Broadband, digital, teleseismic P-wave data from the University of California, Berkeley, stations BKS, MHC, and SAO are used to study the character of teleseismic P coda. Velocity heterogeneity in structure near these stations gives rise to a variety of scattering effects, which include azimuthly dependent receiver functions, large tangential component particle motions, and high levels of P coda. P codas for shallow and deep teleseisms are analyzed for three frequency bands centered at 1.5, 0.75, and 0.375 Hz. Observed coda decay rate and level are fit with an energy flux theory to parameterize scattering effects near the source and at the receiver. Coda decay rate and level for deep earthquakes are seen to be stable observables between stations, implying that the scattered coda field is homogeneously distributed over the receiver network validating a major assumption in the energy flux model. Scattering Q (Qs) is not particularly well constrained by the data and is inferred to be affected by focusing and defocusing effects of the direct P waves. There is a pronounced difference in coda decay rate between shallow and deep earthquakes for 1.5 Hz coda, where coda decay rate for deep events is about a factor of 2 greater than for shallow events. There is no difference in coda decay rate between the two populations of events in the lower frequency bands. The observation at 1.5 Hz is consistent with a simple model of P-wave coda formation, where most coda forms in a scattering layer near the surface of the earth in regions near the source and near the receiver. The homogeneity of coda decay at 0.75 and 0.375 Hz for deep and shallow events suggests that a larger, near-receiver scattering mechanism, such as surface-wave formation, dominates the coda.