4 Data reduction

The data reduction was performed using ESO MIDAS SW (Banse et al. 1983). All frames were filtered to eliminate the cosmic ray events and were corrected for bias and flat field. Each spectrum was individually extracted from the science frame, and the same region of the CCD was extracted from the wavelength calibration frame. The background was defined on both sides of the science target if possible and an average background spectrum was subtracted. This simple procedure was justified, since the sky and interstellar background was homogeneous along the slits for all targets, except for the slits with G96 and K255 which showed a rather inhomogeneous H$\alpha$ background with respect to other background lines. The nebulous objects K799 and K4153 also showed inhomogeneous H$\alpha$ and [NII] line strengths, most probably due to a spatially resolved circumstellar shell. There were only three Neon lines in the short wavelength range of the calibration frame: $\lambda$ 6506.5278, 6532.8799, and 6598.9531. We also collected calibration frames with the two H$\alpha$ interference filters H_Alpha/2500+60 and H_Alpha/4500+61 (red shifted with respect to $\lambda_0 = 6563$ Å by 2500 and 4500 kms^-1 respectively) to collect more calibration lines. It turned out that the three Neon lines in the H_Alpha+59 filter band were sufficient to define a linear wavelength dispersion relation. Spectra were transformed into a velocity scale. A heliocentric correction of $V_{{\rm earth}}=-12.6$ kms^-1 was applied. The short wavelength range spectra of G12, G38, G64, G84 G96 and W2:79 suffer from a bad CCD column, accidently located close to H$\alpha$.