Timepix4: 4D Tracking & Hybrid Pixel ASIC

• Timepix4 Chip is a high-performance hybrid pixel detector that integrates advanced timing, energy measurement, and 4D tracking functionalities.
• It features a fine-pitch 448×512 pixel matrix with per-pixel TDC providing sub-nanosecond resolution and data-driven readout supporting hit rates up to 2.5 Ghits/s.
• Its flexible architecture and TSV technology enable versatile applications in particle physics, medical imaging, and quantum optics with sub-micron spatial precision.

The Timepix4 chip is a hybrid pixel detector developed for high-performance particle tracking, timing, photon science, medical imaging, and a range of high-rate, high-precision detection applications. Designed by the Medipix Collaboration and CERN, Timepix4 features a fine-pitch pixel matrix, per-pixel time and energy measurement, advanced digital and analog front ends, and enables 4D-tracking by providing both high-resolution spatial and temporal information. Its flexible architecture, data-driven readout, through-silicon-via (TSV) technology, and versatile configuration allow deployment across diverse scientific domains requiring sub-micron spatial resolution and sub-nanosecond timing resolution.

1. Architecture, Pixel Matrix, and Signal Processing

The Timepix4 ASIC is fabricated in 65 nm CMOS technology. It consists of a 448 × 512 pixel matrix (some variants report up to 512 × 448), with each pixel having a 55 μm pitch, resulting in a total sensitive area of approximately 7 cm² (Alozy et al., 2021, Cavallini et al., 3 Mar 2025). Each pixel comprises:

• A charge-sensitive preamplifier (with threshold DACs for adjustment)
• A discriminator triggered by the leading edge of the analog pulse
• Per-pixel Time-of-Arrival (ToA) and Time-over-Threshold (ToT) digitization via an integrated time-to-digital converter (TDC)
• Event-driven digital logic for zero-suppressed data output

Timepix4’s unique design allows simultaneous measurement of ToA and ToT for each hit, supporting both photon counting and high-rate event time-stamping (Correa et al., 10 Apr 2024). The analog front end achieves an equivalent noise charge (ENC) of about 50–70 e⁻ (Alozy et al., 2021), and the TDC features a nominal time binning of 195 ps per pixel (Heijhoff et al., 2022).

2. Timing, Spatial Resolution, and Calibration

Pixel timing relies on a multi-tiered digitization scheme involving a 40 MHz system clock, a 640 MHz Voltage-Controlled Oscillator (VCO), and phase subdivision for ultra-fine digitization. Calibrated per-VCO TDC correction reduces intrinsic time resolution to 58–62 ps (Heijhoff et al., 2022). Analog front-end timing is sensor-dependent: for a bare chip in electron-collecting mode, the best measured time resolution is 47 ps at high injected charge; with a bonded sensor, this degrades to 62 ps (Heijhoff et al., 2022, Bolzonella et al., 23 Apr 2024). Cluster-wise oversampling (combining ToA and ToT from several pixels) can reduce event timing to 33 ps r.m.s. (Bolzonella et al., 23 Apr 2024).

Spatial resolution is governed by pixel pitch and charge sharing. In silicon tracking configurations, intraplane residuals of 14–15 μm in thin sensors and 3–4 μm in thick sensors have been demonstrated (Akiba et al., 2022, Akiba et al., 19 Mar 2025). In X-ray and photon imaging, clustering using ToT-weighted centroids achieves 5–10 μm position resolution on the anode (Alozy et al., 2021). For transmission electron microscopy, temporal and amplitude clustering can double or triple the Modulation Transfer Function (MTF) at Nyquist frequency (Dimova et al., 25 Nov 2024).

Timewalk, inherent to leading-edge discrimination, restricts resolution, but per-pixel ToT-based corrections permit sub-100 ps timing (Heijhoff et al., 2022, Akiba et al., 19 Mar 2025). Calibration functions typically take the form:

$\text{Timewalk}(q) = a/(q + b)^c + d$

where $q$ is the measured charge.

3. Data Acquisition, Readout, and Online Processing

Timepix4 supports two major readout architectures:

• Data-driven mode: Each pixel outputs a 66-bit event word upon activation, suitable for sparse, asynchronous signals with up to $2.5\,\text{Ghits/s}$ chip throughput via up to 16 optical links at $10.24\,\text{Gbps}$ per link, aggregating to $160\,\text{Gbps}$ bandwidth (Alozy et al., 2021, Cavallini et al., 3 Mar 2025).
• Frame-based mode: Per-pixel counters accumulate over set shutter intervals, yielding rates up to $40\,\text{kfps}$ (Correa et al., 10 Apr 2024).

Acquisition frameworks such as DataPix4 (C++), SPIDR4 (FPGA), and custom FPGA-based DAQ (often using Xilinx Ultrascale or Zynq 7000) interface with the ASIC for chip configuration, slow control, high-speed UDP communication, buffering, and real-time clustering (Cavallini et al., 3 Mar 2025, Akiba et al., 19 Mar 2025, Alozy et al., 2021).

Online clustering utilizes spatial and temporal proximity criteria, generally of the form $|t_2-t_1|\leq \Delta t_{\text{hits}}$, to group related hits, outputting ROOT files or transmitting data for visualization and rapid experiment feedback (Cavallini et al., 3 Mar 2025).

4. Sensor Integration, TSV Technology, and Embedding in Hybrid Detectors

The Timepix4 chip has been engineered for extreme integration flexibility:

• It is bump-bonded to silicon sensors (standard, thin, LGAD, iLGAD, or fully depleted optical sensors) or coupled to microchannel plate (MCP) electron multipliers and transmission photocathodes (Oppenhuis, 11 Sep 2025, Alozy et al., 2021).
• Through Silicon Vias (TSVs) permit four-side buttable tiling with backside connectivity, eliminating dead regions and allowing large-area detectors to be constructed without wire bonds (Sriskaran et al., 2023, Scharenberg et al., 22 Dec 2024, Scharenberg et al., 16 Mar 2025).
• Embedding directly into micro-pattern gaseous detectors (MPGDs) via lamination in polyimide foils or silicon readout boards (using TSV back-contacts) enables hybrid pixel/gas amplification device architectures with ultra-low material budgets and high spatial granularity (Scharenberg et al., 22 Dec 2024, Scharenberg et al., 16 Mar 2025).

Mechanical tests (chip thinning to ~130 µm; ACP bonding) and electrical tests (slow control register readback, gain mapping) confirm feasibility of these integration paths (Scharenberg et al., 16 Mar 2025).

5. Application Domains and Experimental Results

Timepix4 has demonstrated versatility across experimental domains:

• Particle tracking telescopes: Achieve pointing resolutions of $2.3\pm0.1~\mu\text{m}$ and temporal resolutions of $92\pm5~\text{ps}$ for tracks by combining measurements from multiple sensor planes (Akiba et al., 19 Mar 2025, Akiba et al., 2013).
• Single-photon, X-ray, and optical detection: Provide event-based, time-stamped imaging with sub-nanosecond precision and high spatial resolution; sustain hit rates $>10^9~\text{photons/s}$ (Alozy et al., 2021, Hogenbirk et al., 18 Sep 2025).
• Transmission electron microscopy: Apply temporal and energy clustering to improve spatial resolution and partially correct blurring due to extended electron trajectories (Dimova et al., 25 Nov 2024).
• MPGDs and hybrid detectors: Enable direct charge collection in GEM, µRWell, and related architectures, unlocking new possibilities for X-ray polarimetry (sub-$2~\text{keV}$ regime), rare-event searches, and ultra-low material budget tracking (Scharenberg et al., 22 Dec 2024, Scharenberg et al., 16 Mar 2025).
• High-luminosity collider contexts: When coupled to iLGAD sensors, measured time resolution reaches $377\pm7~\text{ps}$ (perpendicular) and $359~\text{ps}$ (grazing); sensor efficiency is $99.6\pm0.1\%$ (Oppenhuis, 11 Sep 2025).

Experimental validation includes SPS North Area beam tests (charged particle tracking; time stamping efficiency $>98\%$; rates up to $15~\text{kHz}$), X-ray and nuclear resonance scattering experiments at PETRA III and ESRF (resolving bunch structures; time resolutions down to $8.5~\text{ns}$), and laboratory measurements using picosecond lasers (one-pixel timing $107~\text{ps}$; cluster averaging $33~\text{ps}$) (Akiba et al., 2013, Bolzonella et al., 23 Apr 2024, Correa et al., 10 Apr 2024).

6. Signal Corrections, Algorithms, and Data Analysis Techniques

Refined performance requires systematic application of corrections and sophisticated algorithms:

• Charge calibration: ToT values are mapped to deposited charge via surrogate functions, e.g. $ToT(q) = p_0 + p_1 q - \frac{p_2}{q-p_3}$ (Akiba et al., 19 Mar 2025).
• Timewalk correction: Per-pixel and global ToT-based corrections employ empirical forms (e.g., $\Delta t_{\text{timewalk}} = a/(q+b) + c$) (Oppenhuis, 11 Sep 2025, Hogenbirk et al., 18 Sep 2025).
• Per-pixel VCO correction: Accounts for local oscillator frequency variation to refine timestamping (Heijhoff et al., 2022, Bolzonella et al., 23 Apr 2024).
• Spatial (η) correction: Polynomial fits (typically third-order) compensate for charge sharing and pixelation bias in cluster position estimation (Akiba et al., 19 Mar 2025).
• Track and vertex reconstruction: Techniques such as uncertainty-weighted averaging, singular value decomposition, and DBSCAN clustering are used to extract particle trajectories, annihilation vertices, and reject outliers (Kraxberger et al., 16 Aug 2025).

7. Future Directions and Impact

The Timepix4 chip and its derivative technologies (including Medipix4) continue to evolve towards larger arrays, finer pixelation, faster timing, lower material budgets, and seamless sensor integration (Sriskaran et al., 2023). Anticipated advances relate to wafer-scale monolithic pixel detectors, TSV and silicon-readout board deployments, improved sensor design (e.g., thinner LGADs for $\lesssim50~\text{ps}$ time resolution), and fully integrated, FPGA-enabled real-time clustering for high-throughput environments.

Its impact spans:

• Precision particle physics (vertexing, timing-layer trackers at HL-LHC)
• Single-photon, quantum optics, and ultrafast optical imaging (Hogenbirk et al., 18 Sep 2025)
• Biomedical X-ray, electron microscopy, and material analysis (Dimova et al., 25 Nov 2024)
• Gas-based rare-event and X-ray polarimetry detectors (Scharenberg et al., 22 Dec 2024)
• General-purpose scientific instrumentation, supported by flexible software such as DataPix4 (Cavallini et al., 3 Mar 2025)

Timepix4’s capacity for sub-micron spatial and sub-nanosecond timing resolution, sustained at high hit rates with minimal dead area or material, underpins its role as a benchmark hybrid pixel ASIC for high-performance detectors in academic and industrial research.