Andy, G4JNT wrote:
<There is another modulation scheme suitable for this band that
<does allow constant amplitude, and hence highly efficient class D
<amplifiers. That is Continuous Phase Modulation (CPM), of which
<Minimum Shift Keying is an extreme example. Listen to the
<transmission from Rugby on 16kHz or 22.1 kHz for an excellent
<example of a 100 Baud version of MSK.
<CPM is a very processor and memory intensive mode hence it is only
<now becoming popular, but at the speeds used at LF my feeling is
<that a PC could cope even if a 56002 EVM does not have the necessary
<memory capacity. CPM is theoretically a few dB better than PSK in
<S/N terms, and in fact with the right coding is said to be able to
<get to within 0.7db of the Shannon limit. Take this together with
<the constant envelope waveform and it could prove very exciting for
<very weak signal LF dxing.
<If anyone reading this has a more intimate knowledge of CPM please
<get in touch. As far as I 'm concerned its all squiggles in a text
<book at the moment, and as for turning these into DSP code....
CPM ist the general class of continuous phase modulation.
All modulation schemes resulting in constant amplitude
and continuous phase belong to this class. Especially
in conjunction with the Partial Response technique high
data rates at low bandwidth can be achieved.
Although the 136 kHz band seems to be very narrow,
amateur LF-transmissions are limited in power, not in
bandwidth, i.e. the extremely low power forces the
information bit rate into the region of 1 bit per second.
A bandwidth of less than 1 Hz would be possible. But,
on the other hand, Shannons Formula says that LARGER
bandwidth is better if you are limited in power.
Indeed, as Andy, G4JNT, mentioned, the bitrates are so
low, that a PC could do the job of decoding.
I therefore propose to use PSK (and possibly FSK for
transatlantic distances because control of the phase
may be lost), at a baudrate several times the
information bitrate. A good example is the interleaved
deep space code used by Pactor 2. Since the bit rate on
LF is much lower, the same, i.e. soft-Viterbi-decoding
of whole packets, could be done at lower code rate
and longer constraint length. The best code for that
purpose known to me is the (56,14,36)-convolutional
code. I am searching for even longer constraint lengths
than 14. (Pactor 2 uses contraint length 9 at rate 1/2).
As a textbook I use K.D.Kammeyer: Nachrichtenuebertragung
(Communication Theory), Teubner,1992. CPM and Partial
Response is well written - may be because the book is
some 8 years old.
At the moment, I am very busy in building up a practical
course on communication technique for students of computer
science at the University of Hamburg based on simulations
in MATLAB including coding and Viterbi-decoding.
Unfortunately, there is no time left for amateur radio
except from writing an e-mail.
72 de Klaus, DJ5HG