I wrote:
> all the symbol error rates are higher than expected
> for the measured carrier S/N, a loss of 2 or 3dB.
I can reproduce this using EbSynth and a weak near-field
9th harmonic signal. 2 to 3dB loss every time on a similar
strength signal to last night's test.
There is also an error floor showing up, about 6 to 8% BER
with a very strong signal that should have negligible chance
of even a single bit error.
Something is wrong with EbSynth.
The symbol errors are bunched,
http://abelian.org/tmp/160215b.gif
and if I repeat the test, the bunches are differently placed.
It looks like the phase of the signal is deviating sufficiently
to cause bit errors on a timescale of several seconds (the
duration of an error bunch). It needs a 10 degree phase error
to produce a bit error on a strong signal (using 9th harmonic).
EbSynth then is not working as well as I'd thought with the
9th harmonic. This will badly affect the decode probability
on signals near the limit.
I expect it can be fixed, perhaps I'd better not take apart the
MF tx just yet so we can do another test.
Meanwhile - good luck Jim. If the EbNaut doesn't decode, the
intervening spells of carrier are a good diagnostic. Hope that
VLF ticket comes through soon - before the season is over.
--
Paul Nicholson
--
|
