The modification attempted to stabilize the signal delays calculated by cross-correlation of the waveforms received from the transducers. We observe considerable sample-to-sample variability in the waveforms. This may be due to interference between sound waves traveling directly from transmitter to receiver and waves reflected from the air-water interface.
Whatever the cause of the variability, averaging of the traces on a digital scope seemed to smooth it out. So we modified the on-line program to carry out a running average of the two waveforms.
The result produced very stable tracking for the phase time. The envelope time has periods of stability, and other times when it is quite unstable.
We chose two two-hour runs and cleaned up the data, removing jumps of multiples of 5 microseconds (the carrier period) in either phase time. Four jumps were removed in the Channel A phase time, and about 15 for Channel B. The resulting smooth curves were shifted to best agree with the envelope times during periods of stability.
This hand-massaging of the data is quite time consuming. However, it lends itself to automation. And, we think that the quality of the data can still be improved considerably, through a refined averaging procedure.