[TowerTalk] 4-square questions

[jimlux]

On 3/6/17 9:40 AM, Grant Saviers wrote:
> Re 4sq receive vs yagi - which to listen to on 80 and 40.
>
> For my site, it depends a bit on the yagi, heights, and arrival angles.
> My 4sq receive, 70' per side, DX Eng, rarely beats the F/S of a 2L 80m
> at 156' or the F/S & F/B of the 3L 40m at 140'.  The 4sq usually beats
> an 86' rotatable 80m dipole at 100'.  The RDF numbers suggest this as well.
>
> OTOH, a 40m 2L Moxon at 102' is more often (25%?) bettered by the 4sq.
>
> On 160 it is no contest 4sq vs T vertical, 4 sq to the rescue. With very
> high angle noise my 4sq generally does less well although overall it is
> a great antenna, RDF seems as advertised.
>
> As Jim(s) note, the best strategy is often to null the noise rather than
> aiming azimuth for maximum gain.


there's a lot of interesting and very powerful approaches to nulling 
noise in an array.  It's not just about forming a "negative beam" in the 
direction.  You can use what's called Space Time Adaptive Processsing 
(STAP) in the radar world and adaptively figure out what the interfering 
signal is, and subtract it from the whole input array.  So what you'd be 
left with is your desired signal(s) and things that are sort of "blob 
like" in the spatial or frequency domains (e.g. lightning noise at a 
distance, or broadband interference).

This kind of processing could very easily knock out slowly drifting 
noise sources with underlying structure (like PWM battery chargers), far 
better than some simple time domain impulse noise notcher or a tone 
notcher. And it has almost no comparable implementation in the analog 
domain - you can't get there without digital processing without a room 
full of incredibly twitchy analog circuitry (look at papers by Widrow, 
Griffiths, etc. from the 60s for analog examples)

Those kinds of interferers can be de-emphasized in the spatial 
processing chain (where you *are* forming a null in that direction).

Then, there's just straight out adaptive beam forming on the desired 
signals.

None of this is trivial, but neither is it mind bendingly complex. 
Someone who's done MUSIC or ESPRIT, for instance, could probably whack 
out an initial implementation in a week.

The key things to figure out for ham use would be
a) how quickly do the interfering signals change (so you can pick 
appropriate adaptation rates for the cancellers)
b) how to make the implementation have constant time delay - the 
Flexradio experience pointed up that a fixed delay isn't a huge problem 
(after all, the signal you're listening to came from 100 milliseconds 
away), but if the delay through the processing chain varies, it will be 
impossible to copy CW at reasonable speeds.
c) how many receiver nodes works, and what should the arrangement be.

In general, you probably want your receiver nodes spread in a 2 D array 
(although having some 3D-ness might be useful.. a node at the bottom of 
your transmit radiator, and another at the top, 10-20 meters into the 
air, might be useful.  Or even a node at the top of a tower, or one at 
each end of a Yagi, in combination with a bunch of ground mounted nodes.

But I see some sort of architecture like this for each node:
1) a active whip - there's a ton of these out there ranging from $20 - $150
2) a bandpass filter, lumped LC, passes the ham bands, but not others - $20
3) a HF receiver dongle	- $25  (I don't know if the RTLSDR has the right 
dynamic range, but it's easy to find out)
4) a small single board computer to drive the dongle and stream the data 
over a long wire (or WLAN) back to the shack. (like a beaglebone green 
or BBG wireless) - $30-40
5) power supply (PoE if you're running network cables would be 
convenient.  Solar power at each node might be better if you're running 
wireless)

So call it $100/node

with $1000 in hardware plus a dedicated (inexpensive $500) PC to run the 
signal processing - For $1500 all told, you'd probably have a receive 
system that would outperform just about any sort of Yagi or pair of 
Yagis, on receive.

You wouldn't necessarily get all the bells and whistles and integration 
with the rest of the shack - not that it couldn't be done, but I'm 
thinking more about the signal processing, with the goal of generating 
an audio stream for listening, or feeding something like CW skimmer or 
your digital modes program -   But, again, looking at the Flexradio 
experience 10 years ago, they spent a LOT of time and effort dealing 
with creeping featurism and user experience and compatibility issues, 
not on the core signal processing.

That is one thing.. a yagi on a rotator is a pretty darn intuitive user 
experience. You turn it until the Signal to (Interference+Noise) ratio 
is the best.  One variable to optimize - azimuth, maybe 2 if you have a 
stack and a switching network.