🔭 astrophysics

The Tartu Observatory Fiber-fed Echelle Spectrograph (TOFES) Data Reduction Pipeline

This paper introduces the PyReduce-based data reduction pipeline and the Spectroscopy-Toolbox for the Tartu Observatory Fiber-fed Echelle Spectrograph (TOFES), a high-resolution instrument designed for the 1.5 m AZT-12 telescope to support scientific goals such as characterizing Herbig AeBe stars, exoplanet host-stars, and massive star atmospheric variability.

Original authors: Sandipan P. D. Borthakur (Tartu Observatory, Estonia, Space Research Institute, Austria, Graz University of Technology, Austria), Tõnis Eenmäe (Tartu Observatory, Estonia), Nikolai Piskunov (Uppsala U

Published 2026-08-03

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Sandipan P. D. Borthakur (Tartu Observatory, Estonia, Space Research Institute, Austria, Graz University of Technology, Austria), Tõnis Eenmäe (Tartu Observatory, Estonia), Nikolai Piskunov (Uppsala University, Sweden), Luca Fossati (Space Research Institute, Austria), Mihkel Kama (University College London, UK, Tartu Observatory, Estonia), Thomas Marquart (Uppsala University, Sweden), Anni Kasikov (Tartu Observatory, Estonia, European Southern Observatory, Chile), Heleri Ramler (Tartu Observatory, Estonia), Anna Aret (Tartu Observatory, Estonia), Christiane Helling (Space Research Institute, Austria), Laurits Leedjärv (Tartu Observatory, Estonia), Linn Boldt-Christmas (Uppsala University, Sweden)

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 single violin playing in the middle of a roaring stadium. To hear the music clearly, you need a way to isolate that specific sound from the crowd, the wind, and the other instruments. In the world of astronomy, stars are those violins, and their "music" is light. When starlight passes through a special prism called an echelle grating, it doesn't just split into a simple rainbow; it gets chopped into dozens of tiny, overlapping slices of color, like a puzzle where the pieces are stacked on top of each other. This is called high-resolution spectroscopy.

Why do we care about this puzzle? Because the way the light is chopped tells us everything about the star. It reveals how fast the star is spinning, what it is made of, and if it has planets hiding around it. However, turning those raw, messy slices of light into a clean, readable story is incredibly difficult. It requires a very smart computer program—a "data reduction pipeline"—to sort the pieces, remove the static, and stitch them back together. Without this software, the telescope is just a fancy camera taking pictures of a jumbled mess.

This paper introduces a new, super-smart computer program designed specifically for a brand-new instrument called TOFES (Tartu Observatory Fiber-fed Echelle Spectrograph). Think of TOFES as a high-tech fiber-optic cable that sucks up starlight and pipes it into a spectrograph sitting in a quiet, temperature-controlled room. The authors, a team of astronomers from Estonia, Austria, Sweden, and the UK, have built a digital toolkit to clean up the data coming from this machine. They tested their new software by pointing the telescope at the Sun (safely, of course, by bouncing its light off the sky) to see if the program could correctly identify the Sun's "voice."

The paper finds that their new pipeline works very well, successfully taking the raw, chaotic data from the spectrograph and turning it into a clean spectrum with a resolution of about 30,000. This means the instrument can distinguish between two colors of light that are incredibly close together, allowing astronomers to measure the speed of a star with a typical uncertainty of about 0.2 kilometers per second. The team also created a separate "toolbox" of smaller programs to help users check the quality of their data, like measuring how much "static" (noise) is in the detector or finding the exact wavelength of every color.

When they tested the system on the Sun, the results were promising but showed some expected imperfections. The software calculated the Sun's speed relative to Earth and found an average shift of nearly zero, but with a scatter of about 0.19 kilometers per second across different parts of the spectrum. While the overall "voice" of the Sun recorded by TOFES matched the known reference recordings very closely, the team noted small systematic trends and variations in the data. These findings confirm that the new instrument and its new software are robust and ready for real science, even if the data isn't quite "perfect" in every single detail.

So, what is this new machine actually going to do? The authors suggest it will be a workhorse for studying a variety of cosmic mysteries. It will help them figure out how young, massive stars are feeding on gas from their surrounding disks, hunt for the host stars of planets discovered by space telescopes, and monitor the wild, churning atmospheres of giant stars that are pulsing and losing mass. Essentially, they have built a new, highly sensitive ear for the universe and the software to make sure we can hear every note clearly.

Technical Summary: The Tartu Observatory Fiber-fed Echelle Spectrograph (TOFES) Data Reduction Pipeline

Problem and Context
Medium-to-high resolution echelle spectroscopy ( $R \gtrsim 30,000$ ) is essential for diverse astronomical applications, ranging from exoplanet detection and characterization to galactic archaeology and stellar atmospheric studies. While existing facilities like the 1.5 m AZT-12 telescope at Tartu Observatory have utilized long-slit spectrographs for decades, the integration of a modern fiber-fed echelle spectrograph is required to achieve higher spectral resolution and broader wavelength coverage in a single exposure. The Tartu Observatory Fiber-fed Echelle Spectrograph (TOFES), a Whoppshel instrument by Shelyak Instruments, has been installed in a dedicated Coudé room. However, to utilize this instrument effectively, a robust, automated data reduction pipeline (DRP) capable of handling the specific optical characteristics of a fiber-fed system and performing high-precision spectral extraction was necessary.

Methodology
The paper introduces the TOFES-DRP, a data reduction pipeline built upon the Python-based PyReduce package. PyReduce is an evolution of the classic reduce package, incorporating capabilities to account for slit tilt and curvature and to perform continuum normalization. The pipeline workflow, detailed in Section 4, follows a standard echelle reduction sequence adapted for TOFES:

Pre-processing: Bias subtraction is performed using master bias frames generated via median combination. The pipeline calculates readout noise and gain using specific statistical methods (Equations 1–3).
Flat Fielding and Order Tracing: Master flat frames are used to correct pixel-to-pixel variations. A clustering algorithm identifies spectral orders on the detector, fitting polynomial functions to trace their locations (Section 4.2).
Background Correction: Scattered light, originating from reflections off the echelle grating holder, is estimated using pixels outside the order traces and removed via 2D polynomial fitting.
Spectral Extraction: The core extraction utilizes a 2D model fitting approach (Horne 1986) where the slit illumination function and spectrum are modeled. The algorithm divides orders into overlapping "swaths" to account for variations in the slit function along the order, weighting the final extraction linearly across these swaths.
Wavelength Calibration: A ThAr lamp spectrum is used to generate a wavelength solution. The authors determined optimal 2D polynomial degrees (dispersion $p_d=4$ , cross-dispersion $p_{cd}=7$ ) by minimizing the root mean square residual velocity ( $\sigma_v$ ) across a grid of polynomial degrees (Section 4.4).
Blaze Correction and Normalization: The blaze function is extracted from flat fields and smoothed using Gaussian kernel regression. To address edge effects where adjacent orders drop in flux, a scaling method using the continuum of neighboring orders is applied. Continuum normalization is performed using either the PyReduce iterative clipping method or the Spectroscopy-Toolbox package, which offers polynomial, spline, or kernel regression options.

Complementing the DRP, the authors developed the Spectroscopy-Toolbox, a modular Python package providing diagnostic tools for readout noise, gain, spectral resolution, wavelength solution generation, and signal-to-noise estimation.

Key Contributions and Results
The primary contribution of this work is the release of the TOFES-DRP and the Spectroscopy-Toolbox as open-source software. The pipeline was validated using solar spectra obtained by projecting the fiber to the sky during daylight, comparing the results against a high-resolution archival reference spectrum from the ESPaDOnS instrument.

Wavelength Calibration: Cross-correlation between the TOFES solar spectrum and the ESPaDOnS reference yielded an average velocity shift of $-0.01$ km s $^{-1}$ with an average scatter of $0.19$ km s $^{-1}$ . The standard deviation within orders was approximately $0.20$ km s $^{-1}$ .
Flux Consistency: The mean flux difference between the TOFES and reference spectra was $0.003$, with a scatter of $0.043$.
Spectral Resolution: Using the resolving power calculator on ThAr lines, the pipeline confirmed an average spectral resolution of $\sim 30,000$ for the 50 $\mu$ m fiber, consistent with manufacturer specifications. The 105 $\mu$ m fiber is expected to achieve $R \approx 16,000$ .
Software Architecture: The Spectroscopy-Toolbox provides a GUI for interactive line identification and wavelength solution generation, while the DRP automates the reduction process, handling bad pixel masking (e.g., a specific streak pattern in column 1158) and parameter management.

Significance and Claims
The paper asserts that the TOFES-DRP successfully processes data from the new fiber-fed echelle spectrograph, enabling the Tartu Observatory to address specific scientific questions. The authors outline the instrument's intended scientific applications, which include:

Stellar Characterization: Measuring accretion contamination in Herbig Ae/Be stars to constrain protoplanetary disk compositions.
Exoplanet Host-Star Analysis: Characterizing stars hosting TESS candidates (specifically those brighter than $V=7$ and earlier than spectral type F2) to support the Ariel space mission.
Massive Star Variability: Monitoring large-scale atmospheric variability and kinematics in massive stars (e.g., Be stars, supergiants) through radial velocity measurements.

The authors emphasize that the pipeline is designed to be flexible and modular, with the Spectroscopy-Toolbox capable of analyzing data from other instruments with similar formats. They note that while the instrument adapter connecting TOFES to the 1.5 m telescope is still under development, the pipeline has been rigorously tested on solar data and is ready for scientific operations once the full system is commissioned. The work does not claim to have solved telluric correction issues (which remain a future task for tools like molecfit) nor does it claim to have performed the aforementioned stellar studies yet; rather, it establishes the necessary data processing infrastructure to enable those future studies.

More like this