A near-field wideband communication system is studied, wherein a base station (BS) employs an extremely large-scale antenna array (ELAA) to serve multiple users situated within its near-field region. To facilitate the near-field beamfocusing and mitigate the wideband beam split, true-time delayer (TTD)-based hybrid beamforming architectures are employed at the BS. Apart from the fully-connected TTD-based architecture, a new sub-connected TTD-based architecture is proposed for enhancing energy efficiency. Three wideband beamfocusing optimization approaches are proposed to maximize spectral efficiency for both architectures. 1) Fully-digital approximation (FDA) approach: In this approach, the TTD-based hybrid beamformers are optimized to approximate the optimal fully-digital beamformers using block coordinate descent. 2) Penalty-based FDA approach: In this approach, the penalty method is leveraged in the FDA approach to guarantee the convergence to a stationary point of the spectral maximization problem. 3) Heuristic two-stage (HTS) approach: In this approach, the closed-form TTD-based analog beamformers are first designed based on the outcomes of near-field beam training and the piecewise-near-field approximation. Subsequently, the low-dimensional digital beamformer is optimized using knowledge of the low-dimensional equivalent channels, resulting in reduced computational complexity and channel estimation complexity. Our numerical results unveil that 1) the proposed approaches effectively eliminate the near-field beam split effect, and 2) compared to the fully-connected architecture, the proposed sub-connected architecture exhibits higher energy efficiency and imposes fewer hardware limitations on TTDs and system bandwidth.