Q-Pix: Pixelated Charge Readout for TPCs

• Q-Pix is a pixelated charge readout architecture that encodes fixed charge increments into reset timings, providing precise 3D imaging in TPCs.
• Its Charge-Integrate/Reset process measures the time needed to accumulate a fixed charge, yielding average current and total collected charge without continuous sampling.
• The system features local clocks, dynamic data networks, and low-leakage electronics, making it scalable for kiloton-scale detectors with high reliability.

Searching arXiv for recent Q-Pix papers and related terminology to ground the article. arXiv search query: Q-Pix liquid argon time projection chamber reset time difference pixel readout Q-Pix is a pixelated charge readout architecture for gas and liquid time projection chambers, especially kiloton-scale liquid argon TPCs, built around a self-triggering Charge-Integrate/Reset process in which each pixel records the times at which a fixed amount of charge has accumulated rather than sampling charge as a function of time. In this “time-to-charge” or “time-per-charge” scheme, reset time differences encode the average current and reset count encodes total collected charge, while local clocks and dynamically established data networks are intended to make the system low power, intrinsically sparse in data output, scalable to very large channel counts, and resilient against single-point failure (Nygren et al., 2018). The concept has progressed from architectural proposal to discrete front-end demonstrations, a 16-channel multi-pixel prototype in a gas TPC, and detector-level studies for supernova and solar-neutrino physics (Miao et al., 2023, Hoch et al., 2024, Kubota et al., 2022, García-Peris et al., 21 Jul 2025).

1. Detector motivation and historical emergence

Q-Pix was proposed for kiloton-scale liquid argon TPCs, especially the DUNE far detector, in response to a conjunction of detector constraints: very large channel count, need for detailed imaging, inaccessibility of in-argon electronics, reliability requirements, low power dissipation, and the need for calibration everywhere in the detector volume (Nygren et al., 2018). For a $4~\text{mm} \times 4~\text{mm}$ pixelization, the 2018 concept paper states that more than $10^8$ channels per 10 kTon module are implied, and later detector studies describe a 10-kton module with roughly 172 million channels in a Q-Pix design space (Nygren et al., 2018, Kubota et al., 2022).

The architecture was framed against limitations of conventional wire readout in large LArTPCs. Multiple planes of long sensing wires provide 2D projections that must be combined with timing to reconstruct 3D events; this is a proven method, but it can suffer from ambiguities in dense or overlapping topologies, from high capacitance associated with long wires, from large continuous data volume, and from substantial engineering complexity (Kubota et al., 2022). Pixelated readout removes many of the geometric ambiguities because charge is localized directly in three dimensions, but conventional pixel digitization can generate high data rates. Q-Pix addresses that tension by encoding charge into reset timing, so only local reset timestamps need to be stored and transmitted (Hoch et al., 2024).

This measurement philosophy was therefore introduced not primarily as a frontend variant of standard waveform digitization, but as a different detector-electronics abstraction: useful events are rare, yet when they occur they can be spatially and temporally complex, so the readout should remain quiescent when no charge is present and become active only when charge arrives (Nygren et al., 2018). In later application studies, this same feature is presented as especially well matched to low-energy, low-rate, background-dominated signals such as supernova and solar-neutrino interactions (Kubota et al., 2022, García-Peris et al., 21 Jul 2025).

2. Charge-Integrate/Reset principle and signal representation

At the core of Q-Pix is a Charge-Integrate/Reset (CIR) loop. Incoming detector current is integrated on a feedback capacitor until a comparator or Schmitt trigger threshold is reached; a reset or replenishment action then returns the channel to baseline or injects a compensating fixed charge, and the time of that event is recorded (Nygren et al., 2018, Miao et al., 2023). The essential datum is the reset time difference (RTD), defined by sequential subtraction of timestamps from successive reset events. For approximately constant current,

$I \cdot \Delta T = \int I(t)\,dt = \Delta Q,$

so the average current is inversely related to RTD, $I \propto 1/\mathrm{RTD}$, while the number of resets is proportional to the total collected charge (Nygren et al., 2018, Hoch et al., 2024).

This is the inversion of conventional waveform digitization emphasized throughout the Q-Pix literature. Conventional ADC-based systems sample amplitude at fixed time intervals; Q-Pix samples time at fixed charge increments. In the gas-TPC prototype, the authors state explicitly that the method measures the time required to accumulate a fixed amount of charge at each pixel, and that shorter RTD means larger average current, while longer RTD means smaller average current (Hoch et al., 2024). In the discrete front-end study, the operational law is written as

$I = \Delta Q \cdot f + I_\mathrm{o},$

where $f$ is the reset pulse frequency and $I_o$ is leakage current; for fixed $\Delta Q$, measuring reset frequency directly gives the current (Miao et al., 2023).

Two circuit realizations appear in the hardware development literature. The simpler form is a reset scheme in which a charge-sensitive amplifier integrates current onto $C_f$, a Schmitt trigger fires when threshold is reached, and a MOSFET rapidly discharges $C_f$ (Miao et al., 2023). The demonstrated board-level implementation ultimately favored a replenishment scheme, in which the MOSFET acts as a controlled current source that replenishes a fixed amount of charge when the threshold is reached. The stated reason is that hard-resetting can cause charge loss and nonlinear errors if input charge arrives during the reset interval, whereas replenishment allows the system to continue integrating input current while the MOSFET injects a fixed compensating charge (Miao et al., 2023).

The front-end paper also emphasizes a close resemblance between the replenishment version of Q-Pix and a first-order Sigma-Delta modulator. In that mapping, the CSA is the integrator, the Schmitt trigger is the quantizer, the MOSFET replenishment current source acts like a one-bit DAC, and the reset pulses form the oversampled bitstream. A direct consequence is that digital signal processing methods used for Sigma-Delta systems, especially CIC filters, can be applied to Q-Pix waveform reconstruction and noise shaping (Miao et al., 2023).

3. System architecture, timing, and scaling model

Q-Pix is not only an analog front-end concept; it is also a detector-scale readout architecture. The original proposal places a free-running local clock on each ASIC rather than distributing a system-wide master clock, with local oscillators expected to run at roughly 50–100 MHz (Nygren et al., 2018). The reset time is captured locally and later calibrated to a global reference. In the supernova study, a 50 MHz clock with a 32-bit timestamp is stated to provide about 43 s before wrap-around, and the expected timing precision is sufficient to obtain about $10^8$0 global timing accuracy, corresponding to roughly 1.6 mm in the liquid-argon drift direction at nominal drift speed (Kubota et al., 2022).

The data network is likewise decentralized. The 2018 design describes dynamically established data networks in which a timing or readout token is injected at the tile boundary and propagated through neighbor-to-neighbor links; each ASIC remembers from whom it accepted the token, and the path can reform around failed chips (Nygren et al., 2018). The resulting electronic architecture is described as flat and self-guided, with high resilience against single-point failure (Nygren et al., 2018).

Scaling is organized through tiles of $10^8$1 ASIC chips. An illustrative geometry in later detector studies uses a $10^8$2-pixel tile, a $10^8$3 pixel pitch, and tile size of order 625 cm$10^8$4 (Kubota et al., 2022). In the solar-neutrino study, a kiloton-scale underground LArTPC is modeled with four drift volumes of 3.5 m each and the innermost anode plane instrumented by contiguous Q-Pix pixels with the same $10^8$5 pitch (García-Peris et al., 21 Jul 2025). The paper describes the Q-Pix scheme there as combining zero suppression, self-triggering pixels, dynamically established data networks, a charge-sensitive amplifier with feedback capacitor $10^8$6, and thresholded reset at a charge increment $10^8$7; only pixel indices and timestamps are stored, and the readout remains continuously active with no fixed acquisition window (García-Peris et al., 21 Jul 2025).

This systems picture is inseparable from the claimed detector advantages. Q-Pix is repeatedly presented as a true 3D imaging readout with sparse data capture, low power, and scalability to very large channel counts, rather than merely a pixel ASIC with unusual pulse processing (Nygren et al., 2018, García-Peris et al., 21 Jul 2025).

4. Front-end demonstrations and prototype measurements

The first hardware demonstrations focused on establishing that the CIR frontend could be realized with practical electronics. The 2023 discrete demonstration used a Texas Instruments LMP7721 CMOS operational amplifier as the CSA, an Analog Devices LT1713 comparator configured as a Schmitt trigger, an external 4-terminal NMOS transistor wire-bonded to the PCB for reset or replenishment, and an FPGA to generate a clocked reset pulse (Miao et al., 2023). The board was engineered for ultra-low leakage with a guarded triple coaxial connector, mechanical PCB grooves, and removal of solder mask above the input region (Miao et al., 2023). In that demonstrator, the effective charge quantum was measured to be $10^8$8 fC, the reported leakage current was as low as about 3 pA, and the measured board showed excellent linearity down to a few picoamperes (Miao et al., 2023). The same paper reports that a TSMC 180 nm NMOS transistor could be biased in the subthreshold region down to roughly $10^8$9 fA, limited by measurement precision, and that the measured behavior was consistent with SPICE simulation, including CIC-based waveform reconstruction and first-order Sigma-Delta-like noise shaping (Miao et al., 2023).

A more detector-like validation followed in the 2024 multi-channel gas-TPC prototype. That work developed a 16-channel Simplified Analog Q-Pix prototype built from commercial off-the-shelf components, with each channel implemented as an IVC102 precision switched integrator followed by an AD8561 comparator (Hoch et al., 2024). The front end was operated with $I \cdot \Delta T = \int I(t)\,dt = \Delta Q,$0 pF and threshold around 1 V, corresponding to about 10 pC, or

$I \cdot \Delta T = \int I(t)\,dt = \Delta Q,$1

while the digital back-end used a Zynq-7000 FPGA SoC on a Digilent Zybo-Z7-20, timestamped reset pulses with a 30.3 MHz clock, sent data via UDP, and handled configuration via TCP (Hoch et al., 2024).

The anode geometry in that prototype consisted of 16 concentric annular electrodes patterned on a custom PCB, chosen because only 16 channels were available and the ionization source was known and near the center (Hoch et al., 2024). A gold photocathode at the cathode was illuminated by a pulsed UV xenon flashlamp delivered through an optical fiber, producing photoelectrons that drifted through P-10 gas (90% Ar, 10% CH$I \cdot \Delta T = \int I(t)\,dt = \Delta Q,$2) in a uniform field (Hoch et al., 2024). In the hardware, the integrator output was divided by four before comparison, and the reset time was set to $I \cdot \Delta T = \int I(t)\,dt = \Delta Q,$3; the reported dead time was negligible, less than 0.1% of the time between resets (Hoch et al., 2024). After channel-by-channel calibration, the prototype reconstructed the radial charge profile of the electron swarm and measured transverse diffusion that agreed with PyBoltz simulations across roughly 200–1500 Torr (Hoch et al., 2024). The authors’ conclusion was that a Q-Pix readout can successfully reconstruct the ionization topology in a TPC from reset timing information alone (Hoch et al., 2024).

5. Physics reach in low-energy neutrino detection

The most developed detector-physics case for Q-Pix concerns low-energy neutrinos in large LArTPCs. For core-collapse supernovae, the 2022 study argues that Q-Pix’s intrinsically 3D readout offers higher reconstruction efficiency, lower energy threshold, considerably lower data rates, and potential pointing information relative to a wire-based readout (Kubota et al., 2022). Using a 10-kpc core-collapse supernova with the Garching electron-capture model propagated through SNOwGLoBES, the study reports 220 $I \cdot \Delta T = \int I(t)\,dt = \Delta Q,$4 charged-current events and 19 $I \cdot \Delta T = \int I(t)\,dt = \Delta Q,$5-electron elastic-scattering events in a 10-kton LAr module (Kubota et al., 2022). The benchmark reset threshold is 1 fC, corresponding to 6250 electrons, or about 147.5 keV in liquid argon (Kubota et al., 2022). For low-energy event identification, the optimal clustering scale is reported as a $I \cdot \Delta T = \int I(t)\,dt = \Delta Q,$6 time window between resets and a threshold of 13 resets per cluster; with that choice, the paper finds purity > 95% at about 80% efficiency for events with $I \cdot \Delta T = \int I(t)\,dt = \Delta Q,$7 MeV deposited energy, and purity > 99% at about 88% efficiency across the full supernova spectrum (Kubota et al., 2022).

The same study emphasizes the data-rate implications. For a 10-kton Q-Pix module, the estimated rate at a threshold of 1 reset is 5.7 MB/s for the full module, and requiring 7 resets reduces the rate by two orders of magnitude (Kubota et al., 2022). At a 10 MeV neutrino-energy threshold—equivalent in the paper to about 5 MeV deposited energy or 34 resets—the annual data volume is quoted as $I \cdot \Delta T = \int I(t)\,dt = \Delta Q,$8 PB/year for Q-Pix versus $I \cdot \Delta T = \int I(t)\,dt = \Delta Q,$9 PB/year for DUNE projective readout, i.e. roughly a six-order-of-magnitude reduction (Kubota et al., 2022). For burst triggering, a DBSCAN clustering with maximum $I \propto 1/\mathrm{RTD}$0, minimum cluster size of 14 resets, and a final requirement of 60 resets in a 10-second window yields less than one fake trigger per month (Kubota et al., 2022). Directional reconstruction is based on RANSAC and PCA, and for a 10-kpc burst the paper reports that about 80% of cases are correctly identified within 25° of the true direction (Kubota et al., 2022).

Solar-neutrino studies use the same readout features in a more difficult background regime. The 2025 paper considers a kiloton-scale underground LArTPC instrumented with Q-Pix pixels of $I \propto 1/\mathrm{RTD}$1 pitch and examines two benchmark configurations: a high-background scenario with a 17.5 kt detector and 10 kt fiducial volume of atmospheric argon, and a low-background scenario based on the SLoMo concept (García-Peris et al., 21 Jul 2025). The readout advantages highlighted are continuous readout of all low-energy events, minimal data rate, manageable storage for offline analysis, and intrinsic 3D topological information (García-Peris et al., 21 Jul 2025). For low-energy neutrinos the study adopts an optimal threshold of $I \propto 1/\mathrm{RTD}$2 $I \propto 1/\mathrm{RTD}$3 6250 electrons, with equivalent energy per reset

$I \propto 1/\mathrm{RTD}$4

a clustering separation along time of $I \propto 1/\mathrm{RTD}$5 corresponding to 5 mm, and an event preselection threshold of 12 resets, corresponding to about 3 MeV (García-Peris et al., 21 Jul 2025). The expected recorded data are estimated as $I \propto 1/\mathrm{RTD}$6 TB/year for events above 3 MeV and $I \propto 1/\mathrm{RTD}$7 PB/year if storing down to $I \propto 1/\mathrm{RTD}$8 MeV (García-Peris et al., 21 Jul 2025).

The solar-neutrino analysis also demonstrates how Q-Pix’s 3D topology can be combined with optical information. A fiducial cut using 4 m of argon as passive shielding leaves about 20% of the active mass, approximately 2.3 kt, while strongly suppressing external backgrounds (García-Peris et al., 21 Jul 2025). For charged-current events on argon, the delayed de-excitation of $I \propto 1/\mathrm{RTD}$9 can emit a 1.64 MeV $I = \Delta Q \cdot f + I_\mathrm{o},$0 with a half-life of 336 ns; the study applies a delayed-light search in a 1500 ns window containing more than 99% of delayed $I = \Delta Q \cdot f + I_\mathrm{o},$1 flashes (García-Peris et al., 21 Jul 2025). It also defines a pulse-shape discrimination parameter with a fast-component integration window of 50 ns and reports that a cut of PSD $I = \Delta Q \cdot f + I_\mathrm{o},$2 gives 99% retention of electron events and 99% rejection of $I = \Delta Q \cdot f + I_\mathrm{o},$3-capture events where the $I = \Delta Q \cdot f + I_\mathrm{o},$4 ionizes, while cautioning that Geant4 has limitations in modeling $I = \Delta Q \cdot f + I_\mathrm{o},$5 transport (García-Peris et al., 21 Jul 2025).

The same paper is explicit about the physics limits. It states that discriminating solar-neutrino signals below 5 MeV is extremely difficult, that below 8 MeV solar neutrinos are overwhelmed, and that below 5 MeV background rates exceed the signal by nearly ten orders of magnitude (García-Peris et al., 21 Jul 2025). Q-Pix therefore appears in that study as an enabling readout, not as a solution to the underlying radioactive-background problem (García-Peris et al., 21 Jul 2025).

6. Calibration, limitations, and scope of the term

A distinctive aspect of the original Q-Pix proposal is absolute charge auto-calibration using intrinsic $I = \Delta Q \cdot f + I_\mathrm{o},$6 decay current in natural argon (Nygren et al., 2018). The 2018 paper treats $I = \Delta Q \cdot f + I_\mathrm{o},$7 not simply as background but as a calibration asset, giving approximate values of about 100 aA per pixel, roughly one decay every several seconds, and a $I = \Delta Q \cdot f + I_\mathrm{o},$8-driven reset rate of about 11 beats per minute (Nygren et al., 2018). In that picture, the pixel has a natural “heartbeat,” and the charge scale follows from

$I = \Delta Q \cdot f + I_\mathrm{o},$9

Later front-end work reiterates that the intrinsic $f$0 background can be used as a heartbeat calibration source to auto-calibrate charge sensitivity pixel by pixel (Miao et al., 2023).

The limitations identified across the literature are technical as well as physics-driven. The architectural papers note heavy demands on ultra-low leakage at cryogenic temperature, accurate local oscillators, careful reset behavior, minimized parasitic coupling, and robust data handling for dynamic routing and token passing (Nygren et al., 2018). The discrete front-end demonstration found that PCB parasitics were a major issue, including a large stray capacitance at the CSA input of roughly 20 pF and a resistor-package parasitic measured around 38 fF; the authors argue that these are engineering, not conceptual, obstacles and that they should be much less severe in a highly integrated ASIC (Miao et al., 2023). That same work concludes that replenishment is preferable to hard reset because it avoids charge loss during reset and is more tolerant of parasitics (Miao et al., 2023). On the detector-physics side, the solar-neutrino study shows that Q-Pix cannot by itself overcome a fundamentally unfavorable background environment, especially below 5 MeV, and that key backgrounds from cavern-wall $f$1 rays and $f$2-capture in radon decay chains remain underconstrained by existing measurements (García-Peris et al., 21 Jul 2025).

A recurring source of confusion is nomenclature. In the LArTPC literature, “Q-Pix” denotes the charge/time-encoding TPC readout architecture described above. It is distinct from QPIXL, a framework for quantum pixel representations and compression for $f$3-dimensional images that unifies encodings such as FRQI, NEQR, MCRQI, and NCQI and uses only $f$4 and CNOT gates with no ancilla qubits (Amankwah et al., 2021). It is also distinct from PhotonPix, a plug-and-play single-photon detector built around an FT MCP-PMT with logical NIM-like output and timing precision down to 10 ps (Orlov et al., 23 Mar 2026), and from $f$5, a single-shot quantitative phase imaging method based on a smartphone quad-pixel PDAF sensor in which each microlens covers a $f$6 pixel group (Bao et al., 31 Oct 2025). These technologies share pixel-centric or timing-centric language, but they address different device classes and measurement problems.

In its established arXiv usage, Q-Pix therefore refers most precisely to a self-triggering, pixelated CIR readout architecture for large TPCs, with fixed-charge quantization, RTD-based current inference, local timing, sparse data output, and a detector-scale design aimed at low power, 3D imaging fidelity, and resilience in very large neutrino detectors (Nygren et al., 2018)