Detection performance of multibranch and multichannel compressive receivers

Some proposed architectures for compressive sampling of time-domain signals involve splitting the signal into multiple branches. This approach is not unlike the idea of splitting a signal into in-phase and quadrature branches for a coherent receiver implementation, but in the case of a compressive receiver the idea is to modulate the signal in each branch against a different measurement kernel. Therefore, multiple projections can be captured in parallel over the same time interval of the signal. In this paper, we evaluate the detection performance of multibranch compressive receiver implementations. The noise statistics over receiver branches are derived and shown to depend on the quality of the low-noise amplifier (LNA). A high-gain, low-noise-figure amplifier causes correlated noise across the receiver branches, but avoids SNR loss due to the power divider. We discuss whether a multibranch approach can provide equal or better detection performance than a single-branch compressive receiver, and include comparisons for varying LNA quality. A comparison to a true multichannel implementation is also provided.