Well done Markus and Paul. Most impressive!
73 & Congrats
On Sun, 18 May 2014, Paul Nicholson wrote:
Another successful test by Markus DF6NM this morning,
2014-05-18. Sending 100 bits at 8270.000 Hz using a rate 1/8
convolutional code with constraint length 25.
The data rate was 24.2 bits per hour and the Shannon capacity
for the channel is around 28 bits per hour.
The 100 bit message was encoded into 992 BPSK symbols sent
coherently at 15 seconds per symbol. The decoder corrected 343
symbol errors - a symbol error rate of 34.6% which corresponds
to Eb/N0 of around -2.0dB.
A one hour carrier immediately following the message enabled
a signal strength measurement: 0.09 fT and S/N of 13.2 dB in
278 uHz. The calculated Eb/N0 for that is -1.57 dB which happens
to be just about exactly at the Shannon limit.
The decoder running on a 32 core AWS EC2 c3.8xlarge took about
3 minutes to find a reference phase and decode the signal.
The message ranked at the 28th maximum likelihood path
through the Viterbi trellis.
This is possibly the strongest forward error correction code
ever transmitted by a radio amateur. In combination with
coherent signalling and Viterbi list decoding, a circa 10uW
ERP signal is decoded with 95% reliability at a range of over
1000km and achieves 85% of the Shannon capacity.