⚛️ high-energy experiments

Readout electronics for SUBMET

This paper presents the design and performance validation of a dedicated data acquisition system for the SUBMET experiment at J-PARC, which utilizes cascaded DRS4 chips to achieve high-speed, low-noise single-photoelectron detection and precise timing synchronization across multiple channels to search for millicharged particles.

Original authors: Claudio Campagnari, Sungwoong Cho, Suyong Choi, Seokju Chung, Matthew Citron, Albert De Roeck, Martin Gastal, Seungkyu Ha, Andy Haas, Christopher Scott Hill, Insung Hwang, Hoyong Jeong, Jaebak Kim, Je

Published 2026-09-09

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Claudio Campagnari, Sungwoong Cho, Suyong Choi, Seokju Chung, Matthew Citron, Albert De Roeck, Martin Gastal, Seungkyu Ha, Andy Haas, Christopher Scott Hill, Insung Hwang, Hoyong Jeong, Jaebak Kim, Jeonghwa Kim, Hyunki Moon, Ryan Schmitz, David Stuart, Eunil Won, Jae Hyeok Yoo, Jinseok Yoo, Ayman Youssef, Ahmad Zaraket, Haitham Zaraket

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the vast landscape of modern physics, the Standard Model acts as a reliable map, charting the known particles and forces that make up our universe. Yet, like any map, it has blank spaces where new territories might lie. One such frontier involves the search for particles that carry a tiny fraction of the electric charge found on an electron. While an electron holds a specific, unchangeable amount of charge, these hypothetical particles, known as millicharged particles, would possess a charge so small it is a mere whisper compared to the standard. If they exist, they could be the hidden ingredients of dark matter, the invisible substance that holds galaxies together but refuses to interact with light. Finding them would rewrite our understanding of the cosmos, but their faint nature makes them incredibly difficult to spot. They would pass through detectors leaving behind only the slightest hint of their presence, a signal so weak it is easily drowned out by the static of the universe.

To catch these elusive whispers, a team of researchers built a specialized listening post at the Japan Proton Accelerator Research Complex. Their goal was to create a system capable of hearing a single flash of light produced when a millicharged particle zips through a block of plastic. The challenge was immense: the particle beam they were watching arrives in a rapid series of eight bursts, each separated by mere fractions of a second, and the signal from a single particle is so faint it amounts to just one photon of light. To solve this, the team developed a custom electronic system that acts like a high-speed camera, capturing the entire sequence of beam bursts in a single snapshot while being sensitive enough to distinguish a single photon from the background noise.

The heart of this new system is a dedicated data acquisition setup designed to handle the unique rhythm of the particle beam. The beam arrives in a specific pattern: eight tight groups of protons, spaced out over a span of about four microseconds. To ensure no signal is missed, the electronics must record a continuous window of time lasting five microseconds, covering the entire sequence of bursts. At the same time, the system must be fast enough to see the tiny, fleeting flashes of light. These flashes last only a few nanoseconds, so the electronics must take thousands of measurements every second to reconstruct the shape of the signal. The researchers solved this by using a specialized chip that acts as a rapid-fire analog memory, taking 4,096 snapshots of the incoming signal at a rate of 820.5 million times per second. This allows them to capture the full five-microsecond window with enough detail to see the smallest possible blip.

Because the signals are so weak, the system had to be engineered to be exceptionally quiet. The researchers designed the electronics to minimize any internal electrical noise, ensuring that the faint pulse from a single photon stands out clearly against the background. They achieved a level of precision where the electrical noise is less than 0.4 millivolts, a threshold low enough to detect a single photon with nearly perfect efficiency. To further improve their ability to see these small signals, they adjusted the baseline voltage of the system so that the negative-going pulses from the detectors could swing across a wider range of the recording scale. This simple adjustment effectively gave them a sharper resolution for the tiny signals they were hunting.

Timing was the other critical piece of the puzzle. Since the experiment relies on finding two matching signals that arrive at the same time in different layers of the detector, the clocks across the entire system had to be perfectly synchronized. The team built a network of ten electronic boards that communicate with a central control unit via fiber optic cables. This setup ensures that every part of the system operates in lockstep. They tested the timing precision rigorously and found that signals recorded on the same chip differ by less than one nanosecond, while signals across different boards stay within eight nanoseconds of each other. This level of precision is far tighter than the thirty-nanosecond window the experiment uses to declare a match, meaning the system can confidently distinguish a real particle event from random background noise.

The system proved its worth by handling the massive flow of data generated during beam operations. Each time the beam fires, the system captures data from 160 separate channels simultaneously, creating a digital record of the entire event. The team demonstrated that their setup could sustain a steady stream of data without losing any information, even when the trigger rate was pushed far beyond what the experiment normally requires. They confirmed that the system could handle a continuous flow of data at speeds up to 50 events per second, which is nearly two orders of magnitude faster than the typical operating rate. This robust performance ensures that the experiment can run smoothly, collecting the vast amount of data needed to search for the rarest of particles.

Ultimately, this work represents a successful marriage of custom hardware design and precise engineering. By building a system that combines high-speed sampling, ultra-low noise, and nanosecond-level timing, the researchers have created a tool capable of probing a region of physics that was previously out of reach. The system is now ready to hunt for millicharged particles with charges as small as one-thousandth of an electron's charge, potentially opening a new window into the dark matter that surrounds us. The successful implementation of this readout system marks a significant step forward, providing the sensitivity required to explore the unknown with a clarity that was not possible before.

Technical Summary: Readout Electronics for SUBMET

Problem Statement
The SUB-Millicharge ExperimenT (SUBMET), located at the Japan Proton Accelerator Research Complex (J-PARC), aims to detect millicharged particles (mCPs) with fractional electric charges $Q = \epsilon e$ where $\epsilon \lesssim 10^{-3}$ . Because the energy deposited by such particles scales with the square of their charge, mCPs produce at most a few scintillation photons, requiring the detector to be sensitive to single-photoelectron (SPE) pulses.

The experiment faces two simultaneous, conflicting technical challenges:

Signal Characteristics: The expected SPE pulses from Hamamatsu R7725 photomultiplier tubes (PMTs) are narrow (5–20 ns full width at half maximum) and low amplitude (5–30 mV). Capturing these requires high sampling rates (to resolve the pulse shape) and high amplitude resolution (to distinguish the pulse from noise).
Beam Structure: The J-PARC beam delivers spills consisting of eight proton bunches spaced 581 ns apart, spanning a total duration of approximately 4 µs. To correlate hits with the beam and identify coincidences between detector layers, the readout system must capture the entire 8-bunch train within a single trigger window.

Commercial off-the-shelf digitizers typically trade sampling rate for buffer depth; high-speed devices often lack the microsecond-scale recording window required to span the full J-PARC spill. Consequently, a custom data acquisition (DAQ) system was necessary to satisfy high sampling rates, long recording windows, and high channel density simultaneously.

Methodology and System Architecture
The authors developed a dedicated DAQ system comprising 160 channels distributed across ten Readout Boards (ROBs) and a central Trigger Control Board (TCB). The system utilizes a hierarchical, star-topology architecture connected via optical fibers.

Sampling Strategy (DRS4 Cascading): The core sampling device is the Domino Ring Sampler 4 (DRS4) chip. A single DRS4 channel provides 1024 cells. To achieve the required 5 µs window, the system cascades four DRS4 channels into a single 4096-cell input. Two such inputs are formed per chip. This configuration operates at a sampling rate of 820.5 MHz, providing a sampling interval of ~1.22 ns. This rate ensures at least four samples per 5 ns pulse while covering the full 4 µs bunch train.
Analog Front-End: The ROB analog front-end converts single-ended PMT signals to differential signals using ultra-low-noise fully differential amplifiers (THS4520). The baseline is deliberately offset upward (to ~1 V) to allow the negative-going PMT pulses to span a larger fraction of the digitizer's dynamic range, improving voltage resolution.
Digitization and Readout: The analog samples stored in the DRS4 capacitors are read out sequentially and digitized by a 12-bit ADC (AD9222) at 10.4167 MHz. The readout process takes approximately 786 µs per trigger, resulting in a maximum trigger rate of ~1.3 kHz, well above the experiment's needs.
Synchronization: The TCB distributes a 125 MHz reference clock and trigger signals to all ROBs via optical fibers using 8b/10b encoding. This ensures phase alignment across the system. To address static timing offsets between boards, an initial calibration is performed, and residual offsets are corrected in offline analysis.
Data Management: The TCB aggregates data from the 10 ROBs into a 4 GiB DDR4 memory buffer before streaming to a back-end server via USB. Each event packet is 128 KiB per ROB (1.25 MiB total for the system), containing 4094 waveform samples per channel after header insertion.

Key Contributions

Custom Cascaded Architecture: The implementation of a cascaded DRS4 configuration to extend the sampling window to 5 µs while maintaining 820.5 MHz sampling, a feature not available in commercial DRS4-based digitizers.
Low-Noise Design: The development of a fully differential signal path and elevated baseline strategy that achieves an RMS noise floor below 0.4 mV across all 160 channels.
Precise Timing Synchronization: A hardware design that couples collinear detector modules (required for coincidence identification) to channels on the same DRS4 chip, minimizing relative timing jitter.
Scalable DAQ System: A modular system capable of handling 160 channels with a throughput margin significantly exceeding the nominal physics requirements.

Results
Performance validation through bench tests and beam operation confirmed the system meets all design specifications:

Noise Performance: The measured pedestal RMS noise is uniformly between 0.34 mV and 0.37 mV across all 160 channels, staying below the 0.4 mV target. This results in a single-photoelectron detection efficiency of 0.995 for the R7725 PMTs.
Timing Precision:
- In-chip: Timing differences between channels on the same DRS4 chip are within 1 ns.
- Chip-to-chip (same board): Differences are within 2 ns.
- Board-to-board: Differences are within 8 ns.
  All these values are well within the experiment's 30 ns coincidence window.
Throughput: The system sustains loss-free data acquisition at rates up to 50 Hz (62.5 MiB/s), which is nearly two orders of magnitude higher than the nominal physics rate of ~0.8 Hz (1.0 MiB/s). The current bottleneck is identified as the sequential write speed of the server's hard disk storage, not the FPGA logic or USB interface.

Significance
The paper claims that the successful implementation and verification of this readout system represent a critical milestone for the SUBMET experiment. By achieving the necessary combination of low-noise waveform sampling, microsecond-scale recording windows, and nanosecond-level timing precision, the system enables the detection of the faint signals expected from millicharged particles. The authors state that this performance provides the sensitivity required to explore a previously inaccessible region of the millicharged particle parameter space, specifically reaching charges down to $\epsilon \sim 10^{-3}$ for masses up to approximately 1.6 GeV/c². The modular design also allows for future expansion to 208 channels without hardware modification.

Technical Summary: Readout Electronics for SUBMET

More like this