At 08:44 AM 3/7/2005, Tom Rauch wrote:
> > towards the signal of interest. Simultaneously, you are
>computing another
> > pattern that has a strong lobe on the BPL power lines
>nearby with a null in
> > the direction of your signal of interest. You use the
>second pattern to
> > filter and extract non-coherent noise information and
>adaptively use it in
> > an LMS algorithm to cancel BPL noise from the first
>pattern.
>
>I doubt that is practical at HF receiving random signals.
>The phasing system would have no idea what was signal and
>what was noise, and you can't have two patterns at once on a
>single output port.
It is practical, it's been done, for HF skywave links. The key is that the
signals are not really random (neither the noise/interference nor the
desired signals are truly "white".. otherwise you wouldn't be able to copy
CW below the average noise/interference floor in a pileup).
Practical systems of this type actually form several adaptive sums. One of
the more common strategies works on a sort of successive removal
scheme. Find the strongest signal, adapt to it, subtract that signal from
the various elements, then find the next strongest, etc. A good tutorial
paper on the state of the art at that time would be
W.F. Gabriel, "Adaptive Arrays-An Introduction", IEEE Proceedings, Feb
1976, page 239-272
At that time, LMS adaptation algorithms were quite popular. Early systems
used arrays of PLLs, probably in the late 40s, early 50s time frame. There
were two groups of folks, one attacking the problem from a control law
approach (Howells-Appelbaum), the other from an adaptive filter approach
(Widrow).
For simple "multiple parallel loop" schemes, there are probably as many
architectures as there are ways to do loops (limiting, or not, cross
multiplying, Costas types, etc.etc.etc.)
At that time, though, the processing was still largely inherently parallel
and in time lockstep. You'd essentially have N complex weights for N
elements, and you'd just adjust those weights according to some rule. (A
complex weight is a phase shift and an amplitude factor). Any time domain
data was used only within a channel (say, by averaging, or a narrow band
filter, in each channel).
There are some even more sophisticated approaches. These days, most of the
techniques rely on combining information across both time and space.
Imagine a tapped delay line on each element, some M taps long, so now
you're combining NxM complex weights. The length of the delay line is
usually chosen to be somewhat longer than the coherence time of the signal
(after propagation). Adding in some a-priori knowledge of the signal can
help a lot (for instance, SSB voice has a lot of internal structure that
can be exploited. CW has even more).
They all rely on the fact that the interference (even coming from skywave
sources) is NOT uniform, particularly on a short time scale. There are a
number of researchers using this to measure lightning densities world wide
(since that's the dominant source of atmospheric noise from long distances...).
Most powerline related noise has strong periodic components that are quite
susceptible to adaptive removal.
The truly horrible thing about BPL, in fact, is that the signals are
expressly designed to make them spectrally white and uniform (to minimize
interference potential and susceptibility), which makes them hard to
adaptively remove.
>You must be talking about an entire receiver chain including
>detectors, rather than the system Jim described. In that
>case you could sample broadband noise (as long as it wasn't
>"hiss") and subtract it out. Of course that involves a
>complex receiver chain rather than the complex adaptive
>phasing system originally described.
I would think that any phased array receiver would do most of the
processing in the digital domain, and I would assume that you're doing both
space and time processing.
In a simple passive phase shifter scheme with a limited number of elements,
one might not do this. It's not clear to me, just right now, whether
sophisticated processing is "worth it" for small numbers of elements (say,
less than 4-6). There's been some promising research.
However, a multi channel receiver can be potentially very cheap (no high
power amplifiers to worry about) on a per channel basis. Each channel
needs a suitable front end mixer (probably with a huge LO drive, to get
good IM3 performance), a reasonable IF strip and then some high performance
A/Ds. Such a thing isn't available at a ham-suitable price yet (and may
never be..).
The real problem is that while the receiver design is seemingly
straightforward, there are an enormous number of traps for the
unwary. It's not just a matter of getting the Mini-Circuits catalog and
ordering up a bunch of ZFY-1 +23dBm mixers.
If one were interested in experimenting, one might start with a rack of
PCR-1000 receivers, which are easily controllable, have a fairly wideband
IF filter available, and can be modified to all lock to the same reference
source. However, I doubt whether the PCR-1000 would be considered "contest
grade", even though they at least have the advantage of being reasonably
well packaged and a mass-produced product. One might also consider using a
rack of SDR-1000 type DDS based radios, but those are even less
refined. You'd be doing a lot of integration work, on top of a pretty
herculean software development effort.
But, based on retail pricing of units like the SDR-1000 and PCR-1000, I can
see a multichannel receiver costing about $200/channel + $1000 for the
shared part (if it were ever to reach production quantities). Given the
enormous interest in "Smart Antennas" for the wireless telecom business, in
the next 5-10 years, there may well be off-the-shelf receiver hardware for
low-microwave bands that hams could scrounge that might serve. You'd
translate your HF signals up to a 1-2 GHz IF. Cellular companies are going
to be pulling out all the first generation DSP systems (if they haven't
already) and they might be easily reprogrammable.
If you had a BIG budget, you could just go to Pentek or Racal and buy the
hardware you'd need.
>Peaking a near noise floor or below noise floor random
>signal (like a weak DX station) is another much more
>difficult matter.
Difficult, but doable (in a theoretical sense).
>Let me know when you get one running!
Probably 5 years in the future, at least. Ham use of software radios is
pretty new still, and that needs to mature a bit, and, with Moore's law,
the processor side keeps getting better.
>73 Tom
_______________________________________________
See: http://www.mscomputer.com for "Self Supporting Towers", "Wireless Weather
Stations", and lot's more. Call Toll Free, 1-800-333-9041 with any questions
and ask for Sherman, W2FLA.
_______________________________________________
TowerTalk mailing list
TowerTalk@contesting.com
http://lists.contesting.com/mailman/listinfo/towertalk
|