Research on Micro-CL Geometric Errors

Jiajun Zhang, Liu Shi, Cunfeng Wei, Baodong Liu, Biao WeiIEEE Access 2021 · DOI

Abstract

Computed laminography (CL) is a nondestructive testing technique for planar objects. It has been applied in applications like printed circuit board inspection and paleontological fossils research. However, the CL system calibration is a challenging task for complex mechanical structure and many degrees of freedom. To evaluated the influence of geometric errors, a new-type microscopic CL (micro-CL) with different kinds of geometric errors was simulated. The results can guide the installation and calibration of a micro-CL scanner.

Introduction: The Challenge of Imaging Planar Objects

Computed Tomography (CT) is widely regarded as the gold standard for 3D internal imaging. However, it faces significant limitations when applied to flat, planar objects such as printed circuit boards (PCBs) or thin fossils. Because these objects are much wider than they are thick, standard CT scans suffer from extreme variations in X-ray absorption at different angles, leading to poor image quality.

Computed Laminography (CL) addresses this by utilizing a tilted rotation axis, making it ideal for "flat" samples. Despite its advantages, CL systems are mechanically complex with numerous degrees of freedom. Even minute misalignments—on the scale of micrometers or fractions of a degree—can introduce "ghosting," blurring, or geometric distortion in the final 3D reconstruction. This research quantifies the tolerance levels for these mechanical errors, providing a practical roadmap for high-resolution micro-CL system calibration.

Methods: Modeling and Simulation

To isolate and evaluate individual geometric errors, the researchers employed a "simulation-first" methodology.

  1. Mathematical Modeling: A rigorous coordinate system transformation was established, mapping the Global Coordinate System (lab space) to the Local Detector Coordinate System. This mathematical framework allows for the precise representation of errors such as detector tilt, slant, and skew.
  2. Simulation Environment: The study utilized the ASTRA Toolbox, a high-performance, GPU-accelerated open-source framework for tomographic reconstruction.
  3. Digital Phantom: To visualize blurring and artifacts clearly, a digital "phantom" was created. It consists of high-contrast line pairs (to test resolution) and small spheres (to test geometric fidelity) within a 580×580×24 pixel volume.
  4. Reconstruction & Metrics: The SIRT (Simultaneous Iterative Reconstruction Technique) algorithm was used for reconstruction. Image degradation was quantified using two metrics: NMSDC (Normalized Mean Square Difference of Contrast) for large localized errors and NAADC (Normalized Average Absolute Difference of Contrast) for widespread errors.
Figure 1: Micro-CL scanning structure diagram and the ideal coordinate system used for simulation.

Key Insights: Identifying Critical Misalignments

The simulation revealed that not all mechanical errors are created equal. Some have a negligible impact, while others are catastrophic for image quality.

  • The "Skew" Danger Zone: A critical finding is that while detector slant and tilt (up to 0.2°) have relatively minor effects, detector skew (rotation within the detector plane) is highly damaging. It introduces "drag" artifacts and rotates the entire reconstructed volume.
  • Z-Axis Sensitivity: While X and Y source offsets primarily shift the image, errors along the Z-axis (the source-to-object distance) lead to a significant loss of fine detail, making narrow line pairs indistinguishable.
  • The Magnification Paradox: Sensitivity to geometric errors increases as the object moves closer to the X-ray source. High-magnification "micro-CL" requires significantly tighter mechanical tolerances than standard industrial CL systems.
  • Detector Offsets: Offsets in the detector plane (XD and YD axes) primarily cause blurring, obscuring the internal structure of the sample.
Figure 5: Coordinate transformations illustrating the rotation angles used to model detector geometric errors.

Key Results: Hardware Calibration Limits

The study establishes a definitive "Error Limit Table" to guide the installation and calibration of micro-CL hardware:

  • Object Motion: Must be restricted to under 1μm.
  • X-ray Source Position: Tolerable up to 3μm in X, Y, and Z directions.
  • Detector Skew: Must be calibrated within 0.2°.
  • Detector Offset: Relatively forgiving, with a tolerance of up to 100μm.

Qualitative analysis via "difference maps" shows that source drift creates directional shadows, while detector misalignments produce complex, flower-like distortion patterns.

Limitations and Future Work

This study was conducted under idealized simulation conditions. In real-world environments, multiple errors often interact non-linearly. Additionally, factors like X-ray beam hardening, photon noise, and environmental variables (thermal expansion, vibrations) were not modeled. Future research will focus on developing auto-calibration algorithms capable of detecting and correcting these geometric errors post-scan by analyzing the projection data itself.

Summary updated

Abstract

Computed laminography (CL) is a nondestructive testing technique for planar objects. It has been applied in applications like printed circuit board inspection and paleontological fossils research. However, the CL system calibration is a challenging task for complex mechanical structure and many degrees of freedom. To evaluated the influence of geometric errors, a new-type microscopic CL (micro-CL) with different kinds of geometric errors was simulated. The results can guide the installation and calibration of a micro-CL scanner.
BibTeX
@inproceedings{zhang2021research,
  author = {Jiajun Zhang and Liu Shi and Cunfeng Wei and Baodong Liu and Biao Wei},
  title = {{Research on Micro-CL Geometric Errors}},
  booktitle = {IEEE Access},
  year = {2021},
  doi = {10.1109/access.2021.3058149},
}

Loading related content...

Figure