[hpsdr] Beam Steering

[Don Jackson AE5K]

The following message by Vic K1LT appeared on the "dttsp-linux" 
discussion list earlier today, and with Vic's permission, I am reposting 
it to the HPSDR list as I believe there is most likely much interest in 
this area by our HPSDR list membership.  Any replies should be posted to 
both our HPSDR list and to Vic directly ( vkean at ds.net ).  Vic is not a 
member of the HPSDR list (yet).  [Do not reply to me!]
--Don, AE5K

-------- Original Message --------
Subject: [dttsp-linux] Beam Steering
Date: Fri, 5 Oct 2007 15:50:54 -0400
From: Victor A. Kean, Jr. <vkean at ds.net>
Reply-To: dttsp-linux at yahoogroups.com
To: <dttsp-linux at yahoogroups.com>

Fellow SDR enthusiasts:

I have constructed an electronically steerable phased array antenna
(and receiver) system using 8 short verticals, 4 softrock v6.1 radios,
a DDS 6.0 synthesizer, a Delta 1010LT sound card, DttSP, SDR-Shell,
and a blob of software that I wrote (and, of course, a computer
running GNU/Linux).  I'm also using WA6UFQ's DDS controller software
on my laptop to tune the synthesizer.

Each vertical is 23 feet tall with 4 23 foot top hat wires sloping at
a 45 degree angle.  Each vertical has 16 23 foot radial wires.  Each
vertical is brought to resonance on 1.83 MHz with a 30.5 uH coil and
matched to 75 ohm RG6 feedlines with a 68 ohm resistor.

Pairs of verticals are separated by 70 feet, and fed through 40 foot
pieces of RG6 to a phasing box with a couple of transformers and a
relay and a 56 foot piece of RG6 as a phasing line.  So a pair of
verticals constitutes an end-fire array that can be remotely switched
in direction by sending 12 volts down the feedline.

There are 4 pairs of verticals in a line, separated by 290 feet, which
is just 54% of a wavelength at 1.83 MHz.  The whole array has a
bore-sight bearing of 65 degrees.  Each end-fire element is fed with
1000 feet of (Dayton special) RG6 back to the shack.

At the shack are 4 softrock v6.1 SDR receivers tied together from the
first receiver's quadrature circuits.  That first receiver is fed
from the DDS 6.0 synthesizer.  The receivers have been manually
tweaked to have similar amplitude and phase responses.  The band-pass
filter at the input of the receiver appears to make the largest
contribution to varying frequency response.

The I/Q signal from each receiver is fed into a Delta 1010LT sound
card.  The Delta 1010LT is the same as the Delta 44, except that it
has twice as many input and output channels (8 versus 4).  It uses the
same driver on Linux as the 44 as well.  The first pair of channels is
configure for microphone level inputs rather than line level,
supposedly removed by jumpers.  However, the first two channels are
inverted in the process with respect to the other 6.

My computer has enough horsepower along with the low-latency version
of the 2.6.18 kernel to run 3 simultaneous DttSP and SDR-Shell
instances.  4 instances causes a lot of xruns.

I've written a chunk of software that sits between JACK and a single
instance of DttSP and SDR-Shell that sums the 4 I/Q signals and
effectively performs the beam steering.  JACK is a wonderful idea.  My
lump of software, which I've dubbed "Phasor" for lack of a clever name
performs channel "equalization", steering, and summing.  Summing adds
channels with the appropriate binomial distribution according to the
number of elements configured.  Steering shifts the phase of each
channel by the appropriate amount to form a beam.  Equalization
adjusts the amplitude and phase of each channel according to an
external reference signal.

The most difficult part of the process is equalization.  Despite
considerable care, each receiver contributes about a 10 degree
variation in phase and up to 3 db variance in amplitude.  Also, each
end-fire element varies a few degrees in phase and has a db or two
variance in amplitude.

Best performance seems to occur when I use a stable over-the-air
signal as a reference for the "calibration" process, which merely FFTs
each channel, selects the strongest signals, and records the
difference in amplitude and phase from channel 1.  Then the
equalization process merely applies the inverse of the calibration
data to each channel.

I wish to use this array on the 160 meter ham band.  However, the
nearby AM broadcast band provides a very convenient collection of well
documented signals with which to test.  Daytime testing lets me listen
to 2 or 3 three AM radio stations on each frequency merely by banging
on my "steer" key.

So far, I've made one ham radio QSO using this array to work a station
(SV3RF) that I could not hear (very well) with my collection of
Beverage antennas.  And that contact was prior to the most recent
round of improvements.  Hopefully, this array will shine when there
are some African stations on the air.  My long, narrow lot allows long
Beverages that favor northern Europe, but not so much due east towards
Africa.

The other difficulty has been keeping the set of short verticals in
tune.  Since the entire array is an experiment, I did not construct
the verticals with a mind to great longevity.  That was perhaps a
significant mistake.  After some effort this summer, I've finished the
array and retuned all of the verticals to be as similar as possible.
It appears that lightning (or maybe my own transmitter) increases the
inductance of a coil wound on type 61 material when said coil is part
of an antenna.

There are several improvements yet to be made: frequency compensated
calibration / equalization and automatic I/Q balance are the two main
areas.  Alex, VE3NEA shared Rocky's I/Q balance scheme with me, but so
far I have not been able to replicate his success with my software,
although his method works very well in a contrived set-up.  I'm also
looking for some insight into making a digital filter that applies an
arbitrary phase and amplitude correction given some frequency
dependent function.

I intend to write a QEX article or maybe a paper for TAPR when this
project is complete, but I wanted to share my current excitement with
some of the people that have made my efforts possible, especially
Alex, Tony, Frank, Bob, and Edson.  I hope I didn't leave anyone out.

I have a few old pictures up on my Web site of the array as it existed
last year.  I'll try to get some more pictures of the other hardware
up there soon.  Also, a few diagrams would probably be useful.  Go to
http://k1lt.com, but don't expect much :-)

Thanks!

Vic, K1LT

PS: the comments about adding DDS6.0 control to SDR-Shell inspired
this post.  I would like to get that software.


Yahoo! Groups Links

<*> To visit your group on the web, go to:
     http://groups.yahoo.com/group/dttsp-linux/