Regarding 56k+ on UHF it can be done, at east commercially. I worked for
DataRadio which was owned by Bob Rouleau VE2PY; and Norm Pearl VE2BQS of
the Montreal Packet Net originators of ham radio packet. When I worked
there in 2000 we were implementing 56k on UHF using parallel decode
technology that used 3 Z80 processors in the radio. I installed a system
in MT and conducted drive testing in a police car at speeds up to 108mph
(guaranteed to 80mph but you tell the guy with the gun to slow down). It
worked with few dropped packets even in areas where multi path was a
problem on their voice systems. I don't know what they had going on in the
DSP and the parallel decode it was all kept very much from us in the
Atlanta office, however it almost seemed to like the multi-path exceeding
the coverage plots generated by the very expensive software predicted.
The radios were manufactured in WI (they were dumb radios at this point
with no firmware to speak of. The radios were then shipped to DataRadio
Canada for DSP and firmware loading then shipped back to DataRadio in
Atlanta for distribution to the specific customer. (not shure but I think
there was some sort of tax advantage to this process or they must not have
wanted the DSP process that was radio specific to go out to the US
companies. We then built the base stations (Tait repeaters with 2 RX units
and a DataRadio Base controller) and shipped them to the local installer.
One of the engineers would show up for testing after the first 10 cars were
installed as well as the base stations. If you see one of these radios for
cheep on Ebay or at a ham fest it is likely 800mhz, but pick one up just
for disection. They will be labeled Gemini PD, or PD+ I would love to look
at the code for what all they were doing with Z80 processors.
http://www.amateurradio.com/upcoming-37th-anniversary-of-packet-radio/
Lin N4YCI