This paper deals with the maximum-likelihood (ML) noncoherent data-aided (e.g., no blind) synchronization of multiple-antenna ultrawideband impulse-radio (UWB-IR) terminals that operate over broadband channels and are affected by multipath fading with a priori unknown number of paths and path-gain statistics. The synchronizer that we developed achieves the ML data-aided joint estimate of the number of paths and their arrival times (e.g., time delays), without requiring any a priori knowledge and/or a posteriori estimate of the amplitude (e.g., module and sign) of the channel gains. The ultimate performance of the proposed synchronizer is evaluated (in closed form) by developing the corresponding Cramer-Rao bound (CRB), and the analytical conditions for achieving this bound are provided. The performance gain for the synchronization accuracy of multipath-affected UWB-IR signals arising from the exploitation of the multiple-antenna paradigm is (analytically) evaluated. Furthermore, a low-cost sequential implementation of the proposed synchronizer is detailed. It requires an all-analog front-end circuitry composed of a bank of sliding-window correlators, whose number is fully independent from the number of paths comprising the underlying multiple-antenna channel. Finally, the actual performance of the proposed synchronizer is numerically tested under both the signal acquisition and tracking operating conditions.