Table 2 shows how the first-break and first-peak times vary with transducer size for the P-wave case shown in the right part of Figure . The time picked in the ideal infinite-transducer case is used to define the ``correct'' arrival time. First of all, note by comparing the ``-12 (peak)'' case in Table 1 with the `` (phase)'' case in Table 2 that both picking methods are sensitive to the phase shifts in the recorded wavelet due to the effective dimensionality of the source. The 12mm transducer width is evidently not quite wide enough to fully mimic the wavelet launched by an infinite planar source, even recorded at the center of the leading flat spot. The first-break method is five times less sensitive to the difference in the wavelet phase than the first-peak method, however.
Similarly in Table 2 we find the first-peak method is more sensitive to variations in the recorded wavelet due to the flat spot partially missing the receiver. This should not be surprising. The first-peak method must wait until a substantial part of the wavefront's leading energy surge encounters the receiver before it can detect the arrival; if the leading focus of energy mostly misses the receiver there will be a delay while the detection method waits for more energy to arrive from the trailing portion of the wavefront to the side of the leading flat spot. The first-break method, on the other hand, ideally can detect an arrival when any of the leading ``flat part'' of the wavelet hits the receiver; anything short of a complete miss should result in the phase-velocity answer. (Of course in practice we have to pick the first break at some finite amplitude level, so there will always be some slight delay.) While it is true that first-break times should be more robust against anisotropic effects, Table 2 shows that using the first-peak time for the 12mm-wide transducers in our P-wave example would still have resulted in a phase-velocity measurement error of only , not that bad.