One of the most characteristic properties of complex systems, including the stock market, is that their temporal evolution comprises both noisy and collective components. Despite the overwhelming noise, dominating especially on short, minutely time scales, identification of the collective components is relatively easy by using the correlation matrix analysis. In this case each correlation matrix eigenvalue can be associated with an independent component of the signal (so-called eigensignal). Those components corresponding to the collective eigenvalues can reveal different properties if compared with the random ones. In the present study we analyze statistical characteristics of the eigensignals for the correlation matrices calculated for stocks of the largest companies listed either on the German or the American stock market. In particular, we examine the differences in properties of the eigensignals corresponding to largest and smaller eigenvalues. We also analize fractal features of the eigensignals.