[Paper Review] M.i.p. detection performances of a 100 us read-out CMOS pixel sensor with digitised outputs

M. Winter, J. Baudot+15 more|ArXiv.org|Feb 16, 2009

Particle Detector Development and Performance3 citations

TL;DR

This paper presents the MIMOSA-22 CMOS pixel sensor with 100 µs read-out time and digitized outputs, demonstrating >99.5% detection efficiency at sub-10⁻⁴ fake hit rates and <4 µm spatial resolution. The sensor achieves high radiation tolerance (1.5 kGy) with only a 30% noise increase, validating its architecture for use in high-rate ILC vertex detectors and future EUDET beam telescopes.

ABSTRACT

Swift, high resolution CMOS pixel sensors are being developed for the ILC vertex detector, aiming to allow approaching the interaction point very closely. A major issue is the time resolution of the sensors needed to deal with the high occupancy generated by the beam related background. A 128x576 pixel sensor providing digitised outputs at a read-out time of 92.5 us, was fabricated in 2008 within the EU project EUDET, and tested with charged particles at the CERN-SPS. Its prominent performances in terms of noise, detection efficiency versus fake hit rate, spatial resolution and radiation tolerance are overviewed. They validate the sensor architecture.

Motivation & Objective

• Develop a fast, radiation-tolerant CMOS pixel sensor for high-rate ILC vertex detectors.
• Achieve sub-100 µs read-out time while maintaining high detection efficiency and low noise.
• Validate a pixel architecture with digitized outputs and zero-suppression logic for high data rate environments.
• Assess performance under ionizing radiation, simulating ILC beamstrahlung conditions.
• Enable integration into the EUDET beam telescope and future ILC detector layers.

Proposed method

• Fabricated a 128×576 pixel CMOS sensor with 18.4 µm pitch and 100 µs frame read-out time using 2008 EUDET FP6 project.
• Integrated discriminators with ±4% threshold dispersion and zero-suppression logic (SUZE-01) for data rate reduction.
• Used a 3-stage pipeline for hit detection: string encoding, bank merging, and 96 kbit memory buffering with LVDS output at 160 MHz.
• Tested sensor performance with 120 GeV π⁻ beams at CERN-SPS and ⁵⁵Fe sources to evaluate signal-to-noise ratio and spatial resolution.
• Exposed sensor to 1.5 kGy of ionizing radiation (10 keV X-rays) to assess radiation tolerance and noise increase.
• Measured detection efficiency, fake hit rate, spatial resolution, and noise across different discriminator thresholds.

Experimental results

Research questions

• RQ1 Can a CMOS pixel sensor with 100 µs read-out time maintain >99.5% detection efficiency at a fake hit rate below 10⁻⁴?
• RQ2 What is the spatial resolution of the sensor, and can it meet the 3 µm requirement for ILC inner layers?
• RQ3 How does ionizing radiation (up to 1.5 kGy) affect the sensor’s noise and detection performance?
• RQ4 Can the integrated zero-suppression logic (SUZE-01) reliably process high data rates without failure at 100 MHz?
• RQ5 Does the sensor architecture remain uniform and stable across the entire sensitive area under beam conditions?

Key findings

• The sensor achieved a detection efficiency of >99.5% at a fake hit rate of <10⁻⁴, with a signal-to-noise ratio of 17–21 (most probable) for various pixel designs.
• Single-point spatial resolution was measured at <4 µm, approaching the 3 µm requirement for ILC inner layers.
• After exposure to 1.5 kGy of ionizing radiation, detection efficiency remained >99.5% at a fake rate <10⁻⁴, despite a ~30% increase in noise.
• Noise increased by ~30% due to amplification circuitry, but the sensor remained functional and stable under radiation.
• The zero-suppression logic (SUZE-01) operated reliably up to 115 MHz without failures, validating its design for high-speed data processing.
• The MIMOSA-22 architecture was successfully validated and integrated into the final MIMOSA-26 sensor for the EUDET beam telescope.