El 10/03/2010 10:51, KH6TY escribió:
Jose,
If you were going to design a mode that filled 2200 Hz, but did not
use SS, and was as sensitive as possible in that bandwidth, how would
you do it?
Tough question. I believe that on HF the best solution so far is Pactor-III
It would have to be highly resistant to fast Doppler shift also, but
minimum S/N would be the most important parameter, as it would be used
at UHF. So far, Olivia 16-500 seems to be the best compromise between
minimum S/N and Doppler shift survival at UHF. The more narrow Olivia
modes, even though more sensitive, do not decode as well if there is
noticeable fast Doppler shift, and sometimes, not at all.
As you add more tones the bin width reduces. The only hope I see is
using wide bins to accomodate Doppler, and perhaps, more tones, but that
is not possible with 3 kHz radios. Perhaps it is a task for some SDR. I
believe wider modes are not a problem in UHF. It may take more CPU
power, and higher powered radios for simultaneous tones.
DominoEx is completely destroyed by the Doppler shift
Doppler is parasitic noise to DominoEx...
and MFSK16 is not tolerant enough to drift to be usable at UHF.
MT63-2000 covers 2000 Hz, has highly redundant FEC, but the minimum
S/N is only -2 dB, so that is not an alternative.
Both seem to have been designed for HF, and MT63 seems to require a
single ray dominant path. At times it works well, but I have not had
luck with MT63, overall. MT63 has many carriers and narrow bins, not
good for multipath with doppler.
What I am looking for is a mode that will copy under the visible and
audible noise on UHF during deep fades, but survives fast Doppler
shift. Olivia 16-500 makes it down to the noise, but not under, during
deep fades. CW by ear is just slightly better than Olivia 16-500, and
the note is very raspy sounding - much like Aurora communications.
But CW requires well trained operators...
There is a paper by Tim Giles about multitone modems for high latitude
HF paths (PhD publication in Sweden) and he avoided sending in
contiguous bins in wide Doppler spread conditions, and reassigned
contiguous bins on the side to have a "wider hat" to catch the path
shifted tones. That sacrifices thruput, but nevertheless, it is
worthless to push nature. In that case, it is better to become its ally,
and to me, wider spaced tones and reusing contiguous bins seems a good
idea. I read it a long time ago and maybe I am not remembering all
details, but it was interesting enough so I haven't lost the big picture.
The 3 kHz channel limit on HF is a straitjacket that might be avoided on
VHF - UHF if clear frequencies are available and you need speed.
Another observation - most stations I copy on ROS 16 are reading a
metric of -12 dB or greater. Only once have I copied a station (using
1 baud ROS) that was measuring a metric under -25 dB. Is the ROS
"metric" supposed to correlate with the path S/N? I ask this because
even the weakest ROS tones at 1 baud are still visible on the
waterfall, whereas weak Olivia 32-1000 signals with a -12 dB minimum
S/N stop decoding just about the time the tones become hard to see in
the noise, but still can be heard faintly. It is a long way from even
-25 dB S/N to -12 dB S/N, so I would expect if the metric is just
another way to say S/N, I would not be able to see the tones, yet I
can, and not only on the ROS waterfall, but on the DigiPan waterfall
as well.
I really don't know what does "METRIC" mean in the ROS case, Skip. I
really did not pay much attention to it, as most times there was packet
or pactor QRM, being ROS so wide. What caught my attention is how bad it
performs under QRM, having seen Olivia 500-16 under similar conditions
unaffected. I believe I know the reasons, as you may as well know, but
won't elaborate further about it on this list.
73,
Jose, CO2JA