A recent digital system identification technique, called discrete-time vector fitting (VFz), is extended to two communication system filter design problems. The first one is the complex infinite-impulse-response (IIR) filter design. The second one is the common-denominator MIMO filter design. In the second problem, a combination of VFz and QR decomposition is proposed to reduce the computational complexity. Robustness in finite-precision arithmetic is imposed in both problems via a pole radius constraint. Numerical examples confirm that VFz exhibits fast convergence and produces highly accurate approximants.