Return to Previous Section
Once one has extracted a 1-D spectrum of an emission-line exposure, one can use that to determine the wavelength calibration for spectra obtained with the same telescope pointing. To do so, one must identify the lines in the lamp spectrum. As an example of the process, I added two (2-D) lamp spectra that were obtained sequentially, with the telescope pointing at the zenith; an 8 second Hg lamp and a 2 minute Ne lamp. I then extracted a 1-D spectrum from a single row (row 125). I show the resulting spectrum in Figure 42. The vertical range has been clipped to show the weaker lines. Also, before generating the displayed spectrum, I flipped the data to produce a spectrum with positive dispersion (wavelength increasing with increasing pixel number).
One then compares the observed line list with a reference list of line wavelengths, and looks for matches that give self-consistent results. For this spectrum, there are a total of 31 line-matches with the HgNe line table. The line-matches are given in Table 8. These provide a set of measures to determine a relation between pixel number and wavelength for the spectrum. The simplest functional form for this is a linear fit. For such a fit, the slope is the dispersion (Angstroms per pixel). In general, there is enough flexure in the overall optical system that one can get substantially better results using higher-order fits. For this spectrum, I found that the residuals continued to improve up to a fifth-order fit. Figure 43 shows the wavelength-calibrated spectra using the linear and fifth order fits.
Shown in the manner, the two fits look indistinguishable. That is very reassuring, as it tells us that flexure-induced nonlinearities in the wavelength calibration are small when the telescope is pointed at the zenith. Figure 44 two spectral regions, with both fits plotted for each region. The first is in the middle of the optical part of the spectrum (5700 to 6500 Angstroms). The match between the two fits is, again, very good. The second region is well into the near infrared (9000 to 11200 Angstroms). Here the linear and polynomial fits diverge. This is unsurprizing, as the longest wavelength line used for the fit is the 8717 Angstrom Ne line. It is also unlikely to be a problem, as the atmosphere will probably preclude any useful astronomical observations from our site at such long wavelengths.
As an example of the details of finding a good polynomial fit, I show in Table 9 the results for the fits from first-order (linear) up to fifth-order. I note that the recovered dispersion of 4.3 Angstroms matches the specification from SBIG for the low-dispersion grating.
|Fit Std. Dev.||Dispersion||2nd Order||3rd Order||4th Order||5th Order|
I conducted similar tests on data taken with the high-dispersion grating on 22 July 2008. Here again, the quoted SBIG dispersion is recovered by our system. Table 10 shows the result of polynomial fits to an example high-dispersion HgNe lamp spectrum.
|Fit Std. Dev.||Dispersion||2nd Order||3rd Order||4th Order|
On 31 July 2008, I obtained a set of lamp spectra at both high and low dispersion, all with the telescope pointing to the zenith, using the wide slit. In Figure 45 I show an example of a summed HgNe spectrum from this set. It is a 1-D spectrum extracted with a 9-pixel aperture. The 2-D spectral image was rectified before extraction. The dispersion recovered for this spectrum is consistent with the nominal value (4.35 +/- 0.023 Angstroms). The fit deviation is much worse than for the narrow slit examples (~10 Angstroms). For high-dispersion spectra, the dispersion is also close to nominal, and the fit deviation is again much worse than for spectra obtained with the narrow slit (~2.5 Angstroms).
Procede to Next Subsection
Return to Main Line-Lamp Page
Return to Main Wavelength Calibration Page
Return to Outline