B. Calibration from Telluric Lines


Emission-line lamps can provide an excellent set of data to determine a wavelength transformation for an optical spectrum, however there are two systematic effects that can have a significant effect on the zero-point of the resulting transformation. First, the lamps are not taken at the same altitude and azimuth as the sky observation; they are usually taken just before and after. Second, the optical path from the lamps to the CCD differs from that of the light from the sky. A common method for addressing these issues is to determine a secondary zero-point correction from telluric (or night-sky) lines. These are emission lines from the sky, due either to the physics of the atmosphere or to parasitic light from man-made light sources. These lines have well known wavelengths, and are observed with exactly the same telescope pointing and optical path as light from the target of the observation. As night-sky lines can also interfere with light from the target, one can think of this as turning an unavoidable problem into a useful resource. It happens that the night sky from Andreas Observatory is fairly clean unless the Soccer field lights are on, but deep integrations show at least a few lines.

On the night of 4 June 2009 (UT 5 June 2009), A. Tammour and I obtained data on the night sky to study the magnitude of any potential zero-point shift from the emission-line lamp observations. These data were obtained with the wide slit, the low-dispersion grating, and the focal reducing lens. The detector temperature was T ~ -10 degrees C. We obtained a series of bias, dark and lamp-flat exposures, and used these to reduce the emission-line and sky exposures in the standard way, as described in the earlier sections of this manual. We also obtained two exposures of the night sky with exposure times of 20 and 30 minutes. These exposures were bracketed by emission-line lamp exposures in the following pattern: Hg -- Ne -- Sky -- Ne -- Hg. After downloading the first (20 minute) sky exposure, we decided to increase the second exposure to 30 minutes. Both exposures were taken with the telescope tracking across the meridian at zenith.

The consecutive pairs of emission-line lamps were added together, giving us four combined HgNe lamp frames. For each of these, 150 rows (51--200) were mashed to produce a 1-D spectrum. Figure 46 shows one such spectrum as an example. The identified lines allowed for a fourth-order polynomial wavelength solution that was then applied to the sky spectrum. Figure 47 shows the resulting wavelength-calibrated sky spectrum.

  • Figure 46: 1-D HgNe spectrum extracted from a 150-pixel aperture of a combined 2-D spectral image, taken at low dispersion, through the wide slit.

  • Figure 47: Wavelength-calibrated 1-D night-sky spectrum extracted from a 150-pixel aperture of the 2-D spectral image, taken at low dispersion, through the wide slit, with an exposure time of 30 minutes.

    There are a number of identifiable emission-line features in this spectrum that correspond to well-known night-sky lines. Table 12 shows the features identified, their reference wavelengths, their measured wavelengths from this spectrum, and the resulting wavelength shift for each line. A simple average of these values gives a resulting zero-point shift of

    Table 12 - Night Sky Line Wavelength Shifts
    Feature Reference Wavelength Measured Wavelength Wavelength Shift
    Angstroms Angstroms Angstroms
    Na I 4980.7 4995.4 -14.7
    Hg 5461.0 5481.0 -20.0
    [O I] 5577.0 5608.4 -31.4
    Na I 5685.8 5707.0 -21.2
    Hg 5780.5 5805.6 -25.1
    Na I 5893.0 5909.0 -16.0
    Na I 6157.5 6157.9 -0.4
    [O I] 6300.0 6312.9 -12.6

    Recalling that the spectral resolution in this mode is ~40 Angstroms, this is a fairly small shift, and is only significant at about the 2-sigma level. It is worth noting that this test was done near the zenith. Large zenith angles could introduce more significant zero-point shifts.

    On the night of 6 August 2012 (UT 7 August 2012) I obtained further data on the night sky. These data were obtained with the narrow slit, the low-dispersion grating, and the focal reducing lens. The detector temperature was T ~ -7 degrees C. I obtained a series of bias, dark and lamp-flat exposures, and used these to reduce the emission-line and sky exposures in the standard way, as described in the earlier sections of this manual. I also obtained three exposures of the night sky with exposure times of 20 minutes. These exposures were bracketed by emission-line lamp exposures in the same pattern as our earlier night-sky test. All three exposures of the night sky were taken with the telescope tracking across the meridian at zenith.

    The consecutive pairs of emission-line lamps were added together, as in the previous night-sky test. For these data, I extracted the full slit range of 250 rows, mashed to produce a 1-D spectrum. Figure 48 shows one such spectrum as an example. The identified lines allowed for a fourth-order polynomial wavelength solution that was then applied to the sky spectrum. Figure 49 shows the resulting linearized, wavelength-calibrated sky spectrum.

  • Figure 48: 1-D HgNe spectrum extracted from a 250-pixel aperture of a combined 2-D spectral image, taken at low dispersion, through the narrow slit.

  • Figure 49: Wavelength-calibrated 1-D night-sky spectrum extracted from a 250-pixel aperture of the 2-D spectral image, taken at low dispersion, through the narrow slit, with an exposure time of 20 minutes.

    Note that there is only one clean, well-detected line in this 20 minute exposure. One of the remaining two night-sky exposures was contaminated by residual charge from the emission-line lamps, and was thus discarded. The third night-sky exposure showed the same general features as that shown in Fig. 49. The one good line has a wavelength of 5758 Angstroms, and 5759 Angstroms in the two spectra. Working on the assumption that this line is Hg 5781 Angstroms implies zero-point shifts of +22 and +21 Angstroms, or about a 2-sigma offset, given a nominal spectral resolution of 10 Angstroms. In Figure 50, I show the merge of the two 20 minute exposures smoothed with a 3-pixel kernel. The SNR is improved over the single exposure, the broad NaI feature around 6000 Angstroms is more clear, and the NaI 5686 Angstrom line is now visible, but no other features are obvious.

  • Figure 50: Merged, smoothed, wavelength-calibrated 1-D night-sky spectrum with 40 minutes total integration time, taken at low dispersion, through the narrow slit.}

    We have only obtained data with the low-dispersion grating. Given the low intensity of most sky lines seen from Andreas Observatory, tests at high dispersion are prohibative. It may be that useful data at high dispersion will be obtained in the future in the course of normal observing.


    Procede to Next Section

    Return to Wavelength Calibration Page

    Return to Outline


    Updated: 2013 April 10 [pbe]