[Lowfer] PSK31 success last night.

[JD]

XIQ wrote: "For me, I am far more interested in the QSO and the content of 
that QSO rather worrying about some astonishingly low detection limit."

I thoroughly endorse that view myself!  Software techniques that are mainly 
about automated reporting of spots, or which can only establish correlation 
with an ID already in a database, are OK as a technical exercise, I suppose, 
but they leave me cold because (IMO) radio should be about communication.

Similarly, though, my question pertained to robustness and performance of 
the raw modulation method itself, rather than interface convenience or 
available features.  These things are also important, and I would not claim 
otherwise.  But they are a separate matter from the modulation technique, 
the coding scheme, and any error correction method employed, that make any 
given mode work.

It's way too easy to end up comparing apples to oranges if one talks about 
GMSK being a "mode" and MSK being another "mode," without specifying actual 
parameters or even which implementation one is using.

The technique identified as "GMSK" in MMVari, for instance, appears to be a 
simple varicode data stream at 15.625 or 31.25 baud with no processing other 
than the Gaussian filtering, before being frequency modulated onto the 
carrier with m=0.5.  (At least, I haven't yet found any mention of there 
being FEC or the like.)  However, there is no reason GMSK data has to be a 
custom varicode as it is in the MMVari engine (it could be RTTY or specially 
timed Morse or most any other coding method).

Likewise, there is no reason why plain MSK necessarily has to be represented 
by the ZL2AFP "CMSK" implementation.  CMSK has convolutional coding and who 
knows what all built in before the data stream is allowed near the carrier. 
Ergo, we can expect it to be touchier about tuning as you noted, and for it 
to fall apart more completely as one drops below detection threshold 
(although perhaps less error prone as its developers claim, when everything 
is nearly ideal), and for the data throughput to be less for a given baud 
rate because of the extra payload.

(Before I continue, I wonder what your transmission speed was in comparing 
these two, John, and what your bandwidth limitation is?  I know a little 
something about the modulation methods themselves, but less about the 
software that's available for amateur use.  The different software packages 
seem to have as much folklore about them on the Web as actual fact.)

Anyway...all else being _strictly_ equal, in principle plain MSK has certain 
robustness advantages over GMSK.  That's why I asked if PSK31 isn't giving 
Neil the kind of coverage he wanted, why he thought GMSK would do better? 
>From the modulation standpoint alone, if the baud rate is the same and the 
coding were also the same, GMSK should perform slightly worse and/or require 
more transmitted power, and plain MSK should perform slightly better.

I suppose the practical difference is that different software packages do 
NOT necessarily have the same coding schemes, baud rates, and data 
throughput.  That's why it's good to hear real-world results such as yours, 
John.  Yet I would reiterate that "better" is relative in comparison of 
modulation methods when different formats and speeds of data are being sent. 
To be meaningful, it should probably also be specified which software is 
being used at what speed, and whether/what form of error correction is being 
employed.

John D.