Networks of random quantum scatterers (S-matrices) form paradigmatic models
for the propagation of coherent waves in random media. S-matrix network mo
dels cover universal localization-delocalization properties and have some a
dvantages over more traditional Hamiltonian models. In particular, a straig
htforward implementation of real space renormalization techniques is possib
le. Starting from a finite elementary cell of the S-matrix network, hierarc
hical network models can be constructed by recursion. The localization-delo
calization properties are contained in the flow of the forward scattering s
trength ('conductance') under increasing system size. With the aid of 'smal
l scale' numerics qualitative aspects of the localization-delocalization pr
operties of S-matrix network models can be worked out.