Frequently Asked Questions

X-ray inspection fundamentals
AI denoising models, applications and validation
System investment and detector compatibility
Real-time inspection and system performance
Future developments in X-ray inspection

Answers to common questions from our recent webinar on industrial X-ray inspection, detector selection and AI-powered image denoising.

Responsive Padding

Watch the webinar

X-ray inspection fundamentals

This section introduces the key principles that influence inspection performance, including X-ray penetration, energy selection, material properties and image quality considerations.

1. How deep can X-rays penetrate during inspection?

There is no single maximum penetration depth for X-rays. X-ray penetration depends on the material density, thickness, and the X-ray energy used. In general, higher X-ray energies provide greater penetration, while lower energies are preferred when inspecting less dense materials and when higher image contrast is required.

The table below shows typical energy ranges used for different materials and inspection targets:

Material / Product	Typical X-ray energy
Food products, packaged food, organic materials	20 - 100 kV
Plastics, polymers, rubber parts	20 - 80 kV
PCB assemblies, solder joints, electronic components	60 - 130 kV
Semiconductor packages	80 - 160 kV
Lithium-ion batteries	100 - 180 kV
Aluminum castings and light alloys	100 - 225 kV
Steel components and welded assemblies	180 - 320 kV
High-density metal parts and large industrial components	225 - 450 kV+

Tube voltage X-ray transmission

Fig.1: As the X-ray energy increases with the increase in tube voltage, the X-ray transmission to the object also increases. A high-contrast X-ray image can be obtained by setting a tube voltage value suitable for the object.

UVTRON operating principle step 1

Focal spot size X-ray image transition

For application-specific recommendations, representative sample testing under real inspection conditions is always recommended. We can carry out tests for you in our demo facilities.

Click here to request a demo

AI denoising models, applications and validation

Learn how AI denoising is applied to X-ray inspection, the neural network technology behind it, and the methods used to verify performance and reliability.

2. Can AI denoising be applied to PCB inspection applications?

Yes, AI denoising can be applied to PCB inspection applications, particularly to improve the visibility of small defects or low-contrast features when images are acquired under noisy conditions. The effectiveness of AI denoising depends on several factors, including X-ray energy, dose, object structure, defect size, defect contrast, and inspection speed.

The commercially available AI denoising dynamic-link library (DLL) is currently compatible with the Hamamatsu C14300 series X-ray line scan cameras. At present, it is not commercially available for TDI-CCD camera platforms. We are considering expanding it for flat panel detectors.

Before deployment in a production environment, we recommend validating performance using representative samples and real inspection conditions.

3. What type of AI model is used for image denoising?

Hamamatsu uses proprietary convolutional neural network (CNN) architectures specifically developed for X-ray image denoising.

These models have been optimized to reduce image noise while preserving fine structural details that are critical for inspection applications. In addition to the network architecture itself, particular attention is given to the training process and data generation methods.

The models are trained using data that accurately reproduces the physical behavior of X-ray imaging systems and their associated noise characteristics. Detailed information regarding the network architecture and training conditions remains proprietary.

4. Can AI denoising create false defects or hide existing defects?

As with any AI-based image processing technology, the possibility of image alterations cannot be eliminated, particularly when operating outside the conditions used during training.

To minimize this risk, Hamamatsu trains its AI denoising models using simulations based on the physical characteristics of X-ray imaging systems and validates performance using standard image quality metrics such as Signal-to-Noise Ratio (SNR), Photon Transfer Curve (PTC) analysis and spatial resolution.

Since inspection performance depends on the application, defect type, imaging conditions and inspection algorithms, we recommend validating AI denoising with representative samples and real production conditions before deployment.

System investment and detector compatibility

Explore the factors that influence return on investment and understand how detector operating limits affect system selection.

5. When does X-ray inspection make economic sense for low-cost products?

The decision to implement X-ray inspection should be based on the overall cost of quality rather than product cost alone.

X-ray inspection may make economic sense when the potential cost of defects, contamination, field failures, or product recalls exceeds the cost of inspection. For high-volume production, even low-cost products can justify inspection where defects create significant downstream costs.

Cost and performance can be optimized by considering the complete imaging chain, including the X-ray source, detector and image processing.

6. What are the X-ray energy compatibility limits of Hamamatsu detectors?

Hamamatsu detector platforms currently support maximum X-ray energies from 160 kV to 180 kV, depending on the detector technology and sensor design. The table below lists the maximum supported X-ray energy for each platform.

Detector	Maximum X-ray energy
CCD-TDI (C12300 series)	Up to 180 kV
Single-line line scan cameras (C14300 / C14960 series)	Up to 160 kV
TDI-CMOS detectors	Up to 160 kV
Flat panel sensor C16401SK-51 (12 cm × 14 cm)	Up to 160 kV

These energy ratings are intended as general compatibility guidelines. The optimum detector choice also depends on factors such as material density, thickness, required image quality, and inspection speed. Hamamatsu is continuously expanding its X-ray detector portfolio, and detector solutions compatible with 225 kV systems are currently under development.

Real-time inspection and system performance

Discover how acquisition speed, detector technology and image processing work together to achieve reliable inspection results in production environments.

7. Does AI denoising work in real time, or only on acquired images?

The AI denoising dynamic-link library (DLL) can process both acquired images offline and images inline within an inspection system. Whether inline processing operates in real time depends on multiple factors, including image size, acquisition rate, PC configuration, GPU performance, and application requirements. As a result, processing time cannot be defined as a single fixed value.

For real-time inspection applications, it is important to verify that the AI denoising process can keep pace with the camera acquisition rate within the target hardware environment. Performance validation should therefore be carried out using the intended system configuration.

8. Which matters most in an X-ray NDT system: the source, detector, or image processing?

No single element matters most in isolation. Inspection performance depends on how well the entire imaging chain is matched to the application.

The source influences penetration and image sharpness, the detector affects sensitivity, resolution and speed, while image processing helps maximize the value of the acquired data. The best results are achieved when all three are selected together based on the inspection objective.

9. How do you choose the right resolution for an inspection?

Choose the resolution needed to reliably detect the target feature or defect. Higher resolution is not always better: focal spot size, detector pixel size, magnification, field of view and inspection speed must be considered together to achieve the required detectability while maintaining a practical and repeatable inspection process.

Fig. 2: This focal-spot-size and magnification image can illustrate how these factors affect the resolution available for inspection.

Focal spot size X-ray image transition

10. Why choose TDI-CMOS over a standard line-scan detector for high-speed inspection?

TDI-CMOS is particularly beneficial for high-speed inspection of moving objects when the available x-ray signal is limited.

Time delay integration (TDI) accumulates the signal over multiple synchronized stages, improving the signal-to-noise ratio without reducing throughput. It is often the preferred choice when higher image quality is required under demanding production-line conditions.

Fig. 3: This focal-spot-size and magnification image can illustrate how these factors affect the resolution available for inspection.

Responsive Padding

Focal spot size X-ray image transition

Future developments in X-ray inspection

An overview of emerging technologies and industry trends that are expected to shape the next generation of inspection systems.

11. Which area will advance most in the next five years: sources, detectors, or AI?

Rather than one technology outpacing the others, the most significant advances are likely to come from the combined evolution of X-ray sources, detectors and AI-based image processing. As these technologies become increasingly integrated, they will enable more optimized and application-specific inspection solutions. Hamamatsu continues to invest in innovation across its portfolio, and several developments are currently underway.

If you have a question that’s not covered here, our NDT specialists can help.