Introduction
Chapter 7 finally summarises the complete project work. The chapter also
presents improvements which can be incorporated in future designs. The entire work
has been finally concluded.
Chapter 2
Overview of DF
2.2 DF Technology
In the most general sense, all non-rotating radio direction finding systems
employ a DF antenna having an array of spatially-displaced aerials (also referred to as
“elements”, three or more being required for non-ambiguous operation) that are
illuminated by the received wavefront [2]. The resulting output voltages produced by
these aerials exhibit characteristics (phase, amplitude, or both) that are then measured.
Since these characteristics are unique for every received azimuth in a properly designed
DF antenna, the wavefront angle-of-arrival (bearing) can be ascertained by
appropriately processing and analyzing the aerial output voltages.
5
DF DF DF BEARING DF BEARING
ANTENNA RECEIVER PROCESSOR DISPLAY
antennas. Referring to a 4-aerial Adcock configuration, the first of these antennas is the
N-S bi-directional array comprising the north and south aerials. As illustrated in Figure
2.2 below, the resulting figure-of eight azimuthal gain patterns consists of circular
lobes with maximum sensitivity to the north and south and nulls to the east and west.
This figure-of-eight gain pattern is obtained by applying the N and S aerial voltages to
a differencing network that vectorially subtracts them (N-S) to produce what will
ultimately become the “Y-axis” voltage.
N-S
E-
W
SENS
E
The
second of these
antennas is the E-W
bi-directional array
comprising the east
and west aerials.
Again as illustrated in
Figure 2.1, its
azimuthal gain pattern is identical to that of the NS bi-directional array, but
perpendicularly oriented (as a consequence of the fact that the two arrays are physically
at right angles to each other). This pattern is again obtained by applying the E and W
aerial voltages to a differencing network that vectorially subtracts them (E-W) to
7
produce what will ultimately become the “X-axis” voltage. The fig 2.3 below shows
how these aerials are placed in an array.
The third of these antennas is the omni-directional sense antenna [6]. This
omni-directional sense azimuthal gain pattern is also illustrated in Figure 3.2. The sense
antenna is required to resolve a 180° ambiguity that would otherwise result. Since the
Watson-Watt DF technique falls into the amplitude-comparison category as discussed
above, the purpose of the remaining elements of the DF system (i.e., the DF receiver,
DF bearing processor, and DF bearing display) is simply to measure the X- and Y-axis
voltages and then compute and display the bearing. As mentioned above in amplitude-
comparison DF systems, the relative amplitudes of these two voltages are unique for
every wave front angle-of-arrival. They can therefore be “mapped” into a
corresponding bearing using an appropriate algorithm that performs a computation
based on their ratio.
In order to produce the DOA estimate, the voltage output from both antenna
pairs of Adcock antenna is compared [3]. The same can be expressed:
πd πd
j sin φ −j sin φ πd (2.1)
VNS = VN − VS = e λ
−e λ
= j 2sin sin φ
λ
j
πd
cos φ −j
πd
cos φ πd (2.2)
VEW = VE − VW = e λ
−e λ
= j 2sin cos φ
λ
8
2π d (2.3)
VNS ≈ sin φ
λ
2π d (2.4)
VEW ≈ cos φ
λ
The north-south pair can be treated as generating the y-axis voltage while
the east-west pair creates the x-axis voltage for the array’s coordinate system. This
operation essentially uses the two voltage measurements to locate a point in an
abstracted plane, the angle of this point corresponding to the DOA of the received
signal. The arctangent of the quotient of the north-south voltage and east-west voltage
can be expressed as:
V
φˆ = tan −1 NS (2.5)
VEW
2πr
r (t ) = +
A cos 2 ftπ cos 2πf
t ( ) (2.6)
0
λ r
Where fr is the
inverse of the time taken to sample around the entire antenna array. The Doppler shift
imposed on the signal will be directly proportional to the rate of sampling around the
array. When the sampling approaches the AOA of signal of interest as well as 180
degree away, the measured Doppler shift will cross zero. If the array is sampled at 90
or 270 degree from true AOA then measured Doppler shift will be at a negative and
positive maximum respectively.
M- ZERO
ELEMENT RF FRONT FREQUENCY
ARRAY CROSSING
END DISCRIMINATOR DETECTOR
directly proportional to the DOA. One of the major drawbacks to this system is
decreased listen-through capability due to FM and AM artifacts in the signal due to
sampling.
2.3.1.1 Site Error Suppression Site errors are fundamentally the result of
anomalous conditions at or near the DF antenna that result in various distortions in the
apparent angle-of-arrival of the received wavefront. As a result, the apparent angle-of-
arrival may be different than the true angle-of-arrival. The biggest contributors to site
errors are usually reflecting objects causing multi-path reception.
Unlike an Adcock-based Watson-Watt DF system whose maximum DF
antenna aperture (Adcock aerial pair spacing) can be no greater than 1.22 wavelengths
at the highest operating frequency, there is no theoretical limit to the aperture of a
pseudo-Doppler DF antenna. The aperture can be increased without bounds provided
that additional aerials are appropriately added (theoretically, the maximum separation
between adjacent aerials must not exceed 0.5 wavelengths at the operating frequency to
avoid ambiguity, although in practice this separation should be considerably less). A
wider aperture with more aerials permits greater wavefront averaging, which in turn
tends to average out errors caused by multi-path reception.
such that aerial re-radiation causes negligible bearing error, with the result that Watson-
Watt DF systems are capable of excellent bearing accuracy even at the resonant
frequency of the aerials. Adcock DF antennas thus do not require shortened aerials or
other measures that compromise sensitivity to preserve bearing accuracy.
2.4 INTERFEROMETRY
The Interferometry DF [1], [9] determines the angle of incidence of a wave by
directly measuring the phase difference between the signals picked up at different
points on the received wavefront by the elements of antenna array as shown below in
figure 3.6.
14
INPUT
SIGNAL
D cos φ
D
ANT 1 ANT 2
V1 V2
2π (2.7)
V2 = V exp jwt − X
λ
Where
V = the initial transmitted signal amplitude
X = the distance traveled
2π
λ = free space propagation constant
Antenna 1 the voltage will be:
2π (2.8)
V1 =V exp jwt
− +
X D cos φ
Where D cos φ λ Represents the
2π
This above equation can ψ= φ
D cos be solved and AOA can be
expressed in terms of
λ frequency:
cos φ = 30Vψ
(2.13)
2π Df
CHAPTER 3
DESIGN CONSIDERATIONS FOR DF SYSTEMS
DF ANTENNA
RECEIVING DF OUTPUT
SYSTEM PROCESSOR DEVICE
Factors to be considered in the selection of antenna arrays [5], [6], [11] are:-
1. Coverage (the range and azimuth sector over which the target emitters
are located). This will determine the form of the array.
2. Expected propagation modes of the arriving signals. These will
determine the required elevation response of the antenna array and the type
of antenna elements to be used.
3. Combination of measurement speed and measurement accuracy. These
will determine the required aperture of the antenna array.
4. Physical requirements of the DF system (fixed or movable). These will
affect the physical form of the elements and the size of the array.
22
ANTENNA ANTENNA
SWITCH
SINGLE CHANNEL
RECEIVER
TO DF PROCESSOR
N - CHANNEL
RECEIVER
TO DF PROCESSOR
CHAPTER 4
DF SYSTEM DESIGN AND RESULTS
DF ANTENNA
PHASE
DETECTOR
FILTER 2- WAY POWER
DIVIDER
2.4 - 2.5 GHz
AMPLIFIER
DF PROCESSOR
DISPLAY
START
PLACE TRANSMITTER
0
AT 0
CALIBRATION TABLE
different time and location and finally the lookup table is developed after averaging all
the readings. Once the lookup table is developed, the test phase starts. During this
phase the source is placed in any direction randomly and again the phase differences
are measured and the correlated with the lookup table data and where maximum
GIVE
accuracy can be increased based NAME
on requirement. The method of 0 DEGREE
interpolation has been
discussed in depth in next chapter.
4.3 Experimental Results
INCREMENT THE
DIRECTION BY 1 DEGREE
27
4.3.1 Antenna system design A uniform circular array (UCA) which provides
360 degree azimuthal coverage has been designed and fabricated for this project and
steps under taken in the designing and various result are as under:
(a) Theoretically a monopole antenna was designed which will act as each
element in the UCA. The design equation is as given below
(6.1)
h + a = 0.24λ0
where
h = height
a = diameter
λ0 = wavelength
This equation is valid for infinite ground plane and hence when the
monopole is placed in a UCA along with the finite ground plane then
optimization is required to be done to get desired results.
Optimization formula is straight forward and is as given below:
(6.2)
f1l1 = f 2 l2
(b) Theoretical design was then simulated on IE3D software. Basically all
three antennas were simulated together in a UCA as shown below and also
the radiation pattern is given below in figure 6.4 & 6.5. The antenna was
fabricated and then tested on network analyzer.
28
Top View
3- D
side view
29
Fig 4.5 Radiation pattern when all three monopoles are excited
Individual monopole antenna and the circular array of antenna were then
tested on network analyzer and the results are as given below.
(b) PCB design of phase detector module was done using Eagle software
and is shown in Figure 4.10.
(c) Phase Detector module was tested using RF transmitter as shown in figure 4.11
33
PHASE
DETECTOR
MODULE
4.3.3 Power divider design Wilkinson power divider [14] model has been
utilized to design the equal split power divider to divide the antenna output and feed it
to the phase detector module. Circuit configuration of N-way Wilkinson power divider
is as shown below in fig 4.12. The N-way Wilkinson power divider/combiner provides
matching of all ports, low loss and high isolation between input and output ports due to
additional ballast resistors.
degree to 0 degree phase difference respectively. The amplifier increases the margin for
the ADC of Atmega 16 microcontroller by 2.47 times and this even helps in the final
correlation algorithm. The amplifier has been designed using a general purpose J-FET
quad operational amplifier TL 084 which is a low noise operational amplifier. Three
out of four amplifiers of TL 084 has been utilized in non-inverting configuration as
shown below. The value of R2 is 10k and R1 is 6.8k.
CHAPTER-5
37
ALGORITHM DEVELOPMENT
& IMPLEMENTATION
The final part of this work was concerned with the proof-of-concept
implementation of the DF algorithm discussed in previous chapter. Initially the DF
algorithm was simulated on matlab by taking readings manually and this provided the
starting point for the implementation of the algorithm on a fully integrated system. In
this chapter, we present the implementation of the algorithm including both hardware
and software platforms, the test environment, as well as the success and failures
encountered in the process.
The phase detector module has been developed using IC AD 8302 and the
details have been discussed in chapter 4. The output of the phase detector is a phase
internally calibrated in terms of voltage and the calibration is given in figure 5.2.
The phase detector unit measures gain/loss and phase up to 2.7GHz and
accurate phase measurement scaling is 10mV/degree. It operates from supply voltage
of 2.7 V-5.5 V and also provides stable 1.8 V reference voltage output. The phase
output varies from 0 degree (1.8 V) to 180 degree (30mV). The developed module is
shown in figure 5.3. The output of phase detector is utilized in the voltage form itself
i.e. it is not being converted to degrees. This output is given to DF processor via
amplifier unit made using TL 084 whose design have been discussed in previous
chapter.
Tx
2.4GHz
Power Phase DF
Amplifier
Divider Detector P
R
O
Phase C LCD
Antenna Power
Detector Amplifier E Display
Base Divider
S
S
O
R
Power Phase
Amplifier
Divider Detector
Actually the lookup table has to be developed in the location which is free
from EMI and even environmental radiations and for this the best suited place is an
anechoic chamber with a turntable which can rotate the DF for 360 degree and the
phase difference data can be stored in the memory as lookup table. But for this project
as we were developing a prototype with just three monopole antenna a simpler method
was adopted to develop the lookup table. Also the readings were taken only in the first
quadrant and at gap of 5 degrees. The complete system was setup in the Antenna lab
and the transmitter was moved manually. The different sets of readings were taken at
different time of the day and then averaging was done to build the lookup table and the
readings along with the lookup table is attached as appendix to this report.
The complete algorithm was simulated on matlab by loading the developed
lookup table and then feeding the phase data for the transmitter direction manually. The
matlab code of the same is attached as appendix. The system was giving 85 % accuracy
in the manual DF; this might be because of the environmental and human error.
The output of the three antennas is given to power divider unit and then
given to the phase detector unit as shown in figure 5.4 and the output of phase detector
is as given below.
41
phase(V1 − V2 ) = θ12
phase(V2 − V3 ) = θ23 (5.2)
phase(V3 − V1 ) = θ31
During the training phase this phase data is stored in the memory in the
lookup table at respective location based on the direction of training source. If it is the
test phase this data is stored in memory and correlation is performed with the lookup
table. The correlation algorithm is as given below for M directions.
TLOOKUP = M M M M (5.3)
ϖ 12 ϖ 23 ϖ 31 M *3 900
M *1
θ12 θ 23 θ31
TTEST = M M M (5.4)
θ12 θ 23 θ31 M *3
( ϕ −θ ) 2 + ( ϕ −θ ) 2 + ( ϕ −θ ) 2
0
0
12 12 23 23 31 31
TCORR = M M (5.5)
0
( ϖ − θ ) 2 + ( ϖ − θ ) 2 + ( ϖ − θ ) 2 90 M *1
12 12 23 23 31 31 M *1
limitation will also give rise to reduced DF accuracy dependent on the resolution of the
lookup table. Now to remove this dependence of DF accuracy method of interpolation
has been used to provide higher resolution for our DOAs.
Interpolations are of different types like linear, polynomial, piecewise
constant, etc. Method of linear interpolation has been utilized in our DF system for
increasing the accuracy. It is the simplest method of getting values at positions in
between the data points. The points are simply joined by straight line segments and
each segment can be interpolated independently based on requirement. The method of
interpolation is required if the minimum Euclidean distance found using equation 5.5 is
not zero because if it is zero then perfect correlation and it will give out exact DOA. If
it non zero then the value and the index i.e. degree shown in equation 5.5 has been
utilized for doing interpolation.
y0 = array (index)
y1 = array (index 2)
d0 = 0
(5.7)
d1 = euclidean _ dist (index 2)
d 2 = euclidean _ dist (index )
y2 = degree output
The above data is required for performing linear interpolation. This data is
then fed in the equation of interpolation to find out the required degree output (DOA).
y −y
y 2 = y1 + 0 1 ( d1 − d2 ) (5.8)
d1 − d0
The method of interpolation has been implemented on matlab and the code
is attached as appendix to this report for future development.
Chapter 6
Advance DF algorithms for multiple signal classification and
Range finding
6. The width and height of the peak does not bear any relation to any
relevant property of the signal.
frequency of the signal, k wavenumber and nm is zero mean additive white complex
Gaussian noise of variance σ 2 . The output of all the sensors forms a matrix Y which is
given as
Y = ( y1 ( t ) y2 ( t ) L L L L ym ( t ) )T (6.2)
Y = As + n (6.3)
where A, the steering vector, s, the signal vector and w the phase delay between two
sensors are as given below
s = [b1 ( t ) e jwt b2 ( t ) e jwt L L L bn ( t ) e jwt ]T (6.4)
A = [a ( w1 ) a ( w2 ) L L L L a ( wn ) ] (6.5)
a ( w1 ) = [1 e jw1 e j 2 w1 L L L e j ( M −1) w1 ]T
wl = kd sin ( θl ) (6.6)
It is assumed that the signal and noise are uncorrelated. The covariance
matrix of data vector would then be
R = E YY H (6.7)
R = AE ss H AH + E nnH
R = APAH + σ 2 I
Now carrying eigen decomposition of matrix R we can partition the signal
subspace and noise subspace eigen vector matrix which is as given below
Λ 0 U H
R = [U V ] 2 H (6.8)
0 σ I V
46
vi + σ 2 L L L i = 1, 2,K K n
λi = 2
σ L L L L i = n + 1,K K m
where U and V are signal and noise subspace vectors respectively. By eigen analysis
we can represent the smallest eigen vectors as
Rvi = σ 2 vi L L L i = n + 1,K K m (6.9)
APAH vi = 0L L i = n + 1,K K m
If R̂ is eigen decomposed then we would find estimate of noise subspace eigen vector
matrix but Vˆ would only be an estimate hence AV H = 0 would not be true but it will
be the minimum at true DOAs and by utilizing this fact the DOAs can be plotted using
the equation given below.
1
Pmusic = H
ˆ ˆ a ( w)
a ( w)VV
H (6.11)
the above data model and result of eigen decomposition has been simulated on matlab
and the results are discussed in succeeding paras.
Fig. 6.2 Music spectrum with two signal sources at 20 & 60 degrees
Fig. 6.3 Music spectrum with three signal sources at 20, 60 & 100 degrees
48
Fig. 6.4 Music spectrum of two near by signal sources at 50 & 54 degrees
The simulation results of the MUSIC algorithm show the following results.
1. The capability to resolve multiple targets with separation angles
smaller then the main lobe width of the antenna array.
2. The estimation errors can be reduced by increasing the number of
snap shots, but there has to be a trade off between the speed and number of
snap shots.
3. The estimation error increases as the angle separation becomes
smaller between signals.
4. Spacing error has to be taken into account during the practical
implementation of the algorithm.
5. The number of signal sources should be less then number of
antenna elements.
6. The performance of MUSIC estimator suffers a progressive
degradation as the SNR is reduced as it causes increase in the covariance
matrix estimation error and corresponding increase in DOA estimation
error.[]
49
Area of
Location
uncertainty
φ1 φ2
DF DF
SITE SITE
(I) (II)
50
SOURCE 1
SOURCE 2
DF 3
DF 2
DF 1
51
Fig. 6.6(a) Result showing exact location with two DFs with error free DOAs
Area
Area of
of Location
Location
uncertainty
uncertainty
52
Target
Target
Fig. 6.6(c) Result showing exact location with three DFs with error free DOAs
Fig. 6.6(d) Result showing exact location of target 2 with three DFs
The above results show that DF accuracy is the most important parameter to
be kept in mind before designing a DF system. Based on operational requirement this
53
parameter can be decided because higher the accuracy means higher cost, design
difficulties, etc and lower accuracy will lead to false target localization.
CHAPTER: 7
Conclusion and Future Work
7.1 Conclusions
The basic principle of Direction finders has been studied. The initial
literature survey covers the various types of DF techniques, pros and cons with
different algorithms. Due to certain disadvantages of amplitude comparison technique
such as complexity in antenna design, limited frequency range (up to 1 GHz), poor site
error suppression, and etc. phase comparison DF technique has been chosen for
implementation. The design parameters and requirements have been studied and
described briefly and the same has been used for further hardware design.
The DF system has been designed for the frequency range 2.4-2.5 GHz.
Being just a prototype the azimuth coverage is only 90 degrees and also the lookup
table has been made for 0 to 90 degree with 5 degree resolution. The modular design of
DF system has been presented in this project report which can be integrated on a single
PCB in future, based on requirement. The monitoring section has not been included in
this DF system due to paucity of time, there are various receivers available in the
market so based on the frequency requirement any receiver can form the monitoring
section of the DF system.
The DF algorithm used for the system has been tested for consistency and
correctness by simulation on matlab and then implemented on the actual system. All
the matlab codes and c codes have been attached as appendix and results have been
included in the report. To increase the DF accuracy, method of interpolation has been
utilized and the same has been tested by developing an algorithm and implementing it
on matlab. Even the range calculation and localization of the source algorithm has been
developed and implemented on matlab and codes attached with the report for future
developments.
REFERENCES
Technique.” www.rdfproducts.com
5. David Adamy, “EW 101 A First Course in Electronic Warfare” Artech House
Boston London.
www.rohde-schwarz.com/www/downcent.nsf/file/chapter1_introduction.pdf
schwarz.com/www/downcent.nsf/file/chapter3
10. Cheol-Sun Park, Dae Young Kim, “The Fast Correlative Interferometer DF
11. Balanis, C. A.: Antenna Theory. New York Harper & Row 1982.
12. RF Products, Web Note WN-003 “Radio Direction Finding receiver and bearing
processor.” www.rdfproducts.com
56
13. Donald E Allen, “Channelised receiver a viable solution for EW and ESM
14. David M Pozar, “Microwave Engineering” Second edition John Wiley & Sons.
Inc 1998.
16 local.wasp.uwa.edu.au/~pbourke/miscellaneous/interpolation