Another alternative is replacing the discrimation of dashes and dots from
the time-domain to the frequency domain. This would mean that 'dashes' and
'dots' have the same length (3 seconds) but that they are transmitted on
different frequencies. Since in visual-CW the frequency discrimination is
very high a shift of 5Hz (or even less) is sufficient. Using this
we do not need the 'time-gaps' between the dashes and dots, only between
characters we would need a 'time-gap' of 1 dot and a 2 dot 'time-gap'
It seems that some similar thinking is going on on both sides of the world
on this topic. We have been using for some time a combination of
Spectrogram and small frquency shifts as a positive means of identifying
weak trans-Tasman LF signals. Also you will have noted Steve Olney's
extensive work in this area.
I am a little confused about the tests you describe Rik. If you eliminate
the time gaps between a string of dots then you just get a long dash so I
assume you are shifting frequency after each "element" rather than one
frequency for dots and another for dashes?