HF Radio
Direction Finding
Dr. David Sadler
25th February 2010
Roke Manor Research Ltd
a Siemens Company
Contents
1. Overview of HF DF
2. Traditional approaches to DF
3. Superresolution DF
4. Antenna array design
5. HF array elements
6. Digital receivers
7. SRDF software
8. Adaptive beamforming for signal separation
9. SRDF and ADBF demonstration
10.Geolocation systems
11.Concluding remarks
12.References
13.Build a DF system
2
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Overview of HF DF
High frequency band nominally 2-30 MHz, 10-150 m
wavelengths
HF band still used for broadcast, marine, aviation, military,
diplomatic, and amateur purposes
HF radio direction finding is needed to monitor and control the
spectrum:
Identifying interfering sources (civilian)
Locating enemy forces (military)
Signals intelligence
Also need to be able to separate out cochannel signals
Need to be able to handle the unique HF environment
Groundwave and skywave propagation
Potential for correlated multipath
Time-varying ionospheric conditions, fading, polarization changes
External noise is not spatially white
3
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Traditional approaches to DF directional
antenna
Simplest approach for DF is to mechanically rotate a directional
antenna
A peak in the response indicates the approximate signal direction
Not easy to rotate directional HF antennas due to large size
Can use an electrically small loop
Not high accuracy
Problems with polarization no good for skywaves
180 ambiguity
Only needs a single receiver
4
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Traditional approaches to DF Watson-Watt
with loops
Two orthogonal loop antennas
Figure of 8 responses
Cosinusoidal for N-S loop
Sinusoidal for E-W loop
Sense
Direction is the arctangent of the
ratio of the E-W signal to N-S
signal
180 ambiguity can be resolved
using a third omnidirectional
antenna
Needs 3 coherent receivers
Small physical size
Can have ~5 accuracy for
groundwaves
Very poor performance for
skywaves with significant
horizontal polarization
N-S loop
E-W loop
5
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Traditional approaches to DF Watson-Watt with
Adcock antenna
Adcock antenna can use the
Watson-Watt principle
4 antennas: monopoles or
dipoles
2 difference combiners are used
to generate the N-S and E-W cos
and sine patterns
Omni sense signal can be
generated by an in phase
combination of all antennas, or a
fifth antenna
N-S
3 coherent receivers needed
Accuracy still ~5 but much
better than loops for skywaves
6
david.sadler@roke.co.uk
Sense
Roke Manor Research Ltd
a Siemens Company
E-W
Traditional approaches to DF pseudo-Doppler
Pseudo-Doppler DF comprises
Circular array with a commutating RF switch
to approximate the circular motion of a
rotating antenna
The antenna signal is frequency modulated
at a rate equal to the rotational frequency
After FM demodulation the rotational tone is
recovered
The phase offset of the recovered tone
compared to the original tone equals the
direction of arrival
Single receiver
Accuracy often worse than Watson-Watt
due to less sensitivity and intolerance to
receiver imperfection
Latency to achieve DF result
7
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Traditional approaches to DF array goniometer
8
david.sadler@roke.co.uk
Pusher CDAA shown
24 antennas per ring
Outer ring 3-10 MHz
Inner ring 10-30 MHz
Mechanical/analogue
goniometer used to
sweep a beam around
360 azimuth
Single receiver
Lots of equipment /
expensive
Roke Manor Research Ltd
a Siemens Company
Superresolution DF block diagram
Digital
receiver 1
Digital
receiver 2
Digital
receiver N
Digital complex data
Array
manifold
Superresolution digital
signal processing
Number of
signals
Signal 1
weights
Signal M
weights
Signal 1
Signal M
Powers
Bearings
A good SRDF/ADBF system can solve the following:
Detection problem
Estimation problem
Reception problem
9
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Superresolution DF for and against
Superresolution means two signals can be resolved which are
less than one beamwidth apart
An antenna array is needed with multiple synchronous receivers
Subspace techniques are applied to achieve superresolution
Requires knowledge of the array manifold
Multiple antennas and receiving equipment
More sophisticated digital processing
+ Order of magnitude increase in resolution
+ Increased DF accuracy (< 1 error)
+ Azimuth and elevation DF
+ Simultaneous DF of multiple cochannel signals
+ Operation with very few data samples
+ Not fixed to a particular array geometry
The array manifold characterizes the antenna array and
fundamentally sets how good it will be for DF
It is the known array calibration function against which the unknown
signals are compared to find the lines of bearing
10
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Superresolution DF the array manifold
Example for a 3
element array
For a signal arriving at the array from a particular direction, the
set of relative gains and phases at the antennas defines an array
response vector
The array manifold is the locus (curve) of the complete set of
array response vectors for all directions
11
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Superresolution DF MUSIC algorithm
There are many algorithms Capon, MUSIC, ESPRIT, IMP
MUSIC is the most well known of the subspace techniques
1. Correlate the IQ data
from each element with
every other element to
form the data
covariance matrix
Array response vector
2. Eigendecompose the
covariance matrix
3. Separate the noise and
Array manifold
signal subspaces
4. Calculate the projection
of the array response
vectors for all directions
into the signal subspace
5. Look for nulls in the
projection function
Signal subspace
when the distance
eigenvector 2
between an array
response vector and the
signal subspace is at a
minimum we have
found a signal
12
david.sadler@roke.co.uk
Noise subspace
eigenvector
Example for a 3
element array
Distance between
array response
vector and the
signal subspace
Signal subspace
eigenvector 1
Signal vector 1
Signal vector 2
Roke Manor Research Ltd
a Siemens Company
Antenna array design
Array design is critical to DF performance defines the array
manifold
Need an array which exhibits low levels of ambiguity
DF ambiguity occurs when an array has a similar response to signals
which arrive from distinct directions
Grating lobes are perfect ambiguities, large sidelobes are also a
problem
Ambiguity patterns are used to analyze different layouts
Ideally arrays of aligned antennas are set up in clear sites to
avoid polarization effects
For difficult electromagnetic environments, additional array
calibration and polarization processing are needed
13
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
C8 array, 5 aperture
10
dB
10
C7 array, 5 aperture
Highly symmetrical C8 array has very poor performance
C7 array is superior even with less antennas
Array layout optimization is possible simulated annealing
14
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
dB
Antenna array design - clear array site
ambiguity patterns
Antenna array design EM modelling for difficult
environments
30
20
50
10
El=12 deg Fr=16.014 MHz H=green V=blue Pk:49.6
30
N
100
20
-10
0
10
150
NEC model of a Type 22 frigate
Radiation pattern for a deck edge
loop antenna @ 16 MHz
Diverse V and H polarization
response
dB
10
0 W
-10
-20
S
-30
-30
-20
-10
10
20
30
dB
Skywaves usually have unknown
15
polarization
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
HF array elements
Antennas can be passive or active
Generally inefficient antennas are used for DF
Keeps the size down
Low mutual coupling reduced effect on the array manifold
Monopoles are more practical than dipoles
Smaller physical size
Monopoles need to work against the ground plane, no need to
elevate
Loops can also be used
Good for higher elevation skywaves
Suitable for NVIS
16
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
HF array elements Sarsen crossed loop
Sarsen antenna for strategic fixed
sites
Requires a poured concrete footing
Feed cables typically run
underground and enter the antenna
underneath the main pillar
Omnidirectional, broadband
elements (1-30 MHz)
Simultaneous or switched vertical
monopole and cross loop outputs
(RHCP and LHCP)
Monopole primarily for 0-45
elevation, cross loops for 25-90
elevation
Ground mesh and 8 ground radials
with ground rods ensure a good
ground for the elements to work
against
17
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
HF array elements Quadrant crossed loop
Quadrant antenna for tactical sites
Self supporting
Fibreglass and aluminium
construction, weight < 35 kg
Deployable in 90 s
Monopole gain falls off at low
frequencies this is typical for
broadband receive only antennas
18
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Digital receivers DWR16
30 MHz wideband spectrum monitoring
4 independent DDCs each provide a
32 kHz narrowband channel
RF in, sampled IQ data out over
USB2.0
High linearity, no images from LOs
and mixers, high dynamic range
19
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Digital receivers DWR16 PCB
FIFO
(delay)
USB2
Interface IC
4 channel DDC
FPGA
Variable gain
amplifier
RF filters
RF IN 1
USB2
RF IN 2
External
clock
Fast interface
Direct to FPGA
EEPROM
16 bit CMOS ADC
80 MSPS
Onboard clock
Power
20
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Digital receivers DWR16 GUI
FFT and
spectrogram
displays
DDR control
Power level
monitoring
Software
demodulation
AGC settings
Data recording and
playback
21
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Digital receivers MCDWR16
9 DWR16s in a 2U, 19 box
All channels are synchronized,
coherent sampling
Two units can be linked together
to support 16 antenna DF
systems
Two USB2.0 outputs
Channels 1-8 narrowband DF
Channel 9 wideband monitor
22
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Digital receivers benefits of N channel direct
digitization
Near instantaneous signal
acquisition
No calibration required
No need for multiple
coherent local oscillators
Supports DF on short
duration / frequency
hopping signals
Can support reconstruction
of frequency hoppers
Provides broadband beamforming without the need for large
coaxial cable delay lines
Supports ADBF for enhanced signal copy
N channels provides 10logN dynamic range enhancement
23
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
SRDF software
DF processor is receiver independent, the data server handles
the receiver interface and outputs packets over TCP/IP
Up to 4 independent DF processors can run simultaneously
supports the 4 DDRs in the MCDWR16
MUSIC DF algorithm for azimuth and elevation estimation of
multiple cochannel signals
4 different ADBF algorithms
24
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
SRDF software MUSIC result
Single signal incident upon the array: 250 azimuth, 60 elevation
MUSIC is akin to steering nulls rather than beams, so the resolution is
greater
15
dB
10
0
80
60
300
40
200
20
Elevation
100
0
25
david.sadler@roke.co.uk
Azimuth
Roke Manor Research Ltd
a Siemens Company
Adaptive beamforming for signal separation
single omni antenna
pattern
direction of
signal 1
omni reception
direction of
signal 3
beamform
and null
direction of
signal 2
26
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Adaptive beamforming for signal separation
Conventional beamforming sets the array steering weights equal
to the array response vector for the signal direction
10logN improvement in the SNR for the wanted signal
Interferers reduced to the beam pattern sidelobe level
Beam plus nulls takes the conventional beamforming weights
and projects them to be orthogonal to the interferer subspace
10logN improvement in the SNR for the wanted signal
Superresolution with regards to interferer cancellation
In theory very high levels of cancellation, in practice 15-20 dB occurs
due to errors in the array manifold
Steer a beam plus minimize the output power: Wiener-Hopf
solution
10logN improvement in the SNR for the wanted signal
Maximizes the SINR
Works best for strong interferers, can provide 40 dB of cancellation
Can attenuate the wanted signal if the beam constraint is erroneous
Higher Order Statistics methods not based on DF results
Needs larger data blocks to accurately estimate the statistics
Good nulling even when there are significant DF errors
27
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
SRDF and ADBF demonstration
8 element array in
New Hampshire
3 cochannel signals
centred on 16 MHz
1. Over-modulated AM
jamming signal
north of array (10
azimuth)
2. Morse signal from
Havana Cuba
3. 75 baud FSK
teletype from Fortde-France
28
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Geolocation systems - triangulation
The intersection of lines of bearing from multiple DF sites defines
a region where the transmitter is located
Azimuth triangulation errors depend on
Lateral azimuth error on individual lines of bearing
Number of DF sites reporting
Locations of the DF sites relative to each other
In general the 2 site geolocation error is an ellipse, defined by
the lines of bearing RMS azimuth errors
Geolocation
error
Geolocation
error
Site 1
Site 1
Non-ideal
arrangement of DF
sites
Ideal arrangement
of DF sites
Site 2
Site 2
29
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Geolocation systems single site location
Single site location of skywaves is feasible if there is knowledge of
the ionosphere
F2 layer is the most
important for skywave
refraction
D layer tends to absorb
HF waves
Large variations
between day and night
Variation with the
sunspot cycle
MUF determines which
layer is active for the
signal of interest
Critical frequency
Elevation angle
30
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Geolocation systems single site location
Usual assumption is that a single hop has occurred, typically up
to 1000 km range, but could be up to 3000 km range
Use an ionosonde to measure the ionospheric height
Use a model to predict the ionospheric height
Geolocation accuracy depends on
Azimuth RMS error
Elevation RMS error
Ionospheric height error
Validity of the single hop assumption
31
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Concluding remarks
Conventional HF DF techniques were developed during WWII
Simple amplitude comparison techniques
Adcock antenna to handle skywaves
Superresolution processing first developed during the 1980s
Increased resolution and accuracy
Multiple signals
Practical SRDF systems did not appear until the 1990s
More complete understanding of the array manifold and DF ambiguity
Array, cable and receiver calibration requirements
High quality digital receivers
Current SRDF systems harness technology developments in
Optimized antenna array layout
Compact antenna designs with multiple elements
N-channel wideband digital receivers
Powerful signal processing in PCs and DSPs
SRDF systems are far more capable than conventional systems
and are increasingly cost effective
32
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
References
RDF Products Web Note WN-002, Basics of the Watson-Watt radio direction finding
technique, http://www.rdfproducts.com, 2007
D. H. Brandwood, Ambiguity patterns of planar antenna arrays of parallel elements, in
proc. IEE conf. Antennas and Propagation, vol. 1, pp. 432-435, Apr. 1995
D. J. Sadler, Planar array design for low ambiguity, in proc. Loughborough Antennas
and Propagation conf., vol. 1, pp. 713-716, Nov. 2009
R. O. Schmidt, Multiple emitter location and signal parameter estimation, IEEE Trans.
Antennas and Propagation , no. 3, pp. 276-280, Mar. 1986
D. H. Brandwood and D. J. Sadler, Superresolution direction finding at HF for signals of
unknown polarization, in proc. IEE conf. HF Radio Systems and Techniques, vol. 1,
pp. 133-137, July 2000
C.J. Tarran, Operational HF DF systems employing real time superresolution
processing, in proc. IEE conf. HF Radio Systems and Techniques, vol. 1, pp. 311319, July 1997
B. D. Van Veen and K. M. Buckley, Beamforming: a versatile approach to spatial
filtering, IEEE Acoustics Speech and Signal Processing magazine, pp. 4-24, Apr.
1988
Roke Manor Research website: http://www.roke.co.uk
33
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
Build a DF system
Watson-Watt method with loops is simple but effective
At a minimum need
2 loop antennas, orthogonally mounted
2 HF receivers, can be low cost COTS, needs a digital output
Simple DF processor implemented in software in a PC
WinRadio or GNU radio? Direct PC interface
Things to remember
Need coherent receivers locked LOs/clocks
No AGC, or at least synchronized AGC across the receivers
Receiver calibration is needed if the RF filters are not consistent
between receivers
34
david.sadler@roke.co.uk
Roke Manor Research Ltd
a Siemens Company
