TimepixCluster - Efficient Algorithms and Pipeline for Event-Driven Imaging

Jochen Küpper, DESY

Peer Stelldinger, HAW Hamburg

We developed a new algorithm for the reduction of 4D data from event-driven detectors, which promises important applications of, e.g., Timepix detectors, both for in-lab and facility-based experiments and well as for commercial use. The current algorithms need to be enhanced, adapted, and implemented for real-time use, which is especially crucial for practical application and faster data processing with increasing data volumes from advanced light sources and new detectors.

Timepix4 is an innovative imaging sensor recently developed, in a collaboration at CERN and including DESY, offers independent data streaming from each pixel, with event rates reaching 160 GBit/s, without being constrained by a specific frame rate. Compared to conventional commercially available camera technologies, e.g., CCD or sCMOS, with a similar sensor size, the Timepix4 camera offers much better time resolution on the order of 200 ps and significantly larger data rates up to 160 GBit/s. This emphasizes the need for increasingly efficient processing algorithms.

Event-driven detectors, such as Timepix4, offer several advantages over conventional cameras: The data recorded inherently contains less noise and it provides much better time resolution. However, the unique nature of event-driven data requires appropriately adjusted handling and analysis strategies. We focus on developing and improving efficient algorithms and a processing pipeline specifically designed for such event-driven imaging data with the goal of enabling enhanced real-time recording and analysis capabilities. To address the challenge of the high data rate, we aim to implement the customized clustering algorithm in hardware, e.g., FPGA. Our approach is scalable for multi-chip modules, meeting the demands of advanced experiments. We plan to protect this development through a patent application, also foreseeing important commercial applications. Moreover, for a timely interpretation of the event-based image data and informed operation of experiments, it is crucial to develop novel real-time-visualization techniques with a customized user interface. This will allow users to quickly obtain feedback for optimizing their experiment while collecting the data.