pyTANSPEC v1.0 and HxRGproc: Updated packages to Clean and Reduce TANSPEC data

This paper introduces upgraded versions of the `pyTANSPEC` and `HxRGproc` Python packages, which provide enhanced data reduction capabilities for the TANSPEC instrument, including support for all slit widths, improved wavelength calibration, flux calibration, and automated cleaning of detector readout frames.

The Gist

Imagine an astronomer trying to extract the SPECTRUM of a faint, distant star — that is, the rainbow of colours the star is emitting, spread out so each colour can be measured separately. A spectrum is what you get when you take starlight and split it into its component wavelengths, the way a prism splits sunlight into a rainbow; from that "fingerprint," you can read off what the star is made of, how hot it is, and how fast it is moving. The instrument that does this for the TANSPEC telescope in India is incredibly sensitive, but because it is so powerful, the raw data comes with a lot of "glitches." It gets "dizzy" from the light, it gets "distracted" by cosmic rays (tiny space particles hitting the sensor), and sometimes the spectral data looks a bit "smudged" or "warped."

This paper describes a new set of digital cleaning and processing tools (software packages called HxRGproc and pyTANSPEC) designed specifically for the massive, specialized TANSPEC spectrograph.

Here is the breakdown of how these tools work, using everyday analogies:

1. The "Digital Scrubbing" (HxRGproc)

Before astronomers can analyze the star, they have to clean the "raw" data. Think of the raw data like a recording of a faint sound in a noisy room.

• Removing the "Static" (Bias Correction): Imagine you're listening to a radio station, but there's a constant, annoying hum in the background. HxRGproc acts like a noise-canceling headphone, identifying that constant hum and filtering it out so the music (the starlight) can be heard clearly.
• Fixing the "Overwhelmed" Sensor (Non-linearity Correction): Imagine you are trying to measure how much rain has fallen by collecting it in a glass whose width changes with height — narrow near the bottom, wider in the middle, and narrow again near the top. Even though the same amount of rain is going in, the water level inside doesn't rise at a constant rate: a small amount of rain near the bottom looks like a big jump, while a lot of rain near the top barely moves the level. The camera sensor behaves the same way — its response is NON-LINEAR, so the brightness it reports at low light isn't on the same scale as the brightness it reports near saturation. The software applies a calibration curve to correct for this, so the measured brightness levels are accurate across the full range.
• Dodging "Space Dust" (Cosmic Ray Correction): Occasionally, a tiny particle from deep space slams into the camera sensor, leaving a bright, fake "blip" on the data. It's like a piece of dust landing on your camera lens right when you take a photo. The software identifies these "fake" bright spots and smooths them over so they don't ruin the spectrum.

2. The "Digital Map Maker" (pyTANSPEC)

Once the data is clean, you need to turn that "spectrum" into actual scientific information (like knowing exactly what a star is made of).

• Tracing the Path (Spectral Extraction): When the starlight enters the spectrograph, an optical element (a prism-like dispersing element) spreads it out by colour. The resulting spectrum lands on the camera detector as a long line — slightly curved across the detector because of the optical geometry — with red on one end and blue on the other. Pulling the spectrum cleanly off the detector means tracing that curved line accurately and adding up the photons along it; the new version of this software is much better at "tracing" that line, even when it is faint or thin.
• The "Universal Translator" (Wavelength Calibration): A star's spectrum is like a barcode that tells us its temperature and chemistry. But the camera doesn't naturally know which colour is which; it only knows "pixel number 500." The old software tried to guess the colours by looking for specific lines, but it often got lost. The new software uses a "Template Matching" system. It's like having a master key: the software compares the messy data to a perfect "master barcode" it already knows, allowing it to instantly and accurately translate "pixel numbers" into "actual colours/wavelengths."
• Setting the Brightness (Flux Calibration): Finally, the software ensures the brightness is "true to life." It compares the star's light to a "Standard Star" — a celestial object whose brightness we already know perfectly — to make sure the final report isn't too bright or too dim.

Why does this matter?

Without these tools, the data from the TANSPEC telescope would be like a blurry, noisy, and incorrectly coloured spectrum. By upgrading these "digital cleaning kits," the researchers have made it much faster and much more accurate for astronomers to study the secrets of the universe — from the chemical makeup of distant stars to the atmospheres of planets orbiting other suns.

Technical Summary

Technical Summary: pyTANSPEC-v1.0 and HxRGproc

This paper details the development and implementation of upgraded data reduction software packages, pyTANSPEC-v1.0 and HxRGproc, designed for the TIFR-ARIES Near-Infrared Spectrometer (TANSPEC) mounted on the 3.6-m Devasthal Optical Telescope (DOT).

1. The Problem

TANSPEC is a unique near-infrared spectrograph-cum-imager (0.55–2.5 µm) that operates in two modes: Low Resolution (LR) and Cross-Dispersed (XD). Previous versions of the data reduction pipeline faced several critical limitations:

• Limited Mode Support: The previous pipeline only supported wavelength-calibrated XD mode spectra and was restricted to only two slit widths (0.5" and 1.0").
• Calibration Fragility: The old wavelength calibration method relied on the manual re-identification of emission lines, which failed when spectral lines became blended (common in LR mode or wide slits) or when the detector's aperture position shifted due to instrument mechanical changes over time.
• Detector Noise: The H2RG detector used by TANSPEC exhibits significant non-Gaussian noise, including non-linearity, bias fluctuations (1/f noise), and cosmic ray events, which require sophisticated pre-processing to ensure data integrity.

2. Methodology

The authors implemented a two-stage software solution:

A. Pre-processing with HxRGproc:
This package processes raw Non-Destructive Readout (NDR) frames to generate "slope images."

• Bias Correction: Uses reference pixels (which do not capture light) to subtract bias. To prevent increasing readout noise, the reference pixel counts are smoothed using a Savitzky-Golay (SG) filter (optimized with a window size of 439 pixels).
• Non-linearity Correction: Addresses the non-linear relationship between photon counts and electrical output. The authors used an inverse B-spline model (fitted via the FITPACK algorithm) to correct the signal.
• Cosmic Ray (CR) Removal: Employs a digital filter combining a first-difference filter and a Mexican hat filter to identify discontinuities in the signal ramp, subsequently estimating the true slope via a weighted average of the segments before and after the event.

B. Spectral Extraction with pyTANSPEC-v1.0:

• Extraction: Uses the SpectrumExtractor tool. The upgrade introduces second-order polynomial trace fitting (improving upon the previous first-order method) to better capture the curvature of the spectral orders.
• Wavelength Calibration: Replaces line-identification with a template-matching method using least-squares minimization. Observed lamp spectra (Argon and Neon) are matched against pre-calibrated templates, making the process robust against pixel shifts and line blending.
• Flux Calibration: Uses observations of the A2V-type standard star HD 37725. For XD mode, a polynomial fit per order is used; for LR mode, a median-filter approach is used (after masking telluric regions) to account for the instrument response.

3. Key Contributions

• Full Instrument Support: The pipeline now supports all available slits and both LR and XD modes.
• Robust Calibration Framework: The introduction of template matching for wavelength calibration and a standardized flux calibration task (Task 7).
• Enhanced Data Quality: Integration of HxRGproc into the TANSPEC server ensures that users receive pre-cleaned, non-linearity corrected data.
• Modular Design: The software is designed to be adaptable to other ground-based NIR instruments by swapping trace files and templates.

4. Results

• Noise Reduction: Bias correction significantly suppressed horizontal noise (by a factor of ~3.4), particularly in Channel 4.
• Flux Accuracy: Non-linearity correction resulted in a ~13% increase in extracted flux compared to uncorrected Quick Look Frames (QLF), demonstrating that previous data was systematically underestimated.
• Calibration Performance: The new wavelength calibration is faster (taking <10 seconds vs. ~1 minute) and more resilient to the observed cycle-to-cycle detector pixel shifts.
• Validation: Testing against the CALSPEC standard spectrum of HD 37725 showed that the extracted XD and LR spectra follow the expected spectral energy distribution (SED) accurately.

5. Significance

This work provides the astronomical community with a professional-grade, automated pipeline for TANSPEC data. By enabling reliable Low-Resolution spectroscopy, the upgrade opens new avenues for studying faint objects, exoplanet transmission spectroscopy, and spectral energy distributions. Furthermore, the software's modularity provides a blueprint for the reduction of other next-generation near-infrared instruments.