Short-term properties of atrial fibrillation (AF) frequency, f-wave morphology, and irregularity parameters have been thoroughly studied, but not long-term properties. In the present work, f-wave morphology is characterized by principal component analysis, introducing a novel temporal parameter defined by the cumulative normalized variance of the three largest principal components (r3). Based on 7-day recordings from nine patients with stable chronic heart failure and persistent AF, long-term properties were studied in terms of r3 AF frequency, and sample entropy (SampEn). The main result of the present study is that detection of circadian rhythms depends on the parameter considered: rhythms were found in six (r3, SampEn) and five (AF frequency) patients, but not always in the same patient. Another important result is that circadian rhythms detected in 7-day recordings could not always be detected in 24-h periods, thus shedding new light on the results in previous studies which all were based on 24-h recordings. Infradian rhythms were found in four (r3, SampEn) and one (AF frequency) patients.