Skipper-CCDs: Low-Noise Silicon Imaging Sensors

• Skipper-CCDs are silicon imaging sensors with a floating-gate output that enables true nondestructive multiple sampling for ultra-low noise performance.
• They are fabricated on high-resistivity, backside-coated silicon and deployed in large mosaics for applications in astronomy, dark matter research, and quantum metrology.
• Synchronized multi-amplifier readout and adaptive sampling techniques reduce noise according to the 1/√N law, supporting precise photon counting and advanced ROI strategies.

Skipper Charge-Coupled Devices (Skipper-CCDs) are a class of silicon-based imaging sensors distinguished by their floating-gate output architecture, enabling true nondestructive, multiple sampling of pixel charge packets. Unlike conventional CCDs, which destructively read each pixel’s charge once per exposure, Skipper-CCDs are capable of reading the same charge packet numerous times without degrading the signal, thereby facilitating ultra-low readout noise and single-electron resolution. This capability has profound implications for applications in astronomical spectroscopy, rare-event searches (dark matter, neutrino physics), ultra-low light imaging, and quantum sensing, as demonstrated in large-format deployments and high-impact experiments (Villalpando et al., 2022, Villalpando et al., 2024, Barak et al., 2021).

1. Operating Principle and Readout Noise Suppression

Central to Skipper-CCD technology is the integration of a floating-gate output stage, which capacitively isolates the sense node from the video amplifier. This arrangement permits repeated, nondestructive measurements of a pixel’s charge packet. Each measurement exhibits an RMS noise $\sigma_1$, and by performing $N$ independent reads per pixel and averaging, the effective noise is reduced according to

$\sigma_N = \frac{\sigma_1}{\sqrt{N}}$

For best-performing amplifiers, $\sigma_1$ ranges from $2.5\,e^{-}$ to $4.5\,e^{-}$ RMS/pixel; with $N=800$ reads, sub-electron noise levels of $\sigma_N \sim 0.16\,e^{-}$ RMS/pixel are obtained, manifesting clear single-electron quantization in the pixel-value histogram (Villalpando et al., 2022, Barak et al., 2021).

This tunable noise property is critical for applications requiring single-electron (photon-counting) sensitivity, as it allows dynamic allocation of readout precision and frame time, including advanced Region-of-Interest (ROI) strategies (Chierchie et al., 2020).

2. Device Architecture, Mosaic Focal Planes, and Packaging

Skipper-CCDs are fabricated on high-resistivity ($>5$ k$\Omega\cdot$cm), fully-depleted $N$0-channel silicon, often thinned to $N$1m for astronomical applications and processed with a backside anti-reflective coating to maximize quantum efficiency. Devices for astronomy feature large imaging formats, e.g., $N$2 arrays of $N$3m pixels (Villalpando et al., 2022, Villalpando et al., 2023). These CCDs are mechanically integrated as mosaics (often $N$4) to cover focal planes up to $N$5.

Packaging involves the use of flexible printed circuits for clocks, biases, and video; precision alignment structures such as gold-plated Invar feet; and modular carrying boxes/test fixtures to facilitate handling and installation within vacuum dewars. Inter-device gaps are kept below $N$6m via custom mounts, ensuring compatibility with existing telescope cryostat geometries (Villalpando et al., 2022).

3. Synchronized Readout Electronics and Scalability

Each Skipper-CCD typically features four independent output amplifiers. For a $N$7-CCD mosaic, synchronized readout across $N$8 amplifiers is achieved via custom preamplifier PCBs, low-threshold acquisition (LTA) boards, and FPGA-based clock and bias generation (Villalpando et al., 2022). The readout electronics employ a “Leader/Follower” topology, guaranteeing sub-$N$9s alignment across all channels, and support 16-bit ADCs and Ethernet for data transfer. Firmware advances facilitate high-throughput ($\sigma_N = \frac{\sigma_1}{\sqrt{N}}$0Mpix/s), ROI, and multiplexed readout modes, enabling instrument-scale arrays with thousands of parallel channels (Botti et al., 2024, Chierchie et al., 2022).

Noise scaling adheres to the $\sigma_N = \frac{\sigma_1}{\sqrt{N}}$1 law up to at least $\sigma_N = \frac{\sigma_1}{\sqrt{N}}$2; deviations at higher $\sigma_N = \frac{\sigma_1}{\sqrt{N}}$3 indicate correlated noise sources, subsequently addressed by hardware and signal-processing optimizations. Key signal-to-noise formulas for spectroscopy incorporate the tunable Skipper-CCD readout noise as a quadratic term, allowing direct control of detector-limited S/N by increasing $\sigma_N = \frac{\sigma_1}{\sqrt{N}}$4 (Villalpando et al., 2022).

4. Performance Metrics: Noise, Quantum Efficiency, Charge Transfer

Representative performance metrics from science-grade Skipper-CCD systems include:

Metric	Typical Value	Notes
Single-sample readout noise ($\sigma_N = \frac{\sigma_1}{\sqrt{N}}$5)	$\sigma_N = \frac{\sigma_1}{\sqrt{N}}$6 RMS/pix	per amplifier (Villalpando et al., 2022)
Sub-electron performance ($\sigma_N = \frac{\sigma_1}{\sqrt{N}}$7)	$\sigma_N = \frac{\sigma_1}{\sqrt{N}}$8 RMS/pix	photon counting (Villalpando et al., 2022)
Full-well capacity	$\sigma_N = \frac{\sigma_1}{\sqrt{N}}$9	voltage-optimized (Villalpando et al., 2024)
Charge transfer inefficiency (CTI)	$\sigma_1$0	averaged (Villalpando et al., 2023)
Dark current	$\sigma_1$1 $\sigma_1$2/pix/s	$\sigma_1$3 operation (Villalpando et al., 2023)
Absolute quantum efficiency (QE)	$\sigma_1$4 (450–980nm), $\sigma_1$5 (600–900nm)	AR-coated (Villalpando et al., 2024)

QE measurements are performed via calibrated photodiodes and integrating spheres; consistency across amplifiers and across wavelength bands is confirmed at the $\sigma_1$6 uncertainty level (Villalpando et al., 2024).

CTI values, determined from extended pixel-edge response, are well below levels impacting astrometric or spectroscopic use. Dark current in deep underground or shielded environments is orders of magnitude lower than surface lab values, supporting rare-event sensitivity in dark matter and neutrino applications.

5. Region-of-Interest Readout and Adaptive Sampling Strategies

Given that readout noise scales as $\sigma_1$7 and readout time as $\sigma_1$8, Skipper-CCDs support adaptive sampling approaches whereby high-$\sigma_1$9 (sub-$2.5\,e^{-}$0 noise) is applied selectively to regions of interest (subarrays, spectral windows), while the remainder of the array is read with minimal $2.5\,e^{-}$1 for speed (Chierchie et al., 2020, Drlica-Wagner et al., 2021). Firmware “recipes” allow per-pixel or per-block $2.5\,e^{-}$2 assignment. For instance, $2.5\,e^{-}$3 of pixels at $2.5\,e^{-}$4 can achieve $2.5\,e^{-}$5 RMS in $2.5\,e^{-}$6min per frame, while background pixels are scanned quickly at single-read noise (Villalpando et al., 2023).

This flexibility enables “smart spectroscopy” workflows, e.g., targeted photon counting on faint emission lines or photometric windows, dramatically improving S/N in readout-noise-dominated regimes. Observations of faint high-$2.5\,e^{-}$7 quasars, emission-line galaxies, and ultra-faint dwarf stars have demonstrated real-world increases in sensitivity: detection of features hidden above $2.5\,e^{-}$8 readout noise, and S/N improvements from $2.5\,e^{-}$9 to $4.5\,e^{-}$0 for ELG lines upon reducing $4.5\,e^{-}$1 from $4.5\,e^{-}$2 to $4.5\,e^{-}$3 (Villalpando et al., 2024).

6. Applications in Astronomy, Particle Physics, and Quantum Measurement

Skipper-CCD arrays have been deployed for:

• Astronomical spectroscopy: Integral field units (IFUs) on large telescopes, with focal planes supporting photon-counting operation and sub-electron noise over $4.5\,e^{-}$4 formats (Villalpando et al., 2022, Villalpando et al., 2023, Villalpando et al., 2024).
• Dark matter and neutrino detection: High-mass arrays with sub-$4.5\,e^{-}$5 thresholds, modular packaging for radio-purity and cryogenic operation, and demonstrated background rates $4.5\,e^{-}$6/pix/day (Lin et al., 8 Sep 2025, Barak et al., 2021).
• Reactor-based CE$4.5\,e^{-}$7NS searches: Low-background, shielded detection at nuclear facilities, leveraging $4.5\,e^{-}$8 RMS noise with thick lead and polyethylene shielding (Depaoli et al., 2024).
• Quantum metrology: Self-calibrating, single-electron current sources for quantum-based ampere realization (Gamero et al., 11 Feb 2025).
• Space missions: Radiation-hardened Skipper-CCD designs for proposed satellite instruments, demonstrating sustained photon-counting performance post-proton irradiation (Roach et al., 2024).

In each domain, the sub-electron noise and associated signal discrimination unlock new sensitivity thresholds and practical operation modes that were previously unreachable with conventional CCDs or CMOS imagers.

7. Future Directions: Multi-Amplifier Architectures, Large-Scale Arrays, Advanced Readout Modes

The evolution of Skipper-CCD technology is focused on scalability and throughput. Multi-Amplifier Sensing (MAS) Skipper-CCDs segment the serial register across $4.5\,e^{-}$9 floating-gate nodes, allowing parallel multi-sample readout and reducing noise as $N=800$0 for $N=800$1 samples per amplifier. This architecture accelerates readout by $N=800$2, enabling full $N=800$3 photon-counting operation in under a minute (Villalpando et al., 2024).

Large-scale instrument arrays, including the $N=800$4 OSCURA experiment, employ wafer-level multi-channel silicon packaging, hierarchical analog and digital multiplexing, and automated module assembly, with design targets of $N=800$5 RMS noise, $N=800$6 yield, and background control to $N=800$7mHz/pixel (Botti et al., 2024, Chierchie et al., 2022).

Firmware and readout schemes continue to advance the exploitation of ROI, dynamic $N=800$8 adaptation, and high-speed synchronization, ensuring the practical viability of next-generation Skipper-CCD focal planes in ground- and space-based observatories and precision low-threshold experiments.

---

Collectively, Skipper-CCDs define the current state-of-the-art in low-noise silicon imaging and detection. Their unique floating-gate, multiple-sampling architecture facilitates dynamic, application-specific management of readout noise and time, single-electron and photon-counting capability, and modular scalability from laboratory prototypes to gigapixel-scale arrays. These characteristics underlie transformative advances in astronomy, particle physics, and quantum metrology (Villalpando et al., 2022, Villalpando et al., 2024, Barak et al., 2021, Botti et al., 2024).