🔭 astrophysics

A bestiary of low-level electrical artifacts in SPHEREx flight data

This paper summarizes the SPHEREx team's analysis of low-level electrical artifacts in the mission's first flight data, detailing confirmed, modified, and newly discovered image-space effects such as cosmic rays, crosstalk, and unresponsive pixels, along with the implemented mitigation strategies and their implications for future space missions using HAWAII-2RG detectors.

Original authors: Candice M. Fazar, Michael Zemcov, C. Darren Dowell, Brendan P. Crill, Chi Nguyen, Howard Hui

Published 2026-08-11

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Candice M. Fazar, Michael Zemcov, C. Darren Dowell, Brendan P. Crill, Chi Nguyen, Howard Hui

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

Imagine the universe as a giant, cosmic library, and astronomers as librarians trying to read the faint, dusty spines of books that are billions of years old. To do this, they don't use regular eyes or cameras; they use super-sensitive "digital eyes" called detectors, which are essentially giant grids of tiny buckets designed to catch individual particles of light. But just like a bucket in a rainstorm, these detectors can get splashed by things other than the light they are trying to catch. Sometimes, a single drop of rain (a cosmic ray) hits a bucket and splashes water into its neighbors, making it look like it rained harder there than it actually did. Other times, if a bucket gets too full, the water overflows and spills into the next one, creating a chain reaction. This paper is about a team of space librarians who are learning how to spot these accidental splashes and spills in the data from a new, high-tech telescope called SPHEREx, so they can clean up the picture and see the real stars and galaxies clearly.

The paper, titled "A bestiary of low-level electrical artifacts in SPHEREx flight data," is essentially a field guide to the weird glitches that happen when the SPHEREx telescope takes pictures of the sky. SPHEREx is a NASA mission that launched in March 2025 to map the entire sky in 102 different colors of light, from near-infrared to mid-infrared. It uses six special detectors (HAWAII-2RGs) that are incredibly sensitive. However, because these detectors are so sensitive and the telescope is moving through space, the data it collects gets "contaminated" by electronic quirks and space particles. The authors, who work on the mission's data team, spent a year analyzing the first year of data to figure out exactly what these glitches look like, why they happen, and how to fix them. They call this collection of glitches a "bestiary," a word usually reserved for ancient books of mythical beasts, because these electronic monsters come in many strange shapes and sizes.

The first monster they describe is Charge Blooming. Imagine a very bright star hitting the detector. The light is so intense that the tiny bucket (pixel) meant to catch it fills up instantly. But the light keeps coming, and since the bucket can't hold any more, the extra charge spills over into the neighboring buckets, filling them up too, which then spill into their neighbors. It's like a game of musical chairs where the music never stops, and the overflow creates a growing puddle around the original star. The team found that this doesn't just happen right next to the star; sometimes, the "puddle" spreads hundreds of pixels away. They suspect that the overfilled pixels might actually start glowing like tiny light bulbs (LEDs) inside the detector material, sending light sideways through the silicon, which then gets caught by other pixels. This creates weird "shadows" or "rays" in the image that look like diffraction spikes but are actually caused by the detector itself. To fix this, they are learning to mask out (ignore) these large, glowing areas so they don't mess up the measurements of faint, distant objects.

The second beast is the Snowball. While blooming is caused by too much light, snowballs are caused by high-energy particles from space, like cosmic rays, hitting the detector. When one of these particles smashes into the detector, it deposits a huge amount of charge all at once. This is different from a star because it happens instantly. The result is a "snowball" effect: a central spot that is completely saturated (overflowing) surrounded by a halo of extra charge in nearby pixels. The authors noticed that these snowballs leave a specific signature: a dark center (where the charge was dumped so hard the data is unreadable) surrounded by a ring of extra signal. They measured how far this "halo" spreads and found a mathematical rule to predict its size based on how big the initial hit was. They now use this rule to automatically flag and remove these snowy patches from the final science images, ensuring that a random particle hit doesn't look like a new astronomical object.

The third category of glitches is Crosstalk, which is like a case of electronic whispering. In a perfect world, when a pixel reads a signal, it should stay quiet. But in SPHEREx, the way the detector reads the data (scanning rows and columns in a specific, non-sequential pattern) causes the electrical signal from a bright star to "leak" into other parts of the image. The team identified three types of this leakage:

Type 1 is like a ghost image appearing exactly 32 rows above the real star, but with the colors inverted (a negative image). This happens because of the specific "row-chopping" technique used to reduce noise.
Type 2 is a direct electrical leak where a bright star makes its neighbors in other channels look like they are also bright, creating a faint copy of the star in a different part of the detector.
Type 3 is a "smear" that happens when a bright star is at the very edge of a row. The electrical system gets confused by the sudden jump from a bright pixel to a dark one, causing a streak of signal to appear in the next row being read.

The paper doesn't just describe these monsters; it provides a "net" to catch them. The team has developed a set of rules and flags to identify these artifacts in the raw data. For example, if a pixel is flagged as "overflow" (too full), they check if it's surrounded by "transient" flags (sudden spikes) to decide if it's a blooming star or a snowball. They also use mathematical models to predict how far the snowball halo or crosstalk smear will reach and automatically mask those areas.

The authors are careful to note that while they have a good handle on these effects, some things are still a bit of a mystery. For instance, the theory that overfilled pixels act like internal LEDs to cause the long-range blooming shadows is a strong suggestion based on the geometry of the shadows, but it hasn't been tested in a lab yet. Similarly, the exact reasons for the Type 3 smear involve complex electronics that they suspect are related to the timing of the readout, but they aren't 100% sure. However, their main conclusion is clear: by understanding these "low-level electrical artifacts," they can clean up the data effectively. This is crucial because if they don't, the "noise" from these glitches could hide the faint, diffuse glow of the early universe that SPHEREx is designed to find. The paper serves as a guide for future missions too, showing that even with the best detectors, space is full of electronic surprises that need to be tamed to see the truth of the cosmos.

Technical Summary: A Bestiary of Low-Level Electrical Artifacts in SPHEREx Flight Data

Problem Statement

The SPHEREx mission, an all-sky spectrophotometric survey utilizing six HAWAII-2RG (H2RG) HgCdTe detectors, requires high-fidelity measurements of diffuse emission to achieve its science goals, particularly regarding the Extragalactic Background Light (EBL). While laboratory characterization of these detectors has identified various systematic anomalies (e.g., image persistence, crosstalk, 1/f noise), flight conditions introduce unique and unanticipated low-level electrical artifacts. Specifically, the interaction between high-flux astrophysical sources, energetic cosmic rays, and the mission's specific sample-up-the-ramp (SUR) readout strategy creates image-space artifacts that were not fully predicted by ground testing. These artifacts, if unmitigated, contaminate science data by introducing spurious signals, distorting point spread functions (PSFs), and obscuring faint diffuse emission.

Methodology

The authors analyzed one year of Level-1 (L1) and Level-2 (L2) flight data from the SPHEREx mission. The methodology involved:

Flag Analysis: Examining the SUR bit flags generated onboard the spacecraft (TRANSIENT, OVERFLOW, SUR ERROR) to identify pixel anomalies. These flags are derived from the slope calculation of the non-destructive readout sequence.
Pattern Recognition: Correlating specific combinations of flag bits (summed as integers) with visual patterns in the flux images to distinguish between astrophysical sources, cosmic ray hits, and electronic artifacts.
Stacking and Modeling: Creating finely gridded stacks of similar events (e.g., snowballs of specific sizes) to quantify the spatial extent of excess signals (halos) and radial profiles.
Empirical Modeling: Deriving mathematical models to describe the magnitude and spatial decay of artifacts, such as the "snowball halo" and various crosstalk types, to establish masking thresholds.
Cross-Validation: Comparing flight data behavior against laboratory predictions and known detector physics (e.g., the brighter-fatter effect, interpixel capacitance) to hypothesize physical mechanisms for new phenomena.

Key Contributions and Results

1. Charge Blooming and Extended "Bloom Shadows"

The study confirms that sources bright enough to trigger the OVERFLOW flag (saturation) induce TRANSIENT flags in neighboring pixels due to charge blooming.

Mechanism: Excess charge from saturated pixels spills into neighbors, accelerating their charge accumulation and tripping transient detection.
New Finding (Extended Bloom): For extremely bright sources (e.g., Saturn), excess signal extends tens to hundreds of pixels beyond the immediate blooming ring. The authors observe "shadow" effects where low-efficiency bad pixel groups block this extended signal, creating wedge-shaped radial shadows.
Hypothesis: The authors propose that forward-biased detector diodes in saturated regions emit photons at the cutoff wavelength (acting like an LED). These photons propagate radially within the HgCdTe material (light-pipe effect) until absorbed by low-efficiency regions.
Mitigation: Source masks for bright objects have been expanded to cover these extended radii.

2. Snowballs and High-Energy Cosmic Ray Halos

The paper distinguishes between standard cosmic ray hits and "snowballs" (high-energy particle impacts or radioactive decay events).

Identification: Snowballs are identified by a core of OVERFLOW and TRANSIENT flags (often with mixed SUR ERROR states) surrounded by a ring of transient pixels. Unlike standard transients, they exhibit a "halo" of excess signal extending into neighboring pixels due to interpixel capacitance (IPC) and sub-threshold charge deposition.
Quantification: The authors modeled the excess current ( $i_{halo}$ ) as a function of radius ( $r$ ) and the number of pixels in the transient cluster ( $A_{snow}$ ):
$i_{halo}(r) = \frac{0.2 A_{snow}}{r^2}$
Mitigation: A HALO mask bit is applied in the L2 pipeline, masking pixels out to a radius where the predicted excess signal drops below 0.1 e⁻/s (the noise floor). This effectively removes the contamination from science analysis.

3. Three Types of Electrical Crosstalk

The authors characterized three distinct crosstalk mechanisms that manifest when bright sources illuminate the array:

Type 1 (Row-Chopping Artifact): A negative imprint of a bright source appears exactly 32 rows away (consistent with the NSKIP parameter). This is a coupling between the source and the "phantom" pixel sampling used for noise suppression. It is most prominent in Band 1.
Type 2 (Simultaneous Readout Crosstalk): Bright sources generate excess charge in neighboring channels read out simultaneously. This is true electrical crosstalk with coupling coefficients on the order of $10^{-3}$ , strongest in Band 1 and the SWIR bands.
Type 3 (Readout Chain Smear): A bright pixel at the edge of a channel causes a smeared signal in the subsequent row (32 rows away). This is attributed to hysteresis or offset decay in the Video 8 readout electronics following a large signal step function. Unlike Types 1 and 2, this can produce coupling coefficients $>1$ and is not true crosstalk.
Mitigation: Empirical models are used to predict the crosstalk signal magnitude. Pixels where the predicted signal exceeds 0.3 e⁻/s are flagged and masked in L2 data.

Significance and Claims

The paper claims to provide the first comprehensive catalog of unanticipated electronic artifacts in SPHEREx flight data, moving beyond laboratory predictions.

Operational Impact: The identification and masking of these artifacts (BLOOM, SNOWBALL, HALO, and crosstalk flags) are critical for the SPHEREx Science Data Center to produce clean L2 images suitable for high-fidelity diffuse emission studies.
Scientific Insight: The proposed "forward-biased pixel glow" theory offers a novel explanation for extended blooming effects and shadowing, suggesting that saturated detectors can emit light that propagates through the substrate.
Broader Applicability: The characterization of these artifacts, particularly the snowball halos and crosstalk types, provides a framework for other space missions using HAWAII-2RG detectors to anticipate and mitigate similar low-level systematic errors.
Limitations: The authors note that while masking strategies are effective for sparse events, crowded fields and particle storms (where effects overlap) remain challenging, potentially increasing the number of affected pixels by a factor of three or more. Consequently, SPHEREx marks specific images for re-observation to mitigate data loss.

The paper concludes that while these artifacts are manageable through the presented masking strategies, further study of the underlying physical mechanisms (such as the forward-bias glow) and optimization of readout strategies are necessary to minimize contamination in future missions.