Wavelength calibration

My website will not be a textbook on amateur spectroscopy. It will only occasionally give hints, tips and examples, which are difficult to find elsewhere. So I do not here go into every detail of the wavelength calibration, especially since there are many methods that one can ever choose in function of the objectives of the measurement and evaluation (required precision). Various analysis routines implement them in various forms.

Theme: temporal drifting of the wavelength scale

If an accurate calibration of the spectra is important, eg in the measurement of absolute radial velocities, is taken before and after every shot of the object a calibration frame (in case of slit spectrographs). With my LHIRES III I use the built-in neon lamp or a external neon lamp.

Often, one has at night continuously falling temperature, which minimally changes the geometry of the equipment including the spectrograph, leading to a thermally induced drift of the diffracted light path, and thus leads to shifts of the wavelength scale on the CCD.

The following tests I have done in preparing a observation campaign for the Periastron of WR 140 and its companion O4 star (September / October 2008). A friend borrowed his LHIRES III for the 4 months of a observation campaign at the MONS telescope in Tenerife. For the campaign, a reducer (Meade 0.63) and a flip mirror (Meade # 647) had to be adjusted at Lhires including observation optics in collaboration with my friend Dr. Berthold Stober.


The accompanying graph shows the same excerpts of two neon spectra. There was been no change in the grating position (grating 1200 g / mm) at this observation of WR 140. The time interval between the 2 neon spectra accounted 20 minutes (before and after a WR 140 frame) . The neon emission lines was modeled as a Gaussian profile (in MIDAS a standard method). The difference is (1602.476 – 1605.211) = -2.735 pixels.

This difference can be composed by different components:

  1. Variable geometric relationships that arise in the undesired displacements of the neon lamp (mechanical tolerances of the foldaway mechanism).
  2. Drift caused by thermal changes.
  3. Drift as a result of bending of the spectrograph on the moving telescope.

The contribution of this effects we can see from the result shown directly.


To measure the drift due to thermal and mechanical effects based on bending influences were six 20-minutes recordings of WR 140 in the wavelength range of the NaD doublet had been calibrated with the same at the beginning of the series shot neon spectrum. Then the sodium lines by Gaussian fitting were modeled and determined their center. An evaluation of the time dependence of the minima of the Na lines (5896 and 5889 angstrom) shows the diagram on the left side. The equations of regression lines are registered.

It follows a trend of about 0.006 pixels / min, which corresponds to about 0.12 pixels within 20 min. The difference between the two neon frames of -2.735 pixel (see above) is thus accounted for only a small part by the thermal trend or bending phenomenons (-0.12 pixels / 20 min). The rest must be attributed to mechanical instabilities of the foldingaway mechanism of the neon lamp.


In comparison to my own LHIRES III, the scattering of neon lines from about 2.5 pixels was too high. So I had improved the folding mechanism of the neon lamp, and made a new series on the reproducibility of neon calibration spectra. First, I created an artificial temperature gradient, in which I heated up the Lhires in my domicile at around 21 ° C and then underwent a measurement directly in the observatory at 6 ° C ambient temperature . That led to a strong trend of about -0.05 Pix / min. After a two-hour cooling of the spectrograph, I repeated the series in the night, remaining in the closed observatory dome (heating effect by body heat, observatory temperature approx. 6°C). The result was the opposite trend, but much weaker (+0.02 Pix / min). I also noticed that the heat of the neon lamp already leads in a drift. One should therefore fold back the neon lamp at the end of neon exposure. And then turn off the neon lamp.

The third series lasted for the next day about 6 hours (10 to 16 clock). At intervals of 15 to 60 minutes, I picked up the same neon spectrum at unchanged grating position and standing spectrograph, with the neon lamp was in service always only just the short moments above the slit. I myself left the dome immediately after the recording . On this cloudy day even a small temperature change was recorded in the measurement period.

It is only to see a minor trend over the 6 hours for the two evaluated neon emission lines of the neon lamp. The scatter is rather random.The standard deviation is rounded to 0.23 pixels. With the dispersion of 0.38 angstrom / pixel when using a 1200 g / mm grating (CCD camera = Sigma 6303 ME with 9µm x 9µm pixels) for the WR140 campaign thus a standard calibration deviation of about 0.23 x 0.38 = 0.08 angstrom is expected (if no other effects play a role).Neon_Repro.gif


Using an external neon lamp (before the telescope)

My slit spectrograph LHIRES III from the company Shelyak has an internal neon lamp, which can be folded in front of the slit. Here the slit is very well lit and the collimator also, because the light is emitted from the lamp around and does not enclose a defined aperture ratio. The situation is different when a neon lamp directly (without further diffuser) will be held in front of the telescope aperture. Only that light reaches the spectrograph, which is within the viewing angle of the telescope. This is on my C14 a solid angle of about 5 °. Since the lamp itself has a finite extent (in the light of my lamp the electrodes have a size of 30mm x 40mm) is only a small part of the telescope aperture illuminated according to the geometric size of the Lamp. In my case, will be led to the slit only a wave front from approximately 30mm x 40mm . Star light, however evenly lit in form of a plane wave front,the entire aperture of the telescope. Therefore it is expected that the neon lines on the CCD will depend on the position of the neon lamp above the telescope aperture.

Here the neon lamp used by me (also known as the Beehive lamp). It is relatively bright.


That’s the theoretical consideration. Better is the measurement of such effects.

To this end, I waited in the evening a twilight period in which the light of the uniformly overcast sky just bright enough to see, while illuminating the telescope aperture with light from the sky and the external neon lamp, both spectra: the spectrum of diffuse daylight and the light (in the recording they appears white) emission lines of neon.

1800g/mm grid, 38 µm slit width, f = 200mm collimator (Lhires III), CCD KAF 1603ME, Dispersion = 0.20 angstrom / pixel.K640_Aufnahme.JPG

An example is evaluated to see below. The red marked spectrum is recorded with the internal neon lamp, the black one is made with the external neon lamp, while it was moving in circles during the recording about the telescope aperture to simulate a uniform illumination of the aperture.

Plotted are the integrated intensities of 10 pixel rows (the same) for all recordings. At Pixel 300 is the H alpha line to see. The sharp absorption lines are mostly terrestrial water lines. The 5 neon lines are superimposed on the twilight spectrum.

At first sight, both spectra are identical. The neons appear on the same place.


Here are now carried out detailed analysis.

The two emission lines are marked black for internal and external lamp, the latter having been led in circles over the entire telescope aperture (60s exposure). They are not shifted and have the same width at half maximum FWHM = 5.5 Pix.

PS: translations (German/English)
kreisend = circling
intern = using the internal neon lamp
quer = across
laengs = along


The other spectra (green, red) were taken with fixed lamp position, with the long axis of the slit was taken as a reference for the position (see graph below). The green spectrum was recorded in „Längs1“ lamp position. The two Reds in position „Quer1“ and „Quer2“. The FWHM’s of the recorded spectra with fixed lamp position was only 3.0 to 3.6 pixel, which corresponds almost exactly to the slit width of 38 μm (9×9µm ² pixel). The spectrum taken in „Längs1“ position is not shifted compared to the internal or circular spectra (differences of the lines with Gaussian functions modeled approximately 0.1 pixels). In contrast, the lamp moved into the two lateral (across/quer) positions the spectra are symmetrically shifted by about + – 2 pixels!

From these findings, I draw the following conclusions:

  • An external neon (or other) calibration lamp is applicable.
  • Either the external lamp is fixed in the longitudinal direction of the slit or it will moved in circles over the telescope aperture, so that no preferred direction relative to the slit axis results.
  • Lamp positions collateral to the slit lead to significant but constant shifts of the calibration spectrum.