🔭 astrophysics

Feasibility of up-the-ramp sampling under variable sky for ground-based spectrographs

This feasibility study demonstrates that while up-the-ramp sampling offers 4–10% observing time savings for read-noise-limited targets in inter-line regions under variable K-band sky conditions, its performance over emission lines is compromised by signal-driven cosmic ray false positives and SNR degradation, though these issues can be mitigated by adapting rejection thresholds.

Original authors: Gaia Gaspar, Marcin Sawicki, Nelson Nunes, Rubén J. Díaz, James E. H. Turner

Published 2026-06-12

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Gaia Gaspar, Marcin Sawicki, Nelson Nunes, Rubén J. Díaz, James E. H. Turner

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 you are trying to listen to a very faint whisper (a distant galaxy) in a room where the background noise (the Earth's atmosphere) is constantly changing. You have a special microphone (a telescope detector) that can take hundreds of snapshots of the sound very quickly.

This paper asks a simple question: Is it better to take all those quick snapshots and average them out later (a method called "Up-the-Ramp"), or is it better to just take a few snapshots and average them immediately (the traditional method)?

The researchers found that for most faint objects, taking all the snapshots and averaging them later is a winning strategy, even if the background noise is a bit wobbly. However, there are some specific spots in the "room" where this trick doesn't work as well.

Here is the breakdown of their findings using everyday analogies:

1. The Two Listening Strategies

The Old Way (Fowler/MCDS-8): Imagine taking two photos of a scene—one at the start and one at the end—and averaging them. It's simple, but if the camera is shaky (read noise), you might miss faint details.
The New Way (Up-the-Ramp/UTR): Imagine taking 60 photos in a row. Instead of just looking at the first and last, you draw a line through all of them to see the trend. This is much better at ignoring camera shake and spotting sudden glitches (like cosmic rays hitting the sensor).

2. The Problem: The "Wobbly" Sky

The Earth's atmosphere isn't a quiet, still room. The "sky brightness" (the background glow) changes by 3% to 10% every few minutes. The researchers were worried that if the background noise changes while they are taking their 60 photos, the "line" they draw might get distorted, ruining the measurement.

3. The Results: What Happened?

A. The "Quiet Corners" (Inter-line Regions)
Most of the time, the galaxy's light falls in the dark spaces between the bright lines of the sky's glow.

The Finding: In these quiet corners, the "Up-the-Ramp" method works beautifully. Even with the wobbly sky, it still gives a clearer picture than the old method.
The Benefit: It saves 4% to 10% of observing time. Think of it like getting a 10% discount on your telescope time. Over a year, this could save astronomers about 15 nights of observation time.
The Accuracy: The measurements are just as accurate as the old method. The "wobbly sky" didn't trick the math.

B. The "Noisy Spots" (Sky Emission Lines)
Sometimes, the galaxy's light falls right on top of a bright, flickering line of the sky's own glow.

The Finding: Here, the "Up-the-Ramp" method gets confused. Because the signal is so bright, the computer starts thinking that normal random noise is a "glitch" (a cosmic ray) and tries to delete it.
The Consequence: The computer accidentally deletes real data, making the picture 10-20% worse than the old method.
The Fix: The researchers suggest that for these specific bright spots, we should turn off the "glitch detector" or tune it differently, rather than using the same settings for the whole image.

4. The "False Alarm" Issue

One of the main perks of the "Up-the-Ramp" method is that it's great at spotting cosmic rays (tiny particles hitting the detector that look like bright sparks).

The Issue: When the sky is bright (over those emission lines), the signal is so strong that the computer gets jumpy. It starts flagging normal, random noise as "cosmic rays" and throws them away.
The Reality: The researchers found that these false alarms aren't caused by the sky changing; they are caused simply because the signal is too loud for the standard settings. It's like a smoke detector going off because you are cooking steak, not because there is a fire.
The Solution: You just need to tell the detector, "If you are in a bright area, be less sensitive to alarms."

5. The Data Mountain

There is one downside. Because the "Up-the-Ramp" method saves 60 snapshots instead of just a few averaged ones, the files are 8 times larger.

The Trade-off: The researchers argue this is worth it. Storing extra data on a hard drive is cheap. But time on a giant telescope is incredibly expensive. Saving 10% of your time is worth the extra storage space.

Summary

The paper concludes that for the Gemini Infrared Multi-Object Spectrograph (GIRMOS) and similar telescopes, using the "Up-the-Ramp" method is a smart move.

For faint objects in dark sky gaps: It's a clear winner, saving time and money.
For bright sky lines: It needs a little adjustment (tuning the glitch detector) to avoid deleting real data.
Overall: The benefits of better data and saved time outweigh the hassle of managing larger file sizes.

Technical Summary: Feasibility of Up-the-Ramp Sampling under Variable Sky for Ground-based Spectrographs

Problem Statement
Modern near-infrared (NIR) instruments increasingly utilize HAWAII-2RG (H2RG) detectors with long integration times (300–600s). While Up-the-Ramp (UTR) sampling offers significant advantages over traditional Fowler (Multi-Correlated Double Sampling, MCDS) modes—specifically superior cosmic ray (CR) rejection and extended dynamic range—it requires fitting linear ramps from 30–60 non-destructive reads. A primary concern for ground-based applications is the temporal variability of the sky background, which has been reported to fluctuate by 3–10% on minute timescales. Such variability threatens to introduce systematic errors in ramp fitting, compromise photometric accuracy, and confuse CR rejection algorithms by mimicking cosmic ray jumps. Furthermore, UTR data formats generate high-dimensional FITS files with significantly larger volumes than standard 2D processed frames, posing challenges for observatory operations and data storage. The impact of these variable sky conditions on UTR performance had not been systematically characterized for ground-based spectrographs.

Methodology
This study presents a feasibility analysis using the GIRMOS Data Simulator (DSIM), a high-fidelity simulation tool developed for the Gemini Infrared Multi-Object Spectrograph (GIRMOS). The study focuses on the K-band (and H-band for comparison) under realistic observing conditions at Mauna Kea.

Simulation Setup: The authors simulated 300 Monte Carlo realizations of noise (shot noise, read noise, dark current) and 100 realizations including cosmic rays. The simulations compared two readout modes: UTR (with 60 saved reads over ~600s) and MCDS-8 (the aggressive Fowler-8 mode used by Gemini).
Sky Variability Model: Empirical K-band sky variations were derived from archival Gemini-NIRI data (10–20s cadence). The model incorporated a linear drift of 2% over 700s and stochastic Gaussian pulses (up to 8% amplitude) to create a conservative, aggressive variability scenario.
Target Configuration: Simulations focused on faint emission-line sources representative of high-redshift galaxy surveys, placed in dark inter-line regions where the signal is read-noise dominated.
Analysis Tools: The fitramp module was used for ramp fitting, employing a $\chi^2$ improvement approach for CR rejection. The study analyzed systematic biases, Signal-to-Noise Ratio (SNR) improvements, false positive rates in CR masking, and CR recovery rates across three sky conditions: no sky, stable sky, and variable sky.

Key Contributions and Results

SNR and Bias in Inter-Line Regions:
- For targets in dark inter-line regions (read-noise dominated), UTR sampling maintains its SNR advantage over MCDS-8 even under variable sky conditions.
- SNR Gains: The study quantified SNR improvements of 4–10% for UTR over MCDS-8. This translates directly to 4–10% savings in observing time to achieve the same SNR.
- Systematic Bias: The bias in count rates between UTR and MCDS-8 was found to be negligible (maximum ~0.1% in H-band, ~0.01% in K-band), well within typical NIR flux calibration uncertainties (1–2%).
Performance Over Sky Emission Lines:
- In regions containing bright sky emission lines (Poisson noise dominated), UTR fitting performance degrades.
- SNR Degradation: A 10–20% reduction in SNR was observed for UTR compared to MCDS-8 in these regions.
- False Positives: The CR rejection algorithm flagged a high rate of pixels (10–40%) as cosmic rays in emission line regions. Crucially, the study determined that these false positives are driven by the high signal level (shifting pixels to the shot-noise regime) rather than sky variability itself.
- Mitigation: The authors suggest that CR rejection thresholds should be adapted to local signal levels or disabled in sky-line regions to mitigate false positives.
Cosmic Ray Recovery:
- For faint inter-line targets, CR recovery rates remained consistently high (>97%) across all sky modes (no, stable, and variable) when using optimized per-amplifier thresholds.
- The study confirmed that sky variability does not significantly disrupt CR detection in the read-noise dominated regime, provided thresholds are tuned to the specific amplifier noise characteristics.
Data Volume Impact:
- UTR readout generates significantly larger data volumes. A single night of GIRMOS operations with UTR produces approximately 421 GB of raw and processed data, compared to ~54 GB for MCDS-8 (an ~8-fold increase).
- Despite the storage burden, the authors argue that the time savings (4–10%) outweigh the costs of increased data volume, as telescope time is a more scarce resource than archive capacity.

Significance and Claims
The paper concludes that UTR sampling is feasible and advantageous for ground-based NIR spectrographs like GIRMOS, specifically for read-noise-limited targets located in inter-line regions. The study asserts that:

The advantages of UTR (SNR gain, CR rejection) are preserved under realistic, variable sky conditions.
The systematic biases introduced by sky variability are negligible for photometric accuracy.
The primary challenge lies in sky-line regions, where the high signal level, not sky variability, degrades performance; this can be managed through region-specific thresholding.
The trade-off favors UTR implementation, offering substantial efficiency gains (estimated at ~15 nights recovered annually for a 10-year instrument lifetime) despite the increased data volume.

The authors note that these results rely on GSAOI read noise values as a proxy and that future work should incorporate $1/f$ noise and seeing variations once actual detector data becomes available. However, the current findings provide a robust basis for implementing UTR readout in next-generation ground-based instruments.