[Paper Review] Sixteen Multiple-Amplifier Sensing Charge-Coupled Devices and Characterization Techniques Targeting the Next Generation of Astronomical Instruments

Agustín J. Lapi, Blas Irigoyen Gimenez+9 more|arXiv (Cornell University)|May 29, 2024

CCD and CMOS Imaging Sensors4 citations

TL;DR

This paper presents a 16-multiple-amplifier-sensing charge-coupled device (MAS-CCD) designed for next-generation astronomical spectroscopy, enabling sub-electron noise through multiple non-destructive readouts per charge packet. It achieves ~1 e⁻ rms noise with single sampling and sub-electron resolution via optimal noise averaging, introducing novel models for charge transfer inefficiency and amplifier noise characterization critical for future low-noise instruments.

ABSTRACT

We present a candidate sensor for future spectroscopic applications, such as a Stage-5 Spectroscopic Survey Experiment or the Habitable Worlds Observatory. This type of charge-coupled device (CCD) sensor features multiple in-line amplifiers at its output stage allowing multiple measurements of the same charge packet, either in each amplifier or in the different amplifiers. Recently, the operation of an eight-amplifier sensor has been experimentally demonstrated, and we present the operation of a 16-amplifier sensor. This new sensor enables a noise level of ~1e-rms with a single sample per amplifier. In addition, it is shown that sub-electron noise can be achieved using multiple samples per amplifier. In addition to demonstrating the performance of the 16-amplifier sensor, we aim to create a framework for future analysis and performance optimization of this type of detectors. New models and techniques are presented to characterize specific parameters, which are absent in conventional CCDs and Skipper CCDs: charge transfer between amplifiers and independent and common noise in the amplifiers and their processing.

Motivation & Objective

• To develop a low-noise CCD sensor for future spectroscopic surveys requiring single-electron resolution.
• To address the readout noise limitation in current CCDs, especially in blue-spectrum observations with high sky background.
• To enable fast, low-noise readout for space-based exoplanet detection with few photons per pixel.
• To characterize unique noise sources in MAS-CCDs, such as gate removal inefficiency and amplifier-specific noise.
• To establish a theoretical and experimental framework for optimizing performance in multi-amplifier CCD architectures.

Proposed method

• The 16-MAS-CCD uses in-line amplifiers to perform multiple non-destructive readouts of the same charge packet, reducing readout noise through averaging.
• An optimal noise averaging algorithm is developed to combine multiple samples per amplifier, minimizing total noise while preserving signal integrity.
• A common noise subtraction technique is introduced to isolate and remove correlated noise between amplifiers.
• A new model for gate removal inefficiency (GRI) is proposed, quantifying charge loss during transfer from sense node to serial register.
• Experimental characterization uses normalized pixel values in the overscan region to infer GRI and charge transfer efficiency.
• Theoretical models and fitting algorithms (e.g., in Appendix B) are used to extract parameters like ε from measured data across varying clock voltages and charge levels.

Experimental results

Research questions

• RQ1 How can multiple non-destructive readouts from 16 amplifiers in a CCD reduce readout noise below 1 e⁻ rms?
• RQ2 What are the dominant noise sources unique to multi-amplifier-sensing CCDs, and how can they be modeled and mitigated?
• RQ3 How does gate removal inefficiency (GRI) affect charge transfer from the sense node to the serial register, and what is its quantitative impact on signal fidelity?
• RQ4 What is the optimal method for combining multiple amplifier samples to achieve sub-electron noise performance?
• RQ5 How can clock voltage configurations be optimized to improve charge transfer efficiency and reduce GRI?

Key findings

• The 16-MAS-CCD achieves a readout noise of approximately 1 e⁻ rms with a single sample per amplifier, meeting the target for Stage-5 Spectroscopic Survey Experiment (Spec-S5).
• Sub-electron noise performance is demonstrated using multiple samples per amplifier, with the optimal averaging method significantly reducing total noise below 1 e⁻.
• A new source of charge transfer inefficiency, termed gate removal inefficiency (GRI), is identified and modeled, with ε values ranging from <5×10⁻⁵ to 3×10⁻⁴ depending on clock voltage settings.
• The GRI model shows a proportional dependence on charge level, with ε = 2.26×10⁻⁶Q + 3×10⁻⁴ in extreme cases, indicating measurable charge loss during sense node discharge.
• Experimental data confirm that the first overscan pixel in each amplifier shows no visible charge residue even under severe GRI, supporting the hypothesis that GRI is specific to the charge transfer process post-readout.
• The study demonstrates that reducing clock voltages below -10 V could further improve GRI performance, suggesting a path for future hardware optimization.