This paper is concerned with the performance of differential amplify-and-forward (D-AF) relaying for multi-node wireless communications over time-varying Rayleigh fading channels. A first-order auto-regressive model is utilized to characterize the time-varying nature of the channels. Based on the secondorder statistical properties of the wireless channels, a new set of combining weights is proposed for signal detection at the destination. Expression of pair-wise error probability (PEP) is provided and used to obtain the approximated total average bit error probability (BER). It is shown that the performance of the system is related to the auto-correlation of the direct and cascaded channels and an irreducible error floor exists at high signal-to-noise ratio (SNR). The new weights lead to a better performance when compared to the conventional combining scheme. Computer simulation is carried out in different scenarios to support the analysis.