Delay alignment modulation (DAM) is a novel transmission technique for wireless systems with high spatial resolution by leveraging delay compensation and path-based beamforming, to mitigate the inter-symbol interference (ISI) without resorting to complex channel equalization or multi-carrier transmission. However, most existing studies on DAM consider a simplified scenario by assuming that the channel multi-path delays are integer multiples of the signal sampling interval. This paper investigates DAM for the more general and practical scenarios with fractional multi-path delays. We first analyze the impact of fractional multi-path delays on the existing DAM design, termed integer DAM (iDAM), which can only achieve delay compensations that are integer multiples of the sampling interval. It is revealed that the existence of fractional multi-path delays renders iDAM no longer possible to achieve perfect delay alignment. To address this issue, we propose a more generic DAM design called fractional DAM (fDAM), which achieves fractional delay pre-compensation via upsampling and fractional delay filtering. By leveraging the Farrow filter structure, the proposed approach can eliminate ISI without real-time computation of filter coefficients, as typically required in traditional channel equalization techniques. Simulation results demonstrate that the proposed fDAM outperforms the existing iDAM and orthogonal frequency division multiplexing (OFDM) in terms of symbol error rate (SER) and spectral efficiency, while maintaining a comparable peak-to-average power ratio (PAPR) as iDAM, which is considerably lower than OFDM.