Three different types of frequency response estimators are considered and compared with each other with respect to the inherent resolution bias errors. The behavior of these estimators is shown to differ considerably, especially with respect to peak estimation in resonance systems such as a simple spring-mass system. One of these estimators is shown to exhibit almost no peak bias error at resonance under noise-free measurement conditions whereas the peak bias error of the estimator most commonly used is maximum. However, aside from resonance, the peak-bias-free estimator proves to be the worst of the three estimators. The theoretical predictions are accurately verified by experiments with an electrical LRC-resonance circuit.