[aprssig] "Best" packet decoder solution

[Robert Bruninga]

> I think if we're going to see any FEC scheme 
> succeed for APRS, it needs to be either 
> simple to implement or ...

There is HUGE code redundancy already built into APRS that could
be exploited.  Looking at the following standard APRS oosit"

!DDMM.MMN/DDDMM.MMW$...

Most all of those digits are decimal digits,(only 4 bits) yet
they are encoded in 8 bit ascii.  So a bit error in any digit
only has a 50% chance of messing up the 4 bits needed.  

Further, For the LOWER order bytes, every posit cannot be any
further away from the previuos one by the SPEED times TIME.
Further, even if the error is in a given BYTE, you can at least
convert it to an ambiguous position within the degree of
ambiguity of the messed up byte.

Further, for the upper order bytes, you can infer the position
on RF because your digipeater has to be within a degree of LAT
or LONG to even hear it.

So, simply run your TNC with PASSALL ON, and you will then see
packets even with errors.  Then use the above huristics to
decide on the best estimated posit.  The give it your best shot
and give it the QUESTIONMARK symbol so that you can see that it
is estimated until a solid good packet comes through.

I would think this can greatly improve the receipt of weak
signals.
Of course, it has been shown that AX.25 errors are rarely single
bit errors but runs of bits and so it is not quite so
simplistic, but it sure is a first step and can be done on ALL
EXISTING PACKETS on the air...

Bob, WB4APR