Frequency or phase modulated electrical signals arise in multiple applications in communication systems as well as in measurement setups such as interferometry / vibrometry, velocimetry etc. Common demodulation techniques are often based on analog techniques (e.g. a PLL in the RF band) or - after mixing into the base band and generating IQ signals - on digital signal processing algorithms. Two DSP based algorithms for the frequency demodulation of lQ baseband signals will be presented and compared in this paper. One of them is the synthetic heterodyne approach, based on the digital derivation of the arctan of the angle described by the I and Q component. It will be shown, that this algorithm is very sensitive to noise in the baseband signals. The other, more modern algorithm uses a Kalman filter as an "optimal" estimator of the instantaneous phase and frequency of the two signals. It will be shown that the Kalman filter shows a drastically improved noise performance.