To model the cold dust emission observed in the diffuse interstellar medium, in dense molecular clouds or in cold clumps that could eventually form new stars, it is mandatory to know the physical and spectroscopic properties of this dust and to understand its emission. This work is a continuation of previous studies aiming at providing astronomers with spectroscopic data of realistic cosmic dust analogues for the interpretation of observations. The aim of the present work is to extend the range of studied analogues to iron-rich silicate dust analogues. Ferromagnesium amorphous silicate dust analogues were produced by a sol-gel method with a mean composition close to Mg_1-x_Fe_x_SiO_3_ with x=0.1, 0.2, 0.3, 0.4. Part of each sample was annealed at 500{deg}C for two hours in a reducing atmosphere to modify the oxidation state of iron. We have measured the mass absorption coefficient (MAC) of these eight ferromagnesium amorphous silicate dust analogues in the spectral domain 30-1000um for grain temperature in the range 10-300K and at room temperature in the 5-40um range. The MAC of ferromagnesium samples behaves in the same way as the MAC of pure Mg-rich amorphous silicate samples. In the 30-300K range, the MAC increases with increasing grain temperature whereas in the range 10-30K, we do not see any change of the MAC. The MAC cannot be described by a single power law in {lambda}^-{beta}^. The MAC of the samples does not show any clear trend with the iron content. However the annealing process has, on average, an effect on the MAC that we explain by the evolution of the structure of the samples induced by the processing. The MAC of all the samples is much higher than the MAC calculated by dust models. The complex behavior of the MAC of amorphous silicates with wavelength and temperature is observed whatever the exact silicate composition (Mg vs. Fe amount). It is a universal characteristic of amorphous materials, and therefore of amorphous cosmic silicates, that should be taken into account in astronomical modeling. The enhanced MAC of the measured samples compared to the MAC calculated for cosmic dust model implies that dust masses are overestimated by the models.