Description

Over 16,800 high-dispersion spectrograms were exposed at the coudé focus of the 1.2-m DAO telescope and McKellar spectrograph between 1962 and 2000, employing a variety of grating and camera configurations. A full description of the spectrograph has been given by Richardson (1968).

Over 93,000 spectroscopic plates were secured at the Cassegrain focus of the DAO 1.8-m telescope and spectrograph between 1918 and 1984. Before 1947 the dispersing element was a prism. A Littrow (grating) spectrograph (see Beals, Petrie and McKellar 1946) was introduced in 1947 (plate 32276) and in turn that was replaced by a conventional Cassegrain grating spectrograph in 1967, for plate 63489.

The very great majority of those plates is now in the HIA/DAO plate archive, and a programme to digitize them with the modified in-house PDS has recently commenced.

The various spectrographs required plates of different shapes and sizes. Figures 1 through 4 illustrate the four basic types. Each shows a spectrum of a single object, flanked on both sides by a wavelength-reference spectrum (generally an iron-argon source), and includes an exposure of a multi-stepped source to provide intensity-calibration information.

Note: Details of the spectrograph "modes" mentioned in this document can be found here.

Figure 1: Spectrogram of the near UV of VV Cep, exposed on 1972 August 31 with the McKellar spectrograph in the "32122" mode. Original dispersion: 4.9Å/mm. The double absorption feature in the centre is the Ca II H–Hε blend; the Ca II K line is well to its left.

Figure 2: Spectrogram of the blue region of the Ap star 17 Com, exposed on 1947 March 11 with the 1.8-m telescope and original prism spectrograph in the "IIL" mode. The prominent absorption features are Hδ and Hγ.

"GII" mode.

Figure 3: Spectrogram of the Hα region of ε Aur, exposed on 1956 September 14 with the 1.8-m telescope and Littrow spectrograph in the

Figure 4: Spectrogram of the Ap star 73 Dra, exposed on 1974 August 7 with the 1.2-m telescope and McKellar spectrograph in the "9682" configuration. The prominent stellar absorption lines are the Balmer series H12 – H10, at λλ3750, 3771 and 3798Å. Original dispersion: 2.4Å/mm.

Digitizing Procedures

Plate Scanning

The glass side of the plate is cleaned thoroughly with soft linen cloth, and the emulsion side is gently brushed with a camel-hair brush. The plate is taped onto the glass platten of the PDS, and the platten is rotated until the spectrum is correctly aligned with the X-direction of the travel of the carriage. An appropriate analysing aperture is selected. Those supplied have rather coarse intervals, so choice is limited; since the PDS step-interval is 6μ, the usual choice is a slit-width of 8.5μ.

In normal operation the PDS scans in X, and makes incremental steps in Y before commencing a fresh scan. The Y-increments are supplied as adjustable parameters by the driving software. Both axes are fitted with encoders so that the carriage can return to a known point for each fresh scan. If (as is almost always the case with high-dispersion spectra) the spectrum to be scanned is taller than the height of the slit, a raster of scans is programmed. It may or may not be considered important to incorporate a small overlap between adjacent rasters. Scans of a single plate include up to 5 different sections: the spectrum itself, a wavelength-reference arc on either side of the star spectrum, and "clear plate" (empty background), also on either side but as close as can reasonably be managed to the star spectrum. The clear-plate regions are normally scanned with a raster of at least 3, while the arcs are sampled along their entire height – which can often be much greater than that of the star spectrum.

The plate is then rotated through 90° in order to make a set of scans through the photometric calibration exposures at pre-determined wavelengths. (see Figs. 1 through 4). The recommended intervals for photometric sampling are 200Å for high-dispersion plates (better than, say, 3Å/mm), 400Å for more moderate-dispersion plates, or only once at a representative wavelength in the case of low-dispersion ones.

Preliminary Data Reductions

A PDS measures the amount of light transmitted through the plate at every given (quantified) step, and records it as an output voltage in proportion to the logarithm of the instantaneous input current received from a photomultiplier detector. The first step in the data reduction is therefore to decode the logarithms, yielding an output in terms of plate transmission. The next step is to sum the rasters of each different sections of the plate, thus collapsing the 2-D images to 1-D ones.

Wavelength Scales

The wavelength scale on the plate is established by cross-correlating the 1-D arc spectrum with a set of laboratory air wavelengths and relative positions. A reference pixel, obtained by identifying a stellar or arc feature, is determined and the software stretches or shrinks the relative positions of the laboratory features until an obviously corresponding pattern is recognized and the linear dispersion can be checked. Both arcs are thus treated, and the mean positions of each pair of lines are determined.

The same wavelength solution is then applied to the stellar spectrum, and the output is re-sampled at the desired intervals: 10 mÅ is usually chosen for high-dispersion spectra and 50 mÅ for all others. Users should note that the wavelength scale included in the final FITS product has not been corrected for the intrinsic radial veocity of the star and a heliocentric velocity appropriate to the time of the observation has also not been applied. For convenience, however, the value of the latter has been included in the FITS header.

Figure 5: Iron-Argon spectrum, exposed with the 1.2-m telescope and McKellar spectrograph in the "32121" mode.

Background Correction

A photographic spectrogram has its own background level, against which all measurements are made. That background (loosely referred to as "clear plate") has therefore to be removed in order to establish a zero level for the star spectrum. The clear plate usually exhibits a slowly-varying pattern, and (under the reasonable assumption that the same variations also affect the star spectrum) those variations (and their absolute level) are determined by fitting a smooth but low-order function to each 1-D clear-plate scan. In addition, any prominent blemishes, arising (e.g.) from dirt on the plate or pinpricks in the emulsion, are revealed as upward or downward spikes and are removed at this stage. The two clear-plate functions are combined to yield a mean function, which is then divided into the stellar spectrum, yielding an output of wavelength versus transmission.

Direct-intensity Calibration

A photographic emulsion has a non-linear response to light. Those properties have been well-documented and thoroughly understood since they were first explained by Hürter & Driffield (1890).

Appropriate provision for calibrating each plate is an essential prerequisite for deriving intensity information of any reasonable quality. An exception to that rule has sometimes been adopted by observers who require only radial-velocity measurements, since positional information is obviously unaffected by the intensities of the lines. (Unfortunately, archived material with those deficiencies has limited worth for other studies).

The earliest intensity-calibration exposures took the form of a pattern of circular spots, created by shining a uniform light through holes of different but known diameters. An improvement was the 'calibration wedge', involving a spectrograph slit with a wedge-shaped aperture, whose width (controlling the incident intensity) at a given point was determined by the distance from a fiducial point, effectively the 'closing point' of the wedge. Later variants designed some form of stepped slit, the most reliable of all incorporating blanks between the steps so that the plate background could be carefully monitored. At the DAO a calibration exposure was created by shining a light through a rotating stepped sector; in its earliest form adjacent sectors represented a step of 0.1 in log I, but in a later incarnation that step was modified to 0.2. Each collapsed 1-D scan through the calibration strips is divided by the local clear-plate (determined from the ends of the scans), and the height of each strip is measured and paired with its corresponding log I value. A plot of strip height against log I reveals the "H–D curve" that is characteristic for that particular plate; the curve thus established is then used to convert each transmission value in the star spectrum into direct intensity. Normalizing may be applied to keep the output within a manageable numerical range.

Conversion to FITS Files

All relevant parameters about the spectrum are retrieved directly from the logbook and/or plate envelope, or are derived from them, and added to the header of a FITS file. The stellar spectrum and both comparison arc spectra are written to the same FITS file using IRAF's "multi-spec" format; the stellar spectrum is in aperture 1 and the comparison arcs are located in apertures 2 and 3.

Figure 6: Tracing through a calibration-strip exposure, from a plate exposed with the 1.2-m telescope and McKellar spectrograph in the "9682" mode.

Figure 7: Fully-calibrated spectrum of VV Cep corresponding to Figure 1. The normalization has been arbitrarily adjusted in order to keep the emission lines on scale.

Figure 8: Fully-callibrated spectrum of 17 Com corresponding to Figure 2.

Figure 9: Fully-callibrated spectrum of ε Aur corresponding to Figure 3.

Figure 10: Fully-callibrated spectrum of 73 Dra corresponding to Figure 4.