Abstract: Digital data of a specific plurality of data channels (K0-Kn) of the input side supplied in parallel form are converted into a serial data stream (D) and are in turn divided into parallel data channels (K0-Kn) of the output side upon reception with the assistance of a corresponding demultiplexing. In order to enable the allocation of the bits of the data channels (K0-Kn) of the input side read-in in parallel without great circuit-oriented outlay and without additional synchronization information, it is proposed to monitor the serial data stream (D) transmitted in the form of ATM cells for the occurrence of a specific bit sequence that is already transmitted with every cell format. The position of the individual bits of the corresponding data channels (K0 Kn) in the serial, optical data stream can be determined on the basis of this characteristic bit sequence, so that a correct parallelization at the data stream (D) at the output side is possible.

Abstract: The switching network serves for the transmission of digital data signals within a broadband communication system. This switching network is thereby constructed of individual switching elements (KE) that form at least one multi-stage funnel due to a connection to one another. The section for the assemblies of the switching network is selected such that at least one complete funnel is accommodated on the respective assembly. The respective assembly thereby has its external data inputs provided with unidirectionally designed optical/electrical transducers (OML-E) and has its external data outputs provided with unidirectionally designed electrical/optical transducers (OML-S).

Abstract: First packets P2i with a linear index 2i (i≧0) are transmitted in a first channel K1, second packets P2i+1 with a non-linear index 2i+1 are transmitted in a second channel K2 and third packets PXOR which are formed from two consecutive packets P2i, P2i+1, respectively, by bit-wise XOR are transmitted in a third channel K3. In this way, secure parallel switching matrices can be advantageously created with only three switching matrices.