Data Reduction
Observing Procedures
Real Time Extraction for CorMASS:

This page documents a crude real-time extraction algorithm for CorMASS spectra. The program, real.c, has the following features:

  • It assumes that each spectral order has a second-order functional shape on the array that is invarient with time and with the in-slit position of the source. These shapes are fit independently and the coefficients of the orders are inserted in the function order(order,y-position). Orders are numbered in the correct sense (i.e. the longest wavelength order visible on the array is the 2nd order).

  • The input file format is the standard CorMASS two-plane fits output.

  • The routine can operate on either a raw CorMASS fits image or can difference a pair of CorMASS fits images and operate on the result. For bright stars operating on a single image will produce a useful spectrum. For faint stars it is essential to subtract a sky image of equal integration time in order for the routine to find the spectrum and extract it with reasonable signal to noise.

    • real cm054
      would extract a spectrum from a single image

    • real cm054 cm055
      would difference two frames before extraction.

    Differencing two spectra at different slit positions could produce a satisfying result if the spectra are sufficiently separate so that the "negative" spectrum falls in the "sky" aperture (see below).

  • The routine first searches for the in-slit spectrum location using the 4th spectral order. It simply determines which in-slit position has the peak flux when summed along the spectrum.

  • Once the in-slit offset is known the routine extracts the sum of 5 pixels along the slit for each x-pixel location along the order for each order specified (currently 3rd through 6th order).

  • Sky flux is summed from five pixels not contained in the spectral extraction region. The sky flux region is 10 pixels wide centered on the slit (i.e. spectrum pixels falling in this region are excluded from the sum).

  • The routine produces a diagnostic image called "diag" which is a raw short integer of read1 and read2. Read1 is set to zero. Read2 equal to the original Read2-Read1 image with the spectral orders offset by (order*1000) and the sky region offset by +9999. The routine mkfits.c will convert this diagnostic image to a standard fits image. Here is an example for a bright star.

  • The output is written to a file cm###.sp where ### is the frame number of the "signal" image. This file has four columns
        1             2          3           4
    pixel_number  spec_flux  sky flux  spec_minus_sky
    pixel number = order*256 + x-pixel-position
  • The supermongo script file "sm" contains some macros to quickly plot the results of the program.
    cm ###  -- will plot a spectral extraction from 
               four orders of a single image  (I can't 
               wait to get a formula that converts these 
               pixel numbers to wavelength...)
    cm2 ### ### -- will difference two images and plot.
    sky ###  -- will plot sky flux from a single frame.

  • Here are some examples of what the program can do. The plots cover orders 3 (top) through 6 (bottom). Short wavelengths are plotted on the left. Order 3 is basically the H-band. Order 4 contains the J-band but extends usefully to 1.0 microns at the right of the plot. Order 5 covers .83 to 1.15um. Order 6 covers .75 to .95um.

    • A whirlgif of HD161903 extracted at the top, middle, and bottom of the slit. The routine gives a consistent answer. The horizontal shift is presumably due to a shift of the position of the source within the slit.

    • T-Tauri - look at those emission lines!

    • VB10 - note the Ti0 and other broad features in the 6th order. Here's a blink between VB10 and HD161903.

    • WR2 - One of J.D.'s Wolf-Rayet stars.

    • ??? - Mike's mystery source. This was supposed to be a carbon star candidate in the Galactic plane. Looks more like an HII region.

    • Red Quasar - The bright emission line in the bottom order is redshifted H-alpha.

    • T-dwarf - And here it is blinked against HD162208.

    • New - blinked against stellar spectrum. - blinked against previous T-dwarf.

  • Note that the spectrum of VB10 is so good because we did a single sky integration off source to remove the array bias and sky lines to first order. It is possible that we can difference two slit positions and obtain an equally good - if not better - result. If the "negative" image falls outside the spectral extraction region then it will be subtracted (and thus added) from the positive signal and we will reap a double benefit.
  • Sky lines can also be processed out by removing a column trimmed average along the slit making it easier to identify and extract a faint spectrum. Here is code which implements this trimmed average removal prior to spectral extraction. It does a much better job of extracting a faint spectrum from a single frame. I still find that I get a better looking spectrum if I difference two frames with the spectrum offset along the slit.

For additional information please e-mail Mike Skrutskie: mfs4n@virginia.edu.

UVA Astro
Last modified Monday, June 13, 2005
Copyright © 1996-2005 University of Virginia Department of Astronomy.
All rights reserved. Maintained by the webmaster