Online program Overview Session: RTSD-08

Abstract

This summary reviews the evolution of the Timepix family of hybrid pixel detector readout ASICs, from Timepix in 2006 through to the recently released Timepix4 and the forthcoming Picopix demonstrator. The early versions had a regular matrix of 256x256 pixels. Each generation introduced an architectural innovation: per-pixel programmable Time-of-Arrival (ToA) or Time-over-Threshold (ToT) mode selection (Timepix, 250 nm MOS, 2006, Medipix2 Collaboration); data-driven triggerless readout with simultaneous time and energy measurement (ToA/ToT) and a 1.56 ns on-pixel TDC (Timepix3,130 nm CMOS, 2013, Medipix3 Collaboration); simultaneous ToA/ToT with logarithmic gain mode, improved energy resolution, and dead-time-free continuous read/write (Timepix2,130 nm CMOS, 2018, Medipix2 Collaboration); and 195 ps ToA binning over a four-side buttable 448x512 matrix using through-silicon via (TSV) technology (Timepix4, 65 nm CMOS, 2019, Medipix4 Collaboration). The forthcoming Picopix (28 nm CMOS, submission foreseen in 2026) targets ∼35 ps ToA binning with on-pixel clustering, on-chip time calibration, and packet sorting. Key quantitative parameters—timing resolution, hit rate, active area, power consumption, and readout bandwidth—are compared across all five generations. Representative applications and the knowledge-transfer impact of the Medipix Collaborations model are discussed.

Abstract

Timepix4 is the latest generation of the readout ASIC developed by the Medipix4 Collaboration for hybrid pixel detector technology, featuring a 448 x 512 matrix with a 55 μm pitch. Operating in data-driven mode, it provides simultaneous Time of Arrival (ToA) and Time over Threshold (ToT) information for each hit. This allows for multi-pixel signal clustering to mitigate charge sharing and fluorescence escape, while calibrated ToT values provide per-pixel energy measurement. Such capabilities enable high-resolution X-ray spectral imaging, providing accurate energy deposition information for each detected event and greater post-processing flexibility compared to traditional threshold-based photon-counting systems. In this work, we present the first full-spectral micro-CT images acquired with a Timepix4-based detector operating in data-driven mode. The detector is equipped with a 1 mm-thick CdTe sensor, making it suitable for X-ray energies commonly used in biomedical applications. Tomographic scans were performed at the INFN PEPI laboratory using a microfocus X-ray tube (80 kV, 20 μA). The study evaluated a custom plastic phantom containing three distinct contrast agents: silver, iodine, and gadolinium. Projections were acquired over 360° with a reconstructed voxel size of 33 μm. The resulting 3D attenuation spectra were processed using a basis material decomposition algorithm, producing independent 3D density maps for each contrast agent. The results confirm the feasibility of full-spectral micro-CT imaging using the Timepix4 ASIC, demonstrating its ability to capture 3D attenuation spectra at high spatial resolution. While data rate management remains a challenge, the ability to resolve full attenuation spectra in 3D enables precise multi-material identification in a single scan. This capability offers significant potential for advanced applications in biomedical research and materials science.

Abstract

X-ray imaging systems are widely used in industries such as homeland security, non-destructive testing, archaeological research, and medicine, where accurate material identification is often essential. Examples include explosives screening in airport luggage, product quality testing on a production line, and content characterization in historical manuscripts. Photon-counting detectors provide high spatial and energy resolution, along with full spectroscopic information on a per-event basis, and are therefore well suited for machine learning (ML) approaches to material identification. One example of such a detector is the Timepix3, which records single events at fluxes up to 40 Mhits/cm²/s. Previous work, presented at this conference, demonstrated the feasibility of training an ML classifier (Gradient-Boosted Trees) on measured X-ray spectra acquired with a 12x1 Timepix3 (bump-bonded to 1 mm CdTe) detector and reported initial identification results on a multi-material phantom. This contribution extends that work, addressing two limitations of its results: the restricted size of the experimental material database and the lack of quantitative model diagnostics. The database is supplemented with simulated spectra (generated with the Allpix Squared framework) for materials that are difficult to obtain as pure reference samples, and the consistency between simulated and measured data is assessed. The extended database is then used to retrain the classifier, and standard ML diagnostics are reported to characterize model behavior. The performance of this approach will be compared to classical techniques such as base-material decomposition.

Abstract

This contribution extends the technique of the fast energy dispersive X-ray diffraction (XRD) in transmission mode for inspection of macroscopic objects, as described in our previous work [1], published recently by Nature Scientific Reports. This technique utilizes the event-based registration of diffracted X-ray photons with simultaneous timing and energy information as provided by Timepix3/4 detectors. It enables rapid diffraction measurements with a polychromatic X-ray source, eliminating the need for mechanical angular scanning while preserving spatially resolved structural information.

Current XRD systems utilize relatively low-energy monochromatic X-ray beam (5-20 keV) which can only penetrate shallow layer close to the sample surface. Therefore, thick or dense parts cannot be analyzed through their entire volume using standard approaches. An alternative is the use of synchrotron sources, however, these are generally impractical for industrial applications.

A hard polychromatic X-ray source (X-ray tube) in combination with spectrally sensitive imaging detector such as Timepix3 or Timepix4 equipped with high-Z sensor (CdTe or CZT) overcomes above listed limits. This work focuses on optimizing detector performance by suppressing unwanted effects such as internal X-ray fluorescence and charge-sharing. Both of these effects degrade energy resolution, which directly impacts the quality of recorded XRD data. It will be demonstrated that an optimized detector, together with an event-sorting algorithm, can provide high-quality XRD data within a fraction of a second. Such speed enables 2D scanning across the entire volume of the inspected sample.

[1] Vavrik, D., Georgiev, V., Jakubek, J. et al. Transmission energy dispersive X-ray diffraction as a tool for the laboratory study of fast processes in metals. Sci Rep 15, 31752 (2025). https://doi.org/10.1038/s41598-025-16314-9

Abstract

Grating-based phase-contrast imaging (gbPCI) is a powerful tool for non-destructive testing and material analysis. In conventional X-ray imaging, only the attenuation of the X-rays is measured, resulting in low contrast between structures with similar absorption characteristics. With gbPCI, it is possible to not only get information about the absorption, but also the phase-shift of X-rays when propagating through material. For non-destructive testing and material analysis, the dark-field image provides complementary information, since it is sensitive to structures below the resolution limit of the system, which is primarily given by the pixel sizes of the detector. The pixel pitch is typically 50 μm or higher. With the dark-field image, it is possible to see porous or fibrous structures and estimate their sizes. A photon-counting and energy-resolving Timepix4 detector is used in this work. By setting energy cuts, the sensitivity range can be maximized, allowing a more detailed structural analysis. Additionally, the quality of the imaging system can be improved by cutting unwanted energies. First test measurements with Timepix4 have shown the feasibility of phase-contrast imaging with this detector, by acquiring energy-resolved dark-field images of lycopodium spores and a piece of wood.