Re: Radios for higher speeds,

[Phil Karn <karn@unix.ka9q.ampr.org>]

>It is worth mentioning that telephone modems implement error detecting/
>correcting protocols that really depend on the full-duplex nature of
>the telephone line to achieve reasonable performance.  Also, the

I don't think this is really true. Sure, there's an ARQ protocol
running between the modems to detect and recover from errors. But it's
just another uncreative variant of LAPB, which is also the basis of
AX.25.

If anything, your typical V.42/LAPM implementation is even worse than
your typical AX.25 implementation. Long fixed frame retransmission
timeouts (e.g., 9-10 sec) are the norm. And the mechanism that
negotiates its use at the start of a call is prone to fail on noisy
lines. I.e., when you need it the most, it goes away entirely.

Things work as well as they do only because phone lines are so much
cleaner than your average packet radio channel.

The transmit and receive halves of an actual V.32/V.32bis/V.34 modem,
including the trellis coding, run more or less independently. This is
a neat trick considering that they operate over a 2-wire line using
the full bandwidth in both directions simultaneously.


>really snazzy telephone encoding formats are very sensitive to
>changes in the phase response of the link; on a telco line, this
>don't change that often.  On a radio link, this can change several
>times a second.

Actually, changing phase response is more of a problem on telco lines
than on radio links. Bit slips in a digital carrier are death to most
modern modems. And since they're inaudible to voice, it can be very
hard to convince a telco to fix them.

Radio transceivers have the advantage of being fairly stable in their
phase characteristics. The big problem, of course, is noise and
interference. Both are usually much greater than on most telephone
channels.

Phil