Massive Multiple-Input Multiple-Output (MIMO) systems with hundreds or thousands of antenna elements are fundamental to next-generation wireless networks, promising unprecedented spectral efficiency through spatial multiplexing and beamforming. However, the computational burden of channel state information (CSI) acquisition and processing scales dramatically with array size, creating a critical bottleneck for practical deployments. While previous works demonstrated that Fast Fourier Transform (FFT)-based beamspace processing can exploit the inherent angular sparsity of wireless channels to compress CSI feedback, the digital implementation requires intensive computations that become prohibitive for ultra-large arrays. This paper presents an analog alternative using Butler matrices-passive beamforming networks that realize the Discrete Fourier Transform in hardware-combined with RF switching circuits to select only dominant angular components. We provide a comprehensive analysis of Butler matrix architectures for arrays up to 32 x 32 elements, characterizing insertion losses across different technologies (microstrip, substrate-integrated waveguide, and waveguide) and operating frequencies (10-30 GHz). The proposed system incorporates parallel power sensing with Winner-Take-All circuits for sub-microsecond beam selection, drastically reducing the number of active RF chains. Full-wave simulations and capacity evaluations at 12 and 30 GHz demonstrate that the Butler-based approach achieves comparable performance to FFT methods while offering significant advantages in power consumption and processing latency. For a 256 x 256 array, FFT computation requires 0.36 ms compared to near-instantaneous analog processing, making Butler matrices particularly attractive for real-time massive MIMO systems. These findings establish Butler matrix front-ends as a practical pathway toward scalable, energy-efficient beamspace processing in 6G networks.