Aeroacoustic sensing is well motivated due to its passive nature and low bandwidth, and processing with array baselines of one to a few meters is well studied and useful. However, arrays of this size present difficulties in manufacture and deployment. Motivated by the desire to develop smaller, cheaper (perhaps "disposable") sensor packages, we consider performance for arrays with small baselines. The performance of angle estimation operating roughly in the [30, 500] Hz regime is limited by the observed signal-to-noise ratio (SNR) and propagation in the turbulent atmosphere. Scattering results in a reduction of spatial coherence, which places fundamental limits on angle estimation accuracy. Physics-based statistical models for the scattering have enabled prediction of network performance including detection, angle estimation, time-delay estimation, and geolocation. In the present work, we focus on angle estimation accuracy for the short baseline case. While a large aperture is desirable to improve angle estimation, the propagation produces a rolloff in the spatial coherence that ultimately degrades the performance despite increasing the array aperture. We characterize this tradeoff analytically via Cramer-Rao bounds, as a function of SNR, frequency range, sensor array geometry, and propagation conditions. The results clearly favor shorter ranges and higher frequencies when employing small array baselines.