3 Plate measurements

3.1 The Tautenburg Plate Scanner

The Schmidt plates were completely digitised using the Tautenburg Plate Scanner (hereafter TPS). The TPS is a new fast plate measuring machine at the Thüringer Landessternwarte Tautenburg coming into operation during the last years. In the following, we shall briefly outline the basic concept and the main operational details of the TPS. A short description of the astrometric properties is given by Brunzendorf & Meusinger ([1998]).

The main components of the TPS comprise (1.) a movable X-Y plate carriage, (2.) a diffuse illumination screen below, and (3.) a direct CCD imaging system above the plate carriage. The TPS digitises plates by moving the plate through the optical path of the stationary imaging system. Plates up to 30cm $\times$ 30cm can be measured.

The X-Y plate carriage consists of two motor/encoder/stage units which permit independent motions along the X and Y directions. Prior to each scan, the X-Y carriage is moved into an appropriate position. The Y stage accelerates until it reaches the final velocity that will be kept throughout the scan. The begin of the data acquisition is triggered when the Y stage reaches the actual start position. The rms positional repeatability of each stage is 0.4$\mu$m. The stages have absolute systematic errors of up to 4 $\mu$m ($\rm rms = 2~\mu$m) with a scale length of $\sim10$cm. Currently, the stages are being upgraded with two linear encoders having an absolute accuracy better than $0.5~\mu$m over 30cm.

The part of the plate to be measured is backside-illuminated by a Fostec cold surface light source powered by a regulated 150W tungsten-halogen lamp via optical fibres. The lamp brightness is electronically stabilized within better than 1%. In addition, its actual flux is permanently monitored by means of a fibre optics that feeds light from the lamp directly onto a fraction of the detector array of the CCD (see below). This reference signal is used to correct the incoming data for remaining fluctuations of the lamp intensity. In this way, a photometric stability of better than 0.1% (0.0004D) over 24h is achieved.

A strip of the illuminated plate area is projected onto a $6\,000\times 1$ photosite array (CCD191 from Fairchild Weston Systems, Inc.) by a telecentric mapping lens system with unit magnification. The optics has a linear field diameter in the object plane of 60mm, a numerical aperture of 0.1, and a focal depth of 30$\mu$m. The telecentric projection has the important advantage of being less susceptible to scattered light, because any light transmitted through the emulsion contributes to forming a correct image of the plate, regardless of the origin of that light (e.g., Hambly et al. [1998]; Miller et al. [1992]).

The size of each CCD pixel is $\rm 10~\mu m\times 10~\mu$m, corresponding to $0\hbox{$.\!\!^{\prime\prime}$}5\times0\hbox{$.\!\!^{\prime\prime}$}5$on a Tautenburg Schmidt plate. Because 500 pixels are reserved for the reference signal (see above), the effective width of the measured strip on the plate amounts to 55mm. The CCD is operated in a continuous scan mode, i.e. the photon-generated electrons are accumulated and read out periodically. The analogue output signal is linearly amplified and fed to a 12bit (=4096 grey levels) analogue-to-digital converter. Thereafter, the digital signal is corrected for temporal light source intensity fluctuations (see above) and for spatial inhomogeneities of both the illumination system and the CCD pixel sensitivities ("flat field correction''). The final peak-to-digitisation noise ratio of the TPS data amounts to $\ge1000$.The obtained data are stored in a 16-bit FITS format file on the control computer. Optionally, the plate can be measured with reduced intensity resolution (8bit).

Before a scan is started, the optimal CCD integration time and the camera focus are to be determined. The integration time is adjusted to the transmittance of the plate background corresponding to a fraction of 0.75 of the maximum CCD intensity range. Automated focussing is performed at 8 different positions on the plate; the median focus value is adopted for the whole scan. Thanks to the telecentric projection (telecentric depth $\approx 1$cm) the astrometric and photometric bias caused by small focus deviations over the plate is negligible.

After every 4cm, the scanning process is interrupted for saving the data from RAM onto disk. This scanning interval is well-defined; it causes a random shift of the order of 1$\mu$m in the Y position of all objects within one scanning section with respect to the neighbouring sections. This shift can be easily evaluated from the positions of the objects in the overlap region of neighbouring strips.

In general, plates are completely scanned in a series of overlapping lanes ("strip mode'') by means of fully automatized control software. Alternatively, a number of subareas defined by a list of plate coordinates and area sizes can be measured ("batch mode''). A 24cm $\times$ 24cm Tautenburg Schmidt plate is digitised within typically less than three hours.

The scanner is operated by a DOS-PC with a Pentium 100 CPU, 64MBRAM and two 2GB hard disks for temporary storage of the incoming pixel data. Complete data files of the scanned plates are stored on DAT or CD-ROM. Final data reduction is done on remote workstations.

3.2 Plate digitisation and selection of galaxies

The 37 plates selected for further reduction (Table 1) were digitised in the "strip mode'' with overlaps of 10mm ($8\hbox{$.\mkern-4mu^\prime$}6$) between adjacent strips. Thanks to those overlaps, complete, non-truncated images are available for all galaxies. In a first run, digitisation was done in the 8 bit mode. Additionally, five selected plates were digitised later in the 12bit mode for photometric and astrometric purposes (see below).

The two deepest digitised plates (Nos. 8753 and 8788), taken under good seeing conditions ($1'' - 1\hbox{$.\!\!^{\prime\prime}$}5$), were independently surveyed by eye for galaxies. Only nonstellar objects found on both plates were selected. The final sample consists of 660 galaxies. The selection procedure may be biased against faint low surface brightness (LSB) galaxies as well as against very compact galaxies. Nevertheless, such a "by eye'' selection method is expected to be more complete than usual automatic object classification methods (cf. O'Neil et al. [1997], for the case of LSB galaxies).

On each digitised plate, the images of all 660 galaxies were extracted into separate frames of $200''\times200''$ size. For the more extended giant galaxy NGC1275, a larger frame of $400''\times400''$ was used. Each frame is centred on the deduced preliminary position of the galaxy's core.