sending speed from 10 WPM in the first 5-digit group down to 9.95 WPM during the 3rd and last group. This may not seem like much, but it is enough to cause loss of synchronization with the timing of dots and dashes near the end of the data.
Finally, more trial cross-correlations were made taking this drifting CW speed into account. Correlations for each trial 5-digit number were calculated with additional time offsets in steps of 40 ms, over a 1-second interval. The revised candidate answers were then listed in order of relative correlation magnitude. Fig 8 summarizes the results; the maximum correlation amplitude found for a given 5-digit number is plotted against the time at which that maximum correlation occurs. The best -100 of the 100,000 trials are shown. It is encouraging to see that the strongest correlation occurs within a few milliseconds of the predicted start of the data. The most probable answers are also clustered at start times separated by intervals of 0.24 seconds, which is
Was this article helpful?