Stefan wrote:

> I will repeat the test in a few days on 6470.1 Hz.

Perhaps you would like to try the next software
revision of EbNaut.

Decoder version 0.7 and encoder version 0.5 at

 http://abelian.org/ebnaut/software_ms.shtml

There is some testing that I haven't finished but it is
fit to use for normal message lengths.

This version has CRC down to 0 bits, and will do up to 1000
chars (although you need a terrible amount of RAM to decode
such long messages).

The actual CRC size to use is not critical, just keep to less
than about 8 bits for 3 chars and less than about 6 bits for
2 chars.   Increasing the CRC reduces the gain of the inner
(convolutional) stage but increases the gain of the outer
(list decoder) stage.  The overall performance is about the
same but the shorter CRC means a shorter list and slightly
quicker decode.  Also the shorter CRC leads to slightly
longer symbols for a given Eb/N0, so less critical symbol
timing.

--
Paul Nicholson
--