[aprssig] More efficient use of channel capicity throughshorterpackets

scott at opentrac.org scott at opentrac.org
Thu Oct 12 11:24:54 CDT 2006

> baud FSK, the detection of the tones is fairly quick, but 
> does the modem 
> actually deliver good data immediately?  Or does it need to 
> see multiple 
> bits fly by before its local bitclock is sync'd, and it can reliably 

Depends on the clock recovery used in the receiver.  Normally that's a DPLL
that has to be synced to the incoming bit rate, and loop bandwidth affects
how fast it can be synced.  Sending a bunch of zeros (alternating high/low
in NRZI) gives the greatest number of transitions and makes it easier to
lock on to the signal.  I think most receivers should be able to lock on in
0-10 bit times.  My Tracker2 prototypes do pretty well in that respect -
I've implemented a kind of fuzzy clock recovery scheme that at least
outperforms the KPC-3.  Hook up the output of a KPC-3's demodulator to a T2
and (at least with my test corpus) it'll decode 1-3% more packets than the
KPC-3 itself.  Of course, generally speaking you've got to set your TXD high
enough to work with the slowest TNC around.

>    With 9600baud modems which use more complicated encoding 
> schemes, I 
> suspect this modem clock recovery/bitclock sync time may be longer as 
> well.

I think it depends a lot on implementation, and the specific scheme used.  I
haven't worked with higher baud rates much myself, but I'd expect that an
old hardware-based linear feedback shift register is going to take a lot
longer to sync up to a scrambled signal than a properly written software
equivalent would.

> Another short-packet data application, the ubiquitous credit card 
> authorization terminal, is still using 300 baud modems to 
> this day, simply 
> because the handshake/lockup time on the faster codecs adds 
> more time than 
> is saved during the data-transfer phase.

Excellent example.  You can force faster modems into shorter negotiation
sequences if they're configured right, though I haven't tried that since my
days of exchanging BBS message bundles at 14.4k and trying to shave as much
as possible from the long distance bill.

For something like credit card authorization, you only have to exchange tens
of bytes.  A 300 baud modem can get the job done before a 28.8 is halfway
through its negotiation.


More information about the aprssig mailing list