The paper analyses the effect of spatial smoothing on the performance
of MUSIC algorithm. In particular, an attempt is made to bring out two
effects of the smoothing: (i) reduction of effective correlation betw
een the impinging signals and (ii) reduction of the noise perturbation
s due to finite data. For the case of a two-source scenario with widel
y spaced sources, simplified expressions for improvement with smoothin
g have been obtained which provide more insight into the impact of smo
othing. Specifically, a pessimistic estimate of the minimum value of s
ource correlation beyond which the smoothing is beneficial is brought
out by these expressions. Computer simulations are used to demonstrate
the usefulness of the analytical results.