The observations were performed in different observing runs carried
out in 1994-1995 at two observatories: the Observatorio de Calar
Alto (Almerıa, Spain) and the Sutherland Observatory (South Africa).
Observations at Calar Alto were made in May and
December 1994 using the Coudé spectrograph
on the 2.2 m telescope. The wavelength
covers 
(
). The
resolving power
was 22000 which corresponds to a reciprocal spectral dispersion
of 0.20 Å/pixel at 
. Observations at Sutherland were made in
July 1995 using the Image Tube Spectrograph and the RPCS detector on the 1.9 m
telescope. The spectra cover the wavelength
range 
. The resolving power was 19000
which
corresponds to a
reciprocal spectral dispersion of 0.23 Å/pixel.
The exposure times in all the runs were limited to 15 minutes for the RR Lyraes in order to prevent ``phase blurring". In general, three spectra were obtained for each RR Lyrae. The first spectrum was arbitrarily assigned as phase zero and then, using the published periods from the GCVS, two further spectra were obtained separated by plus or minus a third of a cycle from the first spectrum.
The spectra were reduced using the FIGARO reduction package: the
raw spectra were bias-subtracted, flatfield-corrected (two or more master
flatfield exposure were taken every night) and background-subtracted
before the spectral extraction. The wavelength calibration of the stellar
spectra was done with thorium-argon comparison spectra. Because of the variations in the velocity zero
point of the Image Tube Spectrograph at Sutherland (probably due to a
mechanical slippage of the grating mount), arcs were taken before and after
every observation. In the Calar Alto runs, arcs were
taken at the beginning and the end of every night. Polynomial calibrations
of wavelength as a function of pixel number were calculated for each
comparison spectrum. A fourth order polynomial fit was used for Sutherland
spectra whereas a second order polynomial fit was found adequate for the
Calar Alto sample. In both cases a precision better than 0.04 Å was
achieved which corresponds to 
at 4200 Å.
The continuum windows were
chosen with the aid of the Atlas of Procyon (Griffin & Griffin
1979) and the continuum level calculated by fitting a second-degree
polynomial. The normalized spectra were then derived by dividing the
extracted spectra by this continuum level. The signal-to-noise ratio of the
spectra varied between 5 and 25 depending on the magnitude of the star and
the quality of the seeing. In addition, several IAU radial
velocity standard stars were observed each night to measure the
zero-point offset in radial velocity of our measurements.