A modest proposal

[Hal Feinstein]

Concerning PSK31.  One characteristic is its "glass teletype" mode
where, when the channel S/N drops the PSK31 garbles the communications.
 From a computer keyboard you cannot tell if the sender is sending smile faces,
is a bad speller or the channel is going bad.  It is most disturbing with
are short periodic fading

I looked over Peter M. paper and the design goal of PSK31, but I wonder if
a better protocol could not be designed.  A lot of protocol design work is one
the books and could be mixed with the PSK31 modulation scheme to
create a protocol supporting near error free transmission.

One solution is the ARQ type protocols but these cut down on bandwidth because
they acknowledge every few characters.  By 1990 so called adaptive 
protocols were
designed which provided a variety of choices among forward error coding, 
retransmission
and positive acknowledgement.  A second mechanism
embedded a heavily encoded command channel within the information flow
to tell the receiver to switch to a more robust combination of mechanisms 
when the channel
degraded or to a lesser robust combination with channel improvements.

AN EXAMPLE ADAPTIVE PROTOCOL
An example of a modern adaptive protocol is the protocol underlying 
automatic link establishment (ALE).
Its design goal is to automate the operation of a radio in the event a 
skilled radio operator
is unavailable and includes facilities to automatically  announce itself 
(called sounding), listen for other ALE station
and build networks of them including who they can hear,  pass short text 
message exchange and
automatically move between frequencies in a channel list.
A variety of ALE automation functions, networking and message passing 
require communications
with other ALE equipped stations.  This communications is carried out using 
an adaptive protocol
that includes not only robust error coding, selective retransmission and 
positive acknowledgment,
but also two sided channel quality assessment to measure and select the 
degree of required
error protection.  The standard commercial high frequency transceiver 
outfitted with ALE utilizes
a specifically designed multi-tone signalling alphabet encoded with half 
rate golay encoding.
ALE protocol messages passed between ALE equipped systems utilizes this 
multi--tone
signaling alphabet.  Once a link is established between two ALE equipped 
stations
the transceivers typically drop back to a conventional modulation scheme 
which may
include SSB or a military data transmission format.  ALE also  possess an 
ability to pass
short  "order wire" style messages using its own signalling alphabet.  For 
amateur use
ALE signaling alphabet is very robust and can  "signal through" several 
types of
interference. However, some believe its multi-tone structure (it is base on a
five tones out of eight signalling alphabet) is not legal under US amateur 
rules. At least
one PC implementation of a selected group of ALE functions completely 
eliminates
the ALE's multi-tone signal structure replacing it with conventional ASCII.

THE IDEA
Many features of ALE are interesting in themselves, however,  selecting 
those ALE
features used to adapt to changing signal quality would improve the quality 
of an PSK31
transmission but would maintain the teletype flavor that Peter M was trying 
to create.
I only want to lay out the idea here rather than provide an exhaustive list.
When conditions were good  a PSK31 connection would operate in a 
lightweight mode.
No error correction,  retransmission or acknowledgement would be transmitted.
When the channel begins to degrade the system would switch to bit interleaving
with some light forward coding to overcome single bit errors.  If the channel
continued to degrade additional forward error coding would be applied 
resulting in
a reduction of the channel throughput for user information.  Busty 
interfering signals might
be dealt with by selective retransmissions.  A mechanism will also be 
required to
monitor channel quality, decide on a more or less robust mechanisms and 
signal the
other side.

CONCLUSION
The PC sound card has allowed several new modulations schemes to be built 
such as JT44. However,
not much work is being done with HF computer protocol because of the 
variable nature of the
HF environment.  Combining a popular modulation technique with a computer 
controlled adaptive
protocol could give us a way to utilize a channel that is not ideal.