Newsgroups: rec.radio.amateur.digital.misc
From: gary@ke4zv.atl.ga.us (Gary Coffman)
Subject: Re: Tropo Scatter Packet?
Keywords: Tropo, tropospheric, scatter, packet, radio, multipath
Reply-To: gary@ke4zv.atl.ga.us (Gary Coffman)
Organization: Destructive Testing Systems
Date: Tue, 6 Jun 1995 17:17:47 GMT
In article <3qtr1p$6f2@solaris.cc.vt.edu> mkeitz@bev.net (Mike Keitz) writes:
>In article <1995Jun3.123659.11005@ke4zv.atl.ga.us>,
> gary@ke4zv.atl.ga.us (Gary Coffman) wrote:
>>In article <3qnpcg$n9h@blackice.winternet.com> rwhiting@winternet.com (Rick
>Whiting) writes:
>>>Is tropo scatter, e.g., on 6M, usable for packet radio links? Is there a
>>>multipath problem that would effect maximum baud rate?
>>
>>Yes and yes, but the multipath affects *minimum* baud rate too. The
>>symbols must be significantly longer or shorter in duration than the
>>average multipath induced fade in order to get through in a form
>>recoverable at the receiver. Very fast or very slow works, it's
>>the mid-speeds that are a problem. Regulations limit our top speeds
>>on 6m, so we have to slow down, and use signal recovery techniques
>>that are tolerant of mid-symbol dropouts, to get good results.
>>
>I'm not quite sure if I follow this. First the troposcatter path is long
>and probably lossy (except for rare times when the band "opens"), so even
>without multipath the received signal could be quite weak. This fact alone
>would limit the maximum baud rate, as a certain "energy per bit" is required.
Troposcatter can have a high path loss (and a band opening wouldn't
be troposcatter, so we can ignore that case). The actual path loss
depends on a number of factors, but can range from not much more than
ordinary ducted path loss for a forward scatter path with favorable
weather conditions, up to a loss on the same order as EME for pure
side scatter.
Either high antenna "gain", or high power is required for good SNR.
High gain 6 meter arrays are very unwieldy, so you're looking at kW
class stations if the receiving setup doesn't implement fairly
advanced techniques. 6 meters is about the worst choice for a band
on which to do troposcatter. The microwave bands are much better
because high antenna gains can be achieved with much smaller antennas,
and because the wavelength more closely matches the most common
scattering objects.
>To be honest, I don't know what kind of path length differences there are in
>troposcatter. But that's the important thing to look at to see how much
>intersymbol interfernce will occur.
Actually, troposcatter is Rayleigh scattering from non-specular
convective cells, dust, or water droplets. At 6 meters, it's
primarily convective cells doing the work since the wavelength
is too large to make effective use of the other mechanisms.
The path length differences are on the order of meters to 100s of
meters, but they are the sum of *many* different paths, each varying
by meters from another. The common volumes of atmosphere shared by
transmitting and receiving antenna lobes determine how many (along
with the weather). Smaller common volumes will reduce the number
of multiple paths, and reduce their length differences. This implies
very high gain antennas with their very narrow lobes will be best.
>>A detector that oversamples across the bit period, not just doing
>>a single sample at bit center, will work best.
>This is plausibe if the fading is rapid (fade duration < 1 bit), likely to
>happen only in mobile situations.
No, the convective cells that do the scattering "boil". This is the
primary factor setting the rate of change of the multipath. The more
turbulent the air, the better the cells are at scattering the signal,
but also the more rapid the changes in path lengths.
>But if the demodulator can "see" a notch
>out of the center of a bit, the filters in the receiver are way too wide for
>the bit rate being used. If slow fading occurs, then the whole bit will be
>gone no matter when it is sampled. For weak, noisy signals, an integrator
>type of processing could be used, maybe that is what Gary means, rather than
>just comparing the receiver output to a decision threshold several times and
>recording the 'oversampled' bits.
Correct. Coherent oversampling with integration is exactly what I meant.
Binning and weighting techniques would be applied before summation, and
intelligence could be applied to discard obvious deviations from predicted
signals, IE signal "holes" and noise "spikes" on the same order as the
sampling pulse duration. This intelligence would need to be adaptive to
the changing conditions of the common volume of atmosphere.
>What is more likely to be useful is a detector that remembers previous levels
>and attempts to subtract out intersymbol interference, i.e. an equalizer. The
>demodulator would probably need to be a linear I&Q synchronous demodulator
>rather than the common frequency discriminator, as when multipath hits one of
>those, the result seems to be very unpredictable.
Yes, but.... The problem with using ordinary equalizers is the rapid
change in the signal path. When ordinary EQ is forced to vary at a rate
on the same order as the symbol duration, it ceases to be very useful.
It can never "train" to the average signal conditions. The binning and
weighting I suggest above would accomplish much the same thing in sample
duration "realtime".
>As a developer of packet modems, this is very interesting to me. I would like
>to see some new standards emerge from this, as ordinary packet equipment and
>protocols seem poorly suited to this type of operation.
It would be an interesting challenge. As I said, I suspect 6 meters is
about the worst band that could be chosen for this type of operation
though. 6 meters would be best suited to *meteor scatter*, and in fact
spectrum just below the 6 meter amateur band is used to transmit hydrological
data by meteor scatter for the government. The advantage of meteor "scatter"
is that it isn't really a scatter mode at all. Rather it is the same sort
of refraction that happens in the E-layer, but from the transitory ionized
trails of meteors. Signal levels are strong, the best antennas have a
main lobe width on the order of 60 degrees to intercept the widest common
volume so the fewest meteor trails are missed, IE modest antennas, and
power levels can be on the order of 100 watts or less. Suitable meteors
arrive in the common volume at the rate of several per minute, and individual
trails last on the order of seconds. So, this is mainly a protocol problem.
We need fast pings to find a trail, then rapidly transmit one or more
packets, then go back to pinging for the next trail. Since SNR is good
during a reflection, heavily overloaded symbols can be used to boost
effective throughput.
Gary
--
Gary Coffman KE4ZV | You make it, | gatech!wa4mei!ke4zv!gary
Destructive Testing Systems | we break it. | emory!kd4nc!ke4zv!gary
534 Shannon Way | Guaranteed! | gary@ke4zv.atl.ga.us
Lawrenceville, GA 30244 | |