Deep level transient spectroscopy (DLTS) measurements on multilayer devices can be influenced by secondary barriers. Our starting point is a simple model used in the literature for simulations of DLTS signals induced by such barriers: the structure is represented by two diodes connected back-to-back and separated by a neutral region. Although the calculations confirm that capacitance transients can occur in multilayer structures without defects, we find that the quantitative modeling of corresponding DLTS peaks resembling the N1 signal in Cu(In,Ga)Se2 solar cells requires unphysical input parameters. Therefore, we carefully analyze the assumptions of the model and investigate the limits of its applicability. We look into details of the formation of a DLTS signal and unveil the connection between observed kinetics, capacitance-voltage profiles, and the shape of DLTS peaks. We show that the signals originating from secondary barriers exhibit a logarithmic dependence on the pulse duration which can mimic a response from extended defects. We also verify the validity of the criterion existing in the literature allowing to distinguish signals coming from a back contact and deep defects and show that it can be used only in very specific situations.