PCBs) or integrated circuits (ICs).
  • Future Applications: The method is expected to be valuable for applications in PCBs, ICs, and superconductor inspection, particularly for identifying current leakage and other electrical anomalies.

Summary updated

Abstract

Magnetic current imaging is deemed an emerging powerful technique for visualizing electrical currents in electronic devices. However, the existing magnetic-field-based Fourier Transform back-evolution method is limited by its mono-function of imaging the magnitude of current density in devices under test, and subject to background noise distortion. Here, we developed a novel vectorial current density imaging method based on the detection of the magnetic field gradient generated by current carrying conductors. A closed form solution of current density inversion was analytically derived and numerically verified. Experiments were conducted by scanning tri-axial fluxgate sensor over different shapes of electrical wires. The results show that a current density resolution of 24.15 mA/mm<sup>2</sup>, probe-to-sample separation of 2 mm, and spatial resolution of 0.69 mm were achieved over a maximum scanning area of 300 mm × 300 mm. Such a method is verified to be capable of simultaneously imaging both magnitude and directions of current density, which is a promising technique for in situ noninvasive inspection for the power electronic and semiconductor industry.
BibTeX
@inproceedings{wu2023vectorial,
  author = {Yangjing Wu and Mingji Zhang and Chung‐Kang Peng and Zehuang Zhang and Yichen He and Wenwei Zhang and Liang Chang},
  title = {{A Vectorial Current Density Imaging Method Based on Magnetic Gradient Tensor}},
  booktitle = {Sensors},
  year = {2023},
  doi = {10.3390/s23135859},
}
Figure