Guide

How to use the spectrograph & analysis platform

A complete walkthrough from assembling the device to capturing spectra, plus how to use the analysis platform

01
Assembly
02
Settings
03
Capture
04
Upload
05
Analysis
Step 1: Assemble the device
Mount the objective prism in front of the telescope

1Mount the prism

Using the 3D-printed frame, fix the 15° apex, 60 mm-diameter circular prism in front of the telescope objective. Make sure the prism apex points downward.

2Connect the camera

Connect the QHYCCD MiniCAM8M cooled monochrome astronomy camera to the telescope’s focal plane, ensure the connection is secure, and enable cooling to reduce sensor noise.

3Set up the mount

Set up the telescope on the EQ6-Pro mount and complete polar alignment. The prism deflects light, so tilt the telescope along the prism apex direction to compensate for the image offset.

4Prepare the software

Use TheSkyX Professional Edition to find your target, and SharpCap to control the camera during capture.

Step 2: Capture settings
Adjust camera parameters to the target star’s brightness
Capture mode
Video / continuous
Gain
Moderate (10-20 recommended)
Exposure time
Match target brightness
Cooling temp
20-30°C below ambient
Target selection
Stars brighter than mag 6
Step 3: Capture the stellar spectrum
Record spectral video with SharpCap

Capture tips

  • Locate the target in TheSkyX, making sure the star still images after the prism deflection
  • Use SharpCap’s live view to confirm the spectrum is visible in the frame
  • The spectrum should appear as a horizontal dispersion band, with color shifting from blue to red
  • Capture at least a few hundred frames; stacking them later improves SNR
  • Record the target star’s name and capture time for later analysis
Step 4: Process & upload data
Turn the video into a spectrum image, then upload it for analysis
1

Frame stacking

Use stacking software (e.g. AutoStakkert, PixInsight) to stack the video frames and reduce noise, producing a single spectrum image (PNG or JPG).

2

Spectrum extraction

Rotate the stacked image to horizontal and crop out the usable spectral region. Make sure the full spectrum is visible, blue end on the left and red end on the right.

3

Upload & analyze

Go to the “Analyze” page, upload the processed spectrum image, and click “Run processing”. The system runs the pipeline and outputs the classification automatically.

Python data-processing toolchain
The processing programs developed alongside the paper, free to download

Below are the Python data-processing programs developed in this research, listed in processing order. Each handles one stage, together covering the full pipeline from raw video to spectral curve.

avi-tif.py— Frame stacking

Reads SER/AVI astronomy video and stacks frames to reduce noise, outputting a single TIFF spectrum image. Supports auto-alignment and bad-frame rejection.

Input: video file (.ser/.avi)Output: stacked single-frame TIFF
rotate_to_aclinic.py— Rotation correction

Detects the spectral band orientation and auto-rotates it to horizontal. Uses Hough-transform line-angle detection to ensure blue is on the left and red on the right.

Input: stacked spectrum imageOutput: horizontal spectrum image
getcenterpix.py— Central-line extraction

Sums vertically across the spectrum to extract the 1-D intensity profile of the central line. Automatically locates the band center.

Input: horizontal spectrum imageOutput: 1-D intensity array + stretched preview
positioning.py— Wavelength alignment

Interactive tool that calibrates wavelength by picking a reference point (e.g. a known absorption line), mapping pixel coordinates to physical wavelength.

Input: 1-D spectrum + reference coordinatesOutput: wavelength-calibrated spectral curve
drawcurve.py— Spectral curve plotting

Plots the calibrated spectral data into a standard spectral curve, producing the final spectrum image ready for upload and analysis.

Input: wavelength-calibrated spectrum dataOutput: standard spectral curve (PNG)
Model programs (experimental)

The two programs below are used to build and validate the spectral classification model. They have been ported to TypeScript and can be explored interactively on the “AI Lab” page:

gen-spectro2.py

Standard spectrum generator: Planck radiation + Sellmeier dispersion + absorption-line simulation

cos_similarity.py

Spectral similarity analysis: cosine similarity + self-convolution + line detection

Step 5: Review the analysis results
The system automatically outputs the following information
Spectral type

Stellar spectral type under the Harvard scheme (O-B-A-F-G-K-M)

Surface temperature

Effective temperature fitted from the continuum shape (Kelvin)

Stellar properties

Estimated mass, luminosity and radius ranges for the matched type

Detected lines

Detected absorption lines, each labeled with the likely element

Match scores

Per-type match scores combining temperature and spectral lines

Key features

Characteristic features of the matched type (He I, Balmer, Ca II, TiO, etc.)

Important notes
Ensuring reliable analysis results
  • •The uploaded spectrum image should be clear — avoid over- or under-exposure
  • •Make sure the spectrum is oriented correctly (blue on the left, red on the right); rotate the image first if needed
  • •Results are for reference only; for science-grade precision use professional software for calibration
  • •Upload only spectra captured with this project’s device, since the calibration is specific to it
  • •For low-confidence results, capture multiple times and cross-check