Table 1 lists the program stars and summarizes their most relevant stellar properties taken either from the recent literature or from the second edition of the catalog of Chromospherically Active Binary Stars (CABS, Strassmeier et al. [1993]) if appropriate. Table 2 identifies the comparison and the check stars, the observing interval, whether a neutral density filter was needed or not, the useable number of differential data points - each the mean of three readings on the variable and the comparison -, and which telescope was used.

The individual panels in Fig. 1 through Fig. 45 present the
photometric data in one bandpass (usually
*V* or *y*), the periodogram from these data, and the phased seasonal light
curve from the best-fit period. We emphasize that the sometimes abnormally
large scatter in the phase plots is almost exclusively due to intrinsic spot
changes and not due to instrumental scatter.

Figure 1:
The y-light variations of HR 5B in 1996/97 (left panel).
The right panel is the periodogram showing the adopted significance level at
a signal-to-noise ratio of 4 according to the criteria of Breger et al.
([1993]). No clear period was found for HR 5.
For individual comments and discussions, see Sect. 4 |

Figure 2:
LN Peg (=SAO 91772). The right panel shows the
phased light curve from the best-fit period listed in Table 1.
The corresponding frequency peak is indicated in the header of the
plot along with the Fourier (half) amplitude and the phase shift
from time T_{0}=0.
Note that the phase plots show the combined data from the entire observing
season and its scatter is usually due to intrinsic amplitude changes |

Table 1 includes the results from our period analysis. We applied
a menu-driven program that performs a multiple frequency search through
Fourier transforms with a non-linear least-squares minimization of the
residuals (Sperl [1998]). The Fourier search range included a
large number of frequencies up to the Nyquist frequency with a
frequency spacing optimized for each individual data set. In most situations
the frequency with the highest amplitude was adopted but in some cases,
e.g. when the light curve appeared double humped,
was used. The best fits are
determined by minimizing the squares of the residuals between trial fits
and measurements, and is done with program `curfit` (Bevington
[1969]). The function that is fit to the data is

(1) |

where

To judge the significance of certain frequency peaks we overplot a running
mean of the frequency distribution for a signal-to-noise (*S*/*N*) ratio of 4:1
which was found empirically by Breger et al. ([1993]). From
numerical simulations with varying amounts of white noise, it
was shown by Kuschnig et al. ([1997]) that frequency peaks with
suggest a 99.9% probability for a real period.
Individual errors are estimated from the width of the frequency peak
at
where *n* is the number of
data points and *m* the number of parameters (*n* - *m* is then the number
of the degrees of freedom; *m*=4 in our case). For details we refer to
Bevington ([1969]).

The periods designated
in Table 1 are the average
photometric periods computed from the individual bandpasses
and are always given up to the last significant digit. The light
curves in the figures are phased with these periods unless otherwise
noted. The corresponding zero point in time is *T*_{0}=0 for all light
curves. The phase shift of the light-curve maximum with respect to
*T*_{0} is indicated in the upper right corner of each phase-curve panel in
Figs. 1-45. A slight shift between light curves for the same
star but from different telescopes is due to slight differences in the
separately obtained periods.

Table 1 also contains a column for the stellar radius that is computed from the value of , the latter taken from the literature, and our photometric period.

Copyright The European Southern Observatory (ESO)