-----Original Message-----
From: [email protected] [mailto:[email protected]]
Sent: 04 January 2002 04:36
To: [email protected]
Subject: Re: LF: Re: m-FSK: SNR vs bandwidth
demanding combination, after all. Bit rates slower than about 1/second
never
really caught on here. And where could we get a clear definition of
Differential BPSK, please?
In normal BPSK, the data is coded so that a phase of 0 degrees represents a
'0' and 180 degrees a '1'. Since we have no idea of the absolute starting
phase, it is obvious that an ambiguity can arise as it is impossible to
determine which phase state is which, and the sequence is then just as
easily decoded with the 1s and 0s transposed.
There are several ways to solve this problem. One is to transmit a known
sequence at the start of, and/or periodically during, the transmission. If
this known sequence is decoded upside down then we know that all subsequent
data has to be inverted. Another allied method would be to define symbols
at specific times, such as the UTC second, as being in a known state. This
latter technique obviously relys on having accurate time information at both
ends of the link.
Alternatively, the data can be encoded in such a way that polarity is
irrelevent. Instead of encoding a '0' or '1' as the absolute phase value,
encode the signal such that no change of phase from one symbol period to the
next represents a '0' and a change of phase represents a '1'. Now, a long
string of '0's would appear as a continuous carrier and a long string of
'1's as a carrier whose phase swapped 180 degrees every period. This is
Differential Binary Phase Shift Keying (DBPSK).
DBPSK is used in most HF and VHF modulations as it is the simplest way to
resolve the phase ambiguity problem. PSK31 uses DBPSK, so does Coherent,
and in each of these modes the idle sequence, when no data is being sent,
consists of a long chain of '1's so the transmitted sequence repeatedly
inverts every symbol, giving a demodulator the maximum likelyhood of locking
correctly to the signal timing.
It does have the following disadvantage however: If a symbol is corrupted
by noise or interference, then not only is that data bit decoded
incorrectly, but so is the next symbol, as its phase shift will be
interpreted incorrectly wrt. the corrupted one. Therefore the Bit Error
Rate (BER) for DPSK is double that of absolute BPSK. It is not, as is often
taken to be the case, equivalent to a 3dB degradation in S/N.
Depending on the point on the S/N ratio versus BER curve at which the
demodulator is sitting, a doubling of the BER can correspond to either a
minute change of a fraction of a dB in S/N if S/N is poor, or a hugh change
of many dB if S/N is large
For our LF, very low bandwidth signalling needs however, we do not have to
go to differential modes and as our S/N will always be 'poor' by definition,
the gain of not having to employ differential coding can be considerable.
With symbols of over 1 second in length, it is quite straightforward to,
just as an example, define the bit sent on the UTC minute or hour as being a
zero. Relying on absolute time also has the advantage of not having to rely
on lock up sequences and procedures - always the weak point in coherent
signalling systems. There is no problem these days in getting time to a
few 10s of milliseconds accuracy, and a low cost GPS module will give better
than 1us accuracy. With such easily available, highly accurate timing
information other measurements of the LF path become possible, such as
flight time from one station to the other which would allow some of the
anomalies of the long paths to be investigated.
Andy G4JNT
--
The Information contained in this E-Mail and any subsequent correspondence
is private and is intended solely for the intended recipient(s).
For those other than the recipient any disclosure, copying, distribution,
or any action taken or omitted to be taken in reliance on such information is
prohibited and may be unlawful.
|
