A phenomenological model for simulating photoplethys-mogram (PPG) during paroxysmal atrial fibrillation (AF) is proposed. A PPG pulse is modeled by combining a lognormal and two Gaussian waveforms. Continuous PPG signals are produced by placing and connecting individual pulses according to the RR interval pattern extracted from annotated ECG signals. This paper presents a practical application of the proposed model for studying the performance of an RR-based AF detector. Physionet databases containing AF episodes serve as a basis for modeling PPG signals. Detection performance was tested for different signal-to-noise ratios (SNRs), ranging from 0 to 30 dB. The results show that an SNR of at least 15 dB is required to ensure adequate performance. Considering the lack of annotated, public PPG databases with arrhythmias, the modeling of realistic PPGs based on annotated ECG signals should facilitate the development and testing of PPG-based detectors.