...bands

In the error calculation, calibration errors were not taken into account, since the main goal was to analyse flux variations, rather than to obtain a precise evaluation of the magnitude.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... above

A $10\rm ~d$ smoothing time scale appeared as the most suitable choice: a shorter one would have been too sampling dependent; a longer one would have smoothed too much the evident long-term variations which are visible in the best-sampled periods. In any case, changes of the binning interval in the range 5-$30\rm ~d$ do not affect the final results sensitively.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.