Test-beam results from MiniCACTUS-v2: A depleted monolithic CMOS timing sensor prototype

The MiniCACTUS-v2 prototype, a depleted monolithic CMOS timing sensor fabricated in a 150 nm process and tested at CERN in July 2025, achieved a record time resolution of 48.88 ps on a 175-micron-thick sensor operated at 500 V.

The Gist

Imagine you are trying to take a photo of a speeding bullet. If your camera's shutter is too slow, the bullet will look like a blurry streak. In the world of particle physics, scientists face a similar challenge: they need to catch tiny, invisible particles (like muons) and record exactly when they pass through a detector. The faster and more precise their "shutter" is, the better they can understand the universe.

This paper is about a new, super-fast camera sensor called MiniCACTUS-v2. Here is the story of how it works and what the scientists found, explained in everyday terms.

The Problem: The "Noisy Neighbor"

In the past, the scientists built a prototype (MiniCACTUS-v1) that was very good at timing, but it had a major flaw. Think of the sensor like a quiet library. The "digital" part of the chip (which does the math and counting) was like a group of rowdy construction workers shouting right next to the "analog" part (which listens to the particles).

Because the wires connecting these two parts were too long, the shouting (electrical noise) was vibrating the sensitive listening equipment. This caused the signals to "ring" like a bell that wouldn't stop, making it hard to get a clean reading. Also, the sensor took too long to "reset" after hearing a particle, which was too slow for the high-speed traffic of the Large Hadron Collider (LHC).

The Solution: MiniCACTUS-v2

To fix this, the team redesigned the chip (MiniCACTUS-v2) with a few clever tricks:

1. Moving the Shouting: They moved the "construction workers" (the digital drivers) right next to the "listening post" (the sensors) and put them in a soundproof box (a deep n-well). This stopped the noise from vibrating the sensitive parts.
2. Shortening the Wires: They made the paths the signals travel much shorter, like replacing a long, winding hallway with a direct door.
3. Thinner Sensors: They took the silicon wafers and shaved them down to different thinness levels (150, 175, and 200 micrometers—about the thickness of a human hair). They then gave the back of the sensor a special "electric shock" (backside biasing) to make sure the whole thickness was ready to catch particles.

The Big Test: The "SPS" Race Track

In July 2025, the team took these new chips to CERN (the giant particle physics lab in Europe) to test them on a real "race track" of particles. They fired a beam of muons (particles that are like heavy electrons) at the sensors.

To measure how fast the sensor reacted, they used two "stopwatches" (Photomultiplier Tubes or PMTs) as a reference. It's like having two expert referees with perfect watches to time a runner. The goal was to see if the MiniCACTUS-v2 could keep up with the referees.

The Results: Breaking the 50-Second Barrier

The results were amazing. The team found that:

• It's Super Fast: The best sensor they tested (a tiny 0.5mm x 0.5mm square from the 175-micrometer thick chip) could time a particle with an accuracy of 48.88 picoseconds.
  • To put that in perspective: A picosecond is one-trillionth of a second. If you took one second and stretched it out to be the age of the universe, a picosecond would be about the time it takes for a snail to move the width of a human hair.
• It Handles High Voltage: They cranked up the voltage to 500 volts (like a very strong electric field), and the sensor didn't break. It stayed stable and kept working perfectly.
• No More Noise: The "shouting" problem from the old version was gone. The signals were clean and crisp.

Why Does This Matter?

Currently, the best timing detectors used in big experiments are expensive and complex (like Low Gain Avalanche Detectors). MiniCACTUS-v2 proves that you can make a timing sensor using standard, cheap computer chip manufacturing (CMOS) that is just as fast, if not faster.

The Bottom Line:
This paper shows that scientists have built a "smart, super-fast camera" for particles that is cheap to make, doesn't get confused by its own internal noise, and can time events with incredible precision. This paves the way for future particle colliders to see the universe in "slow motion" with much higher clarity than ever before.

Technical Summary

1. Problem Statement

The paper addresses the need for high-precision, low-cost timing detectors for future high-energy physics experiments (e.g., LHC upgrades, FCC-ee). While Low Gain Avalanche Detectors (LGADs) are the current standard, they are expensive and complex. Depleted Monolithic Active Pixel Sensors (MAPS) in CMOS technology offer a cheaper alternative but have historically struggled with two main issues in the MiniCACTUS-v1 prototype:

• Digital-to-Analog Coupling: Significant signal coupling from digital buffers to the analog sensor caused output signal ringing.
• Slow Recovery Time: The analog front-end (AFE) recovery time exceeded the 25 ns requirement for LHC applications.
• Reliability Issues: Previous iterations suffered from high leakage currents and unstable breakdown voltages due to manufacturing contamination and PCB layout issues.

2. Methodology

The authors designed and fabricated a new prototype, MiniCACTUS-v2, using the LF 150 nm CMOS process. The methodology involved:

• Sensor Architecture:

  • Active Area: A 4 × 5 pixel array with varying pixel sizes (1 mm², 0.5 mm², 0.25 mm²) and small test structures.
  • Sensing Element: Deep n-well/p-substrate diodes without internal amplification (depleted mode).
  • Front-End Design: To minimize capacitance, the analog front-ends (AC-coupled preamplifiers) and discriminators were moved outside the pixel to the column level.
  • Signal Integrity Improvements:
    • LVDS drivers were moved closer to discriminator outputs and placed inside a dedicated deep n-well to reduce coupling.
    • Two new preamplifier architectures were integrated: an optimized Charge Sensitive Amplifier (CSA) and a Voltage Pre-Amplifier (VPA) to shorten Time over Threshold (ToT).
  • Fabrication & Processing: Sensors were thinned to 150 µm, 175 µm, and 200 µm and underwent backside biasing (boron implantation, thermal annealing, metallization) to ensure homogeneous charge collection.

• Experimental Setup:

  • Test-Beam: Conducted in July 2025 at the CERN-SPS (H4 beamline) using 180 GeV/c muons.
  • Reference Timing: Two Hamamatsu PMTs coupled to a NE110 scintillator served as the timing reference.
  • Data Acquisition: A 12-bit, 1 GHz digitizing scope (LeCroy HDO4104A) recorded analog and digital signals.
  • Biasing: Sensors were tested at bias voltages up to 500 V.

• Data Analysis:

  • Time Walk (TW) Correction: Applied using a 7th-order polynomial fit on the analog signal amplitude vs. time difference.
  • PMT Timing: Constant Fraction Discrimination (CFD) at 10% and 50% levels with polynomial fitting to minimize dispersion.
  • Resolution Extraction: The individual time resolution of the Device Under Test (DUT) was calculated by inverting the covariance matrix of the time differences between the DUT and the two PMTs.

3. Key Contributions

• Design Optimization: Successfully mitigated digital-analog coupling issues by repositioning LVDS drivers and optimizing preamplifier architectures, resolving the ringing problems seen in v1.
• Process Robustness: Demonstrated that backside-biased, thinned depleted CMOS sensors can achieve breakdown voltages > 500 V, ensuring full depletion of the charge collection volume.
• Performance Validation: Proved that a depleted monolithic sensor without internal gain (unlike LGADs) can achieve sub-100 ps timing resolution, challenging the necessity of avalanche gain for high-precision timing.
• Thickness Study: Provided comparative data on timing resolution across three different sensor thicknesses (150, 175, and 200 µm).

4. Key Results

• Breakdown Voltage: All sensor variants (150, 175, and 200 µm) demonstrated breakdown voltages exceeding 500 V with extremely low leakage currents (often below the measurement noise floor of the Keithley 2470).
• Best Time Resolution:
  • The optimal result was achieved on a 0.5 mm × 0.5 mm pixel from a 175 µm-thick sensor biased at 500 V.
  • Measured Resolution: 48.88 ps (statistical error: 1.03 ps).
• Thickness Comparison:
  • The 175 µm and 200 µm sensors showed comparable performance at high voltages.
  • The 200 µm sensor with a 0.5 mm × 0.5 mm pixel achieved ~50 ps.
  • A 0.5 mm × 1.0 mm pixel on the 200 µm sensor achieved a best resolution of 60.23 ps.
• Reference PMT Performance: The extracted time resolutions for the reference PMTs were 51.64 ps (PMT1) and 33.76 ps (PMT2).

5. Significance

The MiniCACTUS-v2 results represent a significant milestone in the development of monolithic timing detectors:

1. Sub-100 ps Achievement: It validates that depleted CMOS sensors without internal amplification can meet the stringent < 100 ps timing requirements for future high-luminosity experiments.
2. Cost-Effective Alternative: By avoiding the complex fabrication steps required for LGADs (such as gain layers), this technology offers a potentially lower-cost, radiation-hard solution for large-area tracking and timing.
3. Future Roadmap: The success of this prototype paves the way for integrating Time-to-Digital Converters (TDCs) and data processing units directly onto the sensor die, enabling fully monolithic, high-performance timing detectors for the High-Luminosity LHC and beyond.