>There's a lot of discussion going on at the moment on the US "LowFER"
>mailing list about the relative merits of slow CW versus BPSK
Indeed there has been. Some of it has gotten rather silly and occasionally a
bit abusive, in fact. I'm glad to see the discussion has been more civilized
here, although there are still evident tendencies to compare apples with
oranges and to draw too many conclusions from accepted premises.
Some of the rather extreme arguments in favor of BPSK and similar modes tend
to count the same chickens twice; perhaps more often, for those claiming 23db
advantages <grin>.
The most credible proponents say that, for a given bit rate, coherent methods
have a 9db advantage. Andy Talbot's post yesterday argues this case well.
Six decibels are the result of the effective doubling of detected signal
output by one state of the received signal adding in-phase and the other
adding out of phase in the detector. Integration of voltage over the symbol
period accounts for the other three. The latter effect is a direct
consequence of coherence (one might call it the definition of coherent
detection), and the former requires coherence to exist before it can be
true...or for any detection to exist at all. With coherence, a BPSK-based
mode has a 9db advantage over on-off keying detected non-coherently, or 6db
over coherent CW.
Does this mean, therefore, that BPSK is inherently "better" than Slow CW?
Andy was careful to preface his comments with the qualification, "To compare
theoretically, firstly we need to make the assumption that data rates are
similar for the two modes." However, there is an additional qualifier that
follows from all this: we must also assume that coherence _can_ in fact be
achieved and maintained over the duration of enough symbols for the message
to be reconstructed.
If this latter condition cannot be met, then PSK-based modes have no
advantage. They simply don't work at all.
The traditional technique of reducing data rates and narrowing communication
bandwidth only work for coherent methods to the extent that the path between
the transmitter and receiver is stable enough to accomodate the slower bit
rate. It doesn't matter if the sender and receiver are using perfectly
synchronized atomic clock frequency standards: if propagation effects distort
the all-crucial phase information enough during each individual symbol, the
symbol will not be recovered.
Hence, for a given propagation mode--and, where the ionosphere is involved,
this is also a function of carrier frequency, path length, launch angle,
geomagnetic conditions, and so on--there will be some _minimum_ data rate
below which a given coherent method will not work. Weak-signal reception
techniques based on power integration over extremely narrow bandwidths (i.e.,
spectral analysis techniques) do not suffer nearly as strongly from the same
effects.
Let's re-evaluate the numbers in this light. Suppose, over a given long
path, we find that BPSK will let us receive an acceptable percentage of a
transmission at 10 bits per second, but faster rates are limited by noise and
slower rates are limited by phase errors in the propagating medium. To
achieve the same result, we would have to slow down our on-off keying by 9db,
to roughly one dot per second.
Now, however, let's suppose that either the noise levels increase, or we
attempt to receive the same signal over a greater distance. We can't slow
the speed of the BPSK transmission any further without losing the ability to
recover it coherently; which is to say, at all. Yet, we can slow down the CW
transmission and tighten the resolution of our receiving software and
continue the exchange of information.
(This can't be done without limit, of course. At some point, propagation
phase shifts will spread the spectrum of the on-off carrier enough to
disperse it across multiple channels, considering how narrow the detected
channels have become at that point, and the analog signal-to-noise ratio will
then rapidly collapse too.)
I don't follow the BPSK e-mail group, so this may not be the most current
information, but as far as I am aware, the best LF DX results with BPSK over
here have been at 100 and 200 milliseconds per bit. Some have tried 500 and
1000 milliseconds, but have not done as well; whereas on HF, rates of 50
milliseconds per bit or faster seem most effective.
If 200 milliseconds per bit turns out to be the lowest practical speed for
longwave DX, then we already know that Slow CW at 3 seconds per bit should be
able to match its data recovery performance if the time penalty is
acceptable. We also know that 10 seconds per dot, or longer, is entirely
feasible with the available receiving software. The only thing we don't know
is how Slow CW performs under conditions of very long paths at LF, on the
order of 1600km or more at 5mw radiated power, for which we do have some
experience with BPSK.
The essential point is that there are conditions below which coherent methods
cannot work at all, but in which other (slow) weak-signal methods may provide
at least a chance of getting signal through. There's room for a lot of
experimentation with these "different horses!"
73,
John KD4IDY
|
