Angle Estimation with a Multi ..

Arm Spiral Antenna

Robert P. Penno, Ph.D.

University of Dayton Dayton, Ohio, 45469-0226 (937) 229 3984 [email protected]

Abstract-

This paper presents an analysis of angle estimation using a single aperture multi-mode spiral. Simple, closed-form expressions for the modal outputs of the multi-mode spiral m-e derived and used in the angle estimation analysis. It is demonstrated that in addition to the classical phase comparison method, modern super resolution techniques such as MUSIC and MLM can be applied to multi-mode antenna technology for angle estimation. It is further demonstrated that MUSIC and MLM algorithms can be adapted to work in the presence of jamming. Simulated results show that these new approaches give lower bias and variance and also extend coverage toward bore sight compared to the traditional comparison method.

TABLE OF CONTENTS

1. INTRODUCTION

2. MULTI-MODE SPIRAL ANTENNA

3. ANGLE ESTIMATION USING MULTI-MODE

ANTENNA

4. RESULTS

5. CONCLUDING REMARKS

6. REFERENCES

1.. INTRODUCTION

A variety of approaches may be used to estimate the angle of arrival (AOA) from an airborne platform. These approaches include interferometry [I], Fourier based classical beamforming techniques and the more modem parameter estimation based super resolution methods [2)-[4] such as the Maximum Likelihood Method (MLM) and the MUSIC method. All these approaches require multiple antenna elements. However, issues of maintainability, reliability and application. versatility coupled with cost, size and weight benefits make it highly desirable to minimize the number of apertures on a. given platform. A single aperture multi-mode

U. S. Government work not protected by copyright

Krishna M. Pasala, Ph,D.

University of Dayton Dayton, Ohio 45469·0226 (937) 229 2683 [email protected]

antenna is therefore an attractive alternative to derive AOA estimates from an airborne platform. A multi-arm spiral is one of the best-developed antennas in this class. The classical approach of utilizing these antennas consists of combining the terminal outputs in a Butler matrix modeformer [5] and then deriving angle information from the outputs of the mode-former. Comparison of the phases of the mode-former outputs produces an estimate of the azimuth, which is often the primary angle of interest Comparison of the magnitudes yields an estimate of the elevation angle.

A detailed study of AOA using multi-arm, multi-mode spiral antenna is carried out. Simple closed-form expressions are derived for the mode-former outputs. Using these expressions, ADA estimates m-e obtained using the classical phase comparison method and the more modern MUSIC and MLM methods. ADA estimates must often be obtained in the presence of jamming. It is demonstrated here that it is possible to modify the MUSIC and MLM methods to yield AOA estimates using a single aperture even in the presence of jamming. It may be pointed out that the classical phase comparison method requires a mode-former, a precision component that is expensive. The techniques discussed here require a multi-channel digital receiver that is coupled directly to the multi-arm spiral and does not require the mode-former. A large number of the simulations are carried out to ascertain the statistical properties ·of the angle estimators derived here and also to address the coverage issues. It is shown that MUSIC and MLM extend coverage towards the boresight compared to the phase comparison method. It is also demonstrated that elevation and azimuth can be estimated using a multi-mode spiral in the presence of a jammer.

The material in this paper is organized as follows: Section-Z derives closed-form expressions for the signals received at the outputs of the mode-former. Using these expressions, in Section-S, three different techniques are developed for estimating the AOA Section-4 summarizes the results derived from a number of simulations. Section-S contains a few concluding remarks,

195

2. MULTIMOnE SPIRAL ANTENNA [6]

The geometry of a four-ann equiangular spiral antenna is shown in figure (1). For the sake of clarity, the figure depicts a loosely wound spiral antenna, although. in practice the spirals may be wound very tightly. The radius of the nth ann of an N-arm spiral is defined by,

2:r

fPn =n

The distance along the spiral to a point whose radius is Pn is given by,

Yaxis

X-axis

Figure-l Geometry of the Multi-Arm Spiral.

Using reciprocity, the open circuit voltage at the terminals of an antenna may be expressed in tenus of the current distribution on the antenna excited in the transmitting mode and the incident field and is given by [7],

1 J- -

Voc:=- Eme -Ldv,

IT vol

Here h is the terminal current and It is the current density distribution on the antenna. The determination of the current distribution on a multi-arm spiral is a formidable problem. There is no closed form solution to the integral equation for the current on the spiral. Numerical solutions based on either integral or differential equation formulations are possible [8] but are computationally quite demanding. Simulation studies of angle estimation dependent upon the numerical modeling of a spiral antenna are impractical. Thus there is a pressing need to determine a viable model that enables an efficient computation of the radiation from a multi-arm spiral. An insight into the principle of operation of this antenna facilitates the development of such a model and is presently considered.

196

2.1. Principle of operation of Multi-arm Spiral antenna:

The multi-arm spiral antenna may be considered to be a multi-conductor transmission line. Assuming that there is no mutual coupling between the arms and that the line is infinite, the current on each arm is given by,

'Ai

1(1) = ITe- JI-' (4)

( 1)

IT is the terminal current and 'I' is the length along the spiral from the terminal. In the normal mode of operation the terminal currents are progressively phased and depending upon this phase shift between adjacent arms several modes of operation are possible. Indeed, for an N -arm spiral N distinct modes of operation are possible. The phase shift, (Xk' between adjacent arms corresponding to mode-k is given by

( 2)

21t

(Xk = -k- , k:= 1,2, ... , N. (5)

N

The terminal currents in the N-arms of a multi-ann spiral operating in mode-l are given by,

-/"k

IT•k ::::: 10 eN, k = 1,2, ... , N . ( 6)

( 3)

Though the currents on the N-arms are initially (ie., at 1=0) out of phase, the relative phase difference between corresponding points on the arms changes as the currents on the anus travel outwards. As a result there are regions on the antenna at which the currents on adjacent anus are

in phase and the radiation from this antenna is essentially derived from these regions. To determine this "active" region, (referring to figure (2)) consider the points Ph P2, P3' .... PN which are adjacent to each other on the N arms of the spiral at a distance 'r' from the center. The spiral is considered to be tightly wound and the separation, &, between adjacent arms is taken to be negligible. The difference in the l'ath length between the points Pj and Pk on the first and the kth spirals is given by,

AI= k 2n:r N

( 7)

The phase difference due to this difference in the path length is given by,

( 8)

Figure 2: Adjacent Spirals of an N-Arm spiral Antenna

The total phase difference due to the phase of the terminal current and the difference in the path length is given by,

21t 21t ( 2n:)

8Cl=8Cl --k=-kx- l--r

INN A.

As a result. when

A

r =-,8a=O, 21t

( 10)

the currents at PI and Pk are in phase. Thus a circular region on the multi-arm spital whose radius is equal to ("JJ21t) corresponding to a circumference of 1 A. constitutes an active region. Radiation from this region is defined to be mode-l radiation. Similarly, it is possible to show that by adjusting the initial phase progression of the terminal currents to (- 2x21t1N) , the active region shifts outward to a circular region of radius 2X1J(21t). corresponding to a circumference of 2A.. The radiation from this region is

defined as mode-2 radiation. In general, for an N-arm spiral N distinct modes of radiation exist, Radiation in

mode-k is obtained by phasing the terminal currents to have a progressive phase shift of (-kx21t1N). Such a phase shift corresponds to an active circular region of circumference leA. Using such a physical picture it is possible to model the currents in the active region and from these modeled currents deduce the radiation patterns of the various modes. The active circular region consists of N segments, each of which corresponds to one of the N-anns. The distance, along the spiral arm, from the actual terminal to this "virtual" terminal is designated by L" and is given by,

(DA h

L = --p 1+-·

n 21t 0 a2

( 11)

( 9)

The azimuthal coordinates of these "virtual" terminals are given by

( 12)

21t (p ) 1 ( nA. )

4> =- -l+-ln-

n,p N a Po

p =1,2, ... ,N.

The currents in the segments comprising the active region are given by

( 13) 21t ct>n,p s <P s <Pn,p + N .

h.p is the terminal current on the pth ann. The current in the active region corresponding to mode-n is given by.

To determine the phase 'JI~ , the equality of this current to the "virtual" terminal current at Rn,l is imposed. That is,

I IA. ) = I eN. e -j"~.,1 Ii

"\'f',,,l 0 '"

= Ioe-i/IL·u",

( 15)

This yields,

Thus, the current in the active region is given by

197

( 17)

where

I : I eNn .

n.O 0

( 18)

2.1 Radiation Properties of a Multi-Ann Spiral:

Explicit closed form expressions are derived here for the received voltages corresponding to any specified mode at the output of the mode-former. These expressions are valid for any given angle of incidence and polarization. By virtue of reciprocity. these expressions also constitute the radiation patterns of the multi-mode spiral array. The antenna is located in the x-y plane and the unit vector in the direction of the propagation of the incident wave is given by:

( 19)

Ui ::::-(sin(So)cos(<po)x. + sin(eo)cos(<po)Y + cos(eo)z) The incident plane wave is given by:

( 20)

The unit vector p specifies the polarization of the incident field. The induced open circuit voltage is given by [7]

( 21)

where V.c,m is the induced voltage at the m" spiral terminal. ~m = t: /1 T,m is the normalized current on the mth spiral arm operated in the transmitting mode, and 1m is the distance along the spiral. The normalized current is given by:

( 22)

1m is the unit vector in the direction of the current at 1",. The mode former output is the weighted sum of the terminal voltages. The output corresponding to the mode-n, MD. of an N-arm spiral is given by:

N

Mn = I. Wn,lV 00,1'

1=1

(23)

where,

_j(l~)1_1)11

W =e N •

n,l

(24)

Thus, the mode-former, in effect, implements the N-point discrete Fourier transform (DFI') of the N-terminal voltages

198

of the Nearm spiral. The voltage received at mode-n is given by:

(25)

N

Mn = L J Eo ~. tm ~j~[x", sin90cos4>O+Ym sin9osin<!lo+Z .. cos901

m=lc",

[w e -j~lm ]dl .

n.m m

Here c", is the contour corresponding to the mth spiral ann. The "active region" contributing to the received voltage corresponding to the nth mode consists of a circular region of

radius ( ~~ ). In this region the normalized current may be modeled as:

T = 1.: i e-M'{l (25)

"Il I n.O 4>'

o

where:

( 26)

The mode-n output is now given by:

( 27)

N

Mn = L J Eo(P . a 41' ~iMx",SineOC05<flo+Ymsineosin<flQ+ZmOOseoJ

m=1s ..

[. -in<fl']dI.

In,Oe m

Sm is the segment of the circular active region corresponding to the mth spiral. The contributions from each of the active circular regions may be combined and the mode-N output voltage may be expressed as

2~

Mn:= JEo{P. a., ~j~rXsineoOOS$o+Ysin9osin$o+zOOs8oJ

$'=0 (28)

[in,oe-ill$'] (:}r.

This integral may be evaluated in closed form and the expressions for the modal output voltages for the e and c:p polarizations may be obtained as:

( 29)

Mn -E' (nt..}.n -jn~o[ 9 (In_l(nsm9o) )~

e - Oln 0 ~ e cos 0 (. e) ,

. 21t +In+1 n sm 0

Mn =E i (nt.. \.n+le-inq,o[(Jn+l(nsin6o) )11.

'" 0 n.O 21t fJ - J n-l (n sin (0) U

Note that the magnitude of the output depends only on the elevation angle eo. The phase of the output depends upon the azimuth CPo and also on the phase of in•o• The phase of in.o depends only upon the geometry of the spiral and the mode number and can be removed, a process that is referred to as phase compensation. The variation of the phase (after phase compensation) with the azimuth and the amplitude with the elevation angle are shown in figure 3:

~'MU,HAI+ PHADE 'P .... TTERN (AF'TER QOMPENPAr.TION)

-70 ,00 ~ ~eo

I'"HltC-E.G~eIi9J •

Figure 3a Modal Output Voltage Phase Variation with Azimuth.

-~ 0 2Q 80

THeTA (O&:IlRIiG'fI)

Figure 3.b Modal Output Voltage Magnitude Variation with Elevation.

3. ANGLE ESTIMATION USING MULTI-MODE SPIRAL

A number of angle estimation techniques are developed here. These are: 1) the "phase comparison method", 2) the Maximum Likelihood method and 3) the MUSIC approach. All these techniques are based on processing the signals at the output of a multi-channel receiver connected to the multiple outputs of a multi-ann spiral. The mode-former is no longer necessary and instead a discrete Fourier transform (OFT) of the terminal outputs is carried out to obtain the modal outputs.

3.1 The Phase Comparison Method.

An examination of the dependence of the voltages at the output of the mode-fanner on the azimuth and elevation angles of the incident radiation reveals a simple and interesting structure. The phase of the output voltage depends only upon the azimuth of the incident field and is independent of the elevation angle. Also, as shown in figure 3 the rate of change of phase with the azimuth of the incident field is proportional to the mode number. Since the phase can only be determined modulo-Zn there is ambiguity in the measured phase of every modal output voltage except that of mode 1. Best accuracy is obtained when a phase characteristic with the highest rate of change is utilized. Hence, using mode-l output to resolve the ambiguity and the phase of the output of the highest mode to estimate the azimuth, it can be shown that (for a N-arm spiral):

( 30)

$0 ;:: :n + 2: [integer part of ( nl\: Sn )]

n = 1,2,3, ... , N -1.

Here 1;. is the (compensated) phase of the mode-n output signal. It is also possible to determine the elevation angle eo by comparing the magnitudes of modal output voltages. The modal output voltage ratio is defined below for the 41- polarized case:

( 31)

I [(J p+1 (n sin (0) - J p_1 (n sin eo»)] I I [(J z (n sin (0) - J 1 (n sin 60 »] I p == 2,3,. .. , N - 1 .

Figure 4 shows the variation of these ratios for a four-arm spiral. These curves constitute a "look up table" to determine the elevation angle from the measured ratio rp-\,

, .e ',........_~.:: .. O:::OA:;::":.::"JJ~lP:..:lJT.;..:"".:;":.::O:.:.S;.:FO.:.::" E=L.EV:..:";:.:":::":..:.:" "~NQ::;;L.E=.""'.:.:""':;,:.;:.."ON::::......-.-----,

I .•

.. /~.-"-.--- .....

"

/

, ..

~' .. ~

r.

;;( •. S

s.,

// /1/

e.a --///i

Co 10 ao

.. '

..

.. ee

THETA (CEQREIiS)

70

..

Figure-4: Modal output voltage ratios.

3.2 The Super Resolution Methods:

199

Two super resolution methods, the Maximum Likelihood and the MUSIC methods are considered here. The effect of a barrage noise jammer is also taken into account.

3.2.1 Maximum Likelihood Angle Estimation:

Let the voltage output at the terminal of the pili element be:

p ::= 1,2, ... , N. , (32)

where Eo and \jI are the amplitude and phase, respectively, of the incident field arriving at the antenna from the angle (901$0)' The modal output voltage vector, M , is given by:

- r 1 2 NlI' - [ JT

M=[M.M , ... ,M J =FV=FV1,V2, ... ,VN (33)

Here F is an NxN unitary DFf matrix and:

(34)

where:

( 35)

Then, the modal output vector may be expressed as:

(36)

Note that MD and M can be computed readily using the closed form expressions derived earlier. Hence, V, too, can be computed using:

(37)

since F being an unitary matrix, pH = Fl. The observed signal is corrupted by interference, 11, consisting of receiver noise as well as a barrage noise jammer.

( 38)

200

The jamming signal is taken to be uncorrelated with both the signal and the receiver noise. The interference, t;, :is modeled by a zero mean Gaussian distributed random variable with a covariance matrix of Cll• Thus, the probability density function (p.d.f.) of the observed signal X is Gaussian with a mean of M (80, f/Jo) and a covariance

of CT\' That is,

( 39)

For a given observation at time, to, corresponding to x(> :::; ;(to) and with the quantities to be determined (Eo. ",.

eo and $0) as variable parameters the likelihood function is given by

where:

The estimates of the parameters Eo, 'V' 90 ,<Po are taken to be the values that maximize the likelihood function. The function Q may be expanded and rearranged to give:

( 42)

Q = XHC-1X - XRC-lM_ MHc-li + MHc-lM

o "0 0 T'I T'I 0 'II

= xRe-lx - a xHc-lg- - a 'g-Hc-Ix + aa 'g-Hc-lg-

o 'II 0 0 T'I 'II 0 T'I

By adding and subtracting the term,

(-H _1_)H

g en g

and after some straight forward algebraic manipulation Q may be put in the following form:

C 43)

[ -H -1- 2 )_(IX_~Hc-n_lgl-12 )

-H -1- -H -1- g Ctj Xo

:.>X" e" x" +g e" g a- -H -1-

g en g g en g

Since Q is positive, it is minimized by choosing:

( 44)

This results in:

( 45)

Since the first term is positive and does not depend upon the angle of arrival (9,41), Ql is maximized by minimizing the

second term. Thus, the estimate of the AOA (Ii, ¢ ) is determined by maximizing Q2, where:

( 46)

Note that the vector, g, is the output of the multi-mode antenna excited by a unit amplitude, plane wave incident from the arbitrary AOA, (9,«1», and is called the steering vector.

There are two distinct cases to be considered: 1) Only internal noise present, and 2) both internal noise and external jamming are present. Corresponding to these two cases the covariance matrix assumes different forms.

In the first case, the covariance matrix is given by:

(47)

where cr is the noise variance. It is assumed that the thermal noise variance is the same in all channels. In the second case, for the case of a narrowband jammer:

( 48)

Here Pj is the jammer average power and g j is the steering vector corresponding to the jammer. In the absence of jammer, C~1 is readily determined to be 1/ a", The likelihood surface, Q2ce,$) is then given by:

(49)

where,

( 50)

The maximum likelihood surface, then, has a very simple interpretation. The angIe estimate, (e,~) is the angle

corresponding to which the unit steering vector,g.i§ most parallel to the observed signal vector. xo' The Q2 surface derived above is based upon a single observation at 1;::;41' This could lead to significant variance. A better choice is to consider the surface based upon E[Q2]' That is,

( 51)

This bears an unmistakable similarity to the output of a conventional beam-former.

3.2.2 The MUSIC Approach to Angle Estimation with the Multi-mode Antenna:

The MUSIC (Multiple Signal Classification) [9] algorithm can be implemented using the outputs of the mode-former (which is similar to beam space) or the outputs of the respective elements. The approach presented here directly processes the multiple terminal outputs rather than the modal outputs.

Signal Model- Let there be k sources illuminating the multimode antenna. First consider the output due to the kth

201

source. The incident field at the origin (phase reference) is given by:

( 52)

It is assumed that the polarization is either 8- or (j)-polarized. Then, the modal output vector is given by:

( 53)

The total output at the elements due to all of the sources is given by:

The addition of white, Gaussian noise produces the total output,

( 55)

The Correlation Matrix - The correlation matrix, Rv. ,is

T

defined as:

R" =RsPR~,

.

[. . . 1

. . .

, . .

H- H- H-

R. = F t ,F :gz ···F ~K

( 60)

where R. is an (M x K) matrix: M is the number of modes (also the number of elements). Also note that the rankeR,) is K. and is assumed to be less than M. The elements of P are the average powers of the K external sources and are given by:

( 61)

The correlation matrix for the observed signal is in the standard form for the application of the MUSIC Algorithm. It may be readily shown [9] that an eigen decomposition of the correlation matrix reveals the presence of the K-signal

eigenvalues and (M-K) noise eigenvalues equal to O"~ • In addition, it may be shown that the steering vector corresponding to any of the true signals is orthogonal to the subspace spanned by the noise eigenvectors. This observation leads to the formulation of the MUSIC pseudospectrum:

MUSIC Spectrum - The eigen decomposition of the

( 56) estimated correlated matrix may be expressed as:

Under the assumption that the noise is uncorrelated with the signal, and after some manipulation, this reduces to:

The true signal correlation matrix, Rv, • is given by:

( 58)

R., =E[v.v:l=E[(~~a'ii' I~~a.g, n

If the external sources are uncorrelated, i.e.:

( 59)

some straight forward manipulation produces:

202

( 62)

The eigenvalues contained in A allow for the separation and identification of the noise and signal+noise subspaces. The matrix Q consists of the K "signal" eigenvectors and (M-K) noise eigenvectors. Let the (M-K) noise eigenvectors be grouped together in a noise matrix defined by:

( 63)

Then, the MUSIC spectrum is given by:

( 64)

1

The peaks of this spectrum correspond to the true angles of arrival. From a computational point of view it is

advantageous to determine the noise subspace from a singular value decomposition of the data matrix rather than the estimated correlation matrix.

MUSIC in the presence of Jamming Signal-

When the scene contains a barrage noise jammer in addition to the signal whose AOA is being estimated, the signal model is still given by the equation-56 but now 11 represents the total interference and includes both the thermal noise component and the jammer component. The correlation matrix of the observed signal is no longer given by the equation-58 but instead is given by:

Ry ==Rv +Cll •

T •

where CTj is the interference correlation matrix. The MUSIC approach is now modified as follows:

• The correlation matrix of the interference, CIl, is estimated from data after eliminating the signal from the composite observed signal by passing it through a notch filter tuned to the signal frequency.

• The composite signal is then passed through a whitening filter defined by:

1

V; =C~2 VT

( 66)

•

The correlation matrix of this transformed data is given by:

1 1

~- --

=C1l2 RVf C1)2

1 1 1 1

= C~2 Rv: C~2 + C~2 C1) C~2

1 1

= C~2 Rv' C~2 + I

.

( 67)

The form of the correlation matrix is now suitable for standard MUSIC approach. Let QN' be the noise subspace. Then, the MUSIC pseudospectrum is given by:

( 68)

p~SlC(e,<i»'_ 1 1 1

gH (e,<p)Fc~2 (Q~Q~H ~~2FHg(e,<p)

Note that the steering vector has been modified to 1

C~2FHg(e,<p) because of the whitening filter.

4. RESULTS

( 65)

The techniques of angle estimation using the multi-mode spiral described earlier have been implemented using MATLAB. A large number of simulations have been carried out with a view to ascertain the statistical properties of the estimators and also to assess the coverage obtainable. All the results are based on averaging over one hundred trials and for a four-arm spiral. Specification of the signal to noise ratio (SNR) is with reference to the signal received at mode-l output from signal at the bore sight.

Illustration-l :

For this case the signal AOA is given by a == 40° and cp=:600 • This direction of arrival is near a maximum of all the mode patterns. The multi-mode spiral is most sensitive to signals coming from this direction. The SNR is varied from 10dB to 40dB. The variation of the bias and the standard deviation in the azimuth and elevation is shown in figure 5. The MLM and MUSIC clearly outperform the phase comparison (PC) method even though the AOA chose for this case is best suited for the application of the PC technique.

203

Elevation Bias vs SNA

II

If

0.5 f I

I 1\

!M ! \

.. I \

~ } i

~O.3 I \

! 1/\ i I

iDO.2. \/ IV

\. \

0.1 J" \ /~~L~

\.,' '<, " ," I"·" . "

o Y '.-,. "'. ' s..l._.~ ,._ .. ~'.,./.~' ,.' ./

10 12 14 16 16 20 22 2' 26 26 30

SNR(dB)

D.S

Elevation Standard Deviation VB SNR

I'hase Oomparison MaJ<imum UkeJihood Music

9.5

~~O--~'2~~14~~'~6--~1~8--~~~~~~--~~--~~~~~~W SNR (dB)

Atlmuth Bias V5 SNR

Azimuth Standard Deviation vs SNR

Phase Comparison MaJ<lmum Likelihood Muslc

0.2

1),0

12

t8 20 ~

SNR(dB)

26

14

18

Figure-S: Statistical Properties of the Three Angle Estimators

lllustration-2:

This simulation is carried out to examine the coverage issues in more detail. For elevation angles in the vicinity of e = 40°, say in the range of 20° -650 , all the modal outputs contain signilicant signal content and accordingly all three angle estimators perform well as typified in the previous illustration. All the modal patterns except mode-l have a null at the boresight, Thus, for angles near the boresight, the

204

multi-mode antenna behaves more and more like a single antenna element leading to angle estimates that are less and less reliable. It is of interest, therefore, to determine if MUSIC or .MLM give increased coverage for elevation angles near the boresight. As may be seen from figure 6 which shows the estimator performance as a function of elevation angles near boresight, MLM and MUSIC, especially the latter, give better coverage.

E!ll\Iation Bias vs Elavation Angle

Q.2SJ-----,----;:==:!:::====~

Phase Comparison MaJ!imum LikeUhoo Music

0.2

0.05

"

\, "

10 15

Eleva.lion Angle (degrees)

Elevation S1andard Dal/la1ion vs Elevaiion Angle

O.9r------,.----;::=::::::=====:l

0.8

Pha.se Comparison Maximum Likelihood Music

0.3

...... ;. ;.; '.' ~"'I'"; ,":'. ~ - ~.,..:...

C2L_------------~--------------~------------~

5 W g ~

EI_tion Angle (degrees)

Azimlrth Blas vs Elevation Angle 2.51-~------,--~=:==~====:=:;l

Phase Comparison MaxJmum Likelihood Music

20

10 1S

Elevation Angle (degrees)

Azimuth Standard Deviation vs Elevation Angle

Phase Comparison Maximum Likelihood Music

"._ c,,;.- ~

:..: .

"- .,' - _'0' .... '':'''',.~ ....;' ..... _ ........ ,_,

\~-------------I~O--------------~15--------------~~· Elevation Angle (degrws)

Figure-6: Performance of Angle Estimators Near Boresight.

Figure 7 shows the performance of the estimators for elevation angles near horizon, 9 ~OO. All the modal patterns have a null at the horizon and the signal content is

not significant in all of the modal outputs and accordingly, the performance of the estimators suffers.

205

Elevation Bias vs Elevation Angle

Phase Comparison . Maximum Likelitlood :

Music

I '.

1~------------~------------~------------~

55 70 75 sa

Elevation Angle (degrees)

Illustration-B:

! .

',

,

/\ . !

1\ 01 ... ·, / V "

"

This simulation is carried to demonstrate that angle estimation may be carried out in the presence of a barrage noise jammer using the multi-mode antenna. While the phase comparison method fails completely in the presence of jamming, both MLM and MUSIC algorithms can be adapted to work when a jammer is present The performance of the MUSIC algorithm is demonstrated here. The signal is present at e :::: 40°, <p ::::600 and a jammer is present at e = 50° .<p =90(). The SNR is 20 dB. Three separate cases are considered corresponding to JNR" ofO,lS and 30 dB. The MUSIC pseudospectra are shown in the figure 8.

206

70 75

Elevalion Angle (degrees)

Elevation Sblndard Deviation vs Elevation Angle

Phase Comparison Maximum likelihood MUSiC

,

:Vv' .'.

,'." ..... .-;

, ..

, ,

sa

Azimuth Bias V5 Elevation Angle

O.4r------r------;==:::!::=====;l

0.35

Phase Comparison Maximum Likslihaod Music

O.S

1\

/. \

j \

i \

...

, ,

(

I

I, I

I

0.1

0.05

,

-,

QL_ ~ ~ __ -=~ ~

05 70 75 W Elevation Angle (degrees)

Azimuth Standard Deviation vs Elevation Angle

Phase Comparison Maximum Ukelihood Music

.... ~ .....

.. '"

......... ,_..-

O~~~------------~ro--------------~~------------~ro

Elevation Angle (dagrees)

Figure-7: Performance of Angle Estimators Near Horizon.

~~"" •. ·'· •• ~""'-TO

Figure-S. MUSIC Spectra for Different JNR values

It may be noted that the spectra reveal a single peak corresponding to the signal and that the jammer is suppressed. The rise in the noise floor with the increase in the jammer power may be noted.

5. CONCLUDING REMARKS

In this paper, it has been shown that a single multi-mode aperture can be used for angle of arrival estimation. The use of single multi-mode apertures offers the potential for greater versatility, reliability, and maintainability at a lower size, weight and cost than traditional interferometer techniques. Three distinct techniques for obtaining angle estimation were discussed, namely: the classical Phase Comparison Method, the Maximum Likelihood Method (MLM) and the MUSIC Method. The Phase Comparison Method is based on the Discrete Fourier Transform. The MLM and MUSIC Method are based on modem super resolution parameter estimation.

It was demonstrated that the super resolution techniques eliminate the need for a mode-former, yield lower bias and variance, and extended the coverage compared to that attainable by the Phase Comparison Method. Further, both MLM and the MUSIC method are fully capable of providing angle estimation in the presence of jamming signals. The added performance of the super resolution techniques is not, however, without cost. The MLM and the MUSIC method require the signal at all the terminals of the antenna to be received simultaneously (i.e. require a multi-channel receiver) and place a higher computational burden on the system. In addition the performance of these algorithms is sensitive to errors in the steering vectors. As a result, either measured steering vectors must be used or provision must be made to calibrate the system periodically.

ACKNOWLEDGEMENTS

The authors acknowledge the many helpful discussions with Dr. Stephen Schneider, AFRUSNRP, WPAFB, Ohio.

REFERENCES

1. E. Jacobs and E. W. Ralston, "Ambiguity Resolution in Interferometry," IEEE Trans. AES Vol. AES-17. No.6. Nov. 1981.

2. S. Haykin, Adaptive Filter Theory, Englewood Cliffs, NJ: Prentice Hall, 1991.

3. D. H. Johnson and D. E. Dudgeon, A"ay Signal Processing, Englewood Cliffs, NJ: Prentice Hall, 1993.

4. R. T. Compton, "Adaptive Maximum Likelihood Angle Estimate Bias with a Monopulse Antenna under Ideal conditions," Tech. Report 95-5-4, WPAFB,Ohio-45433.

5.

P. M. Eyring, "Compact DF Antenna Delivers High AOA accuracy," Microwaves & RF, June 1997, pp.61-68.

R.O. Corzine, J.A. Mosko, "Four-Arm Spiral Antennas", Artech House, 1990, Norwood, Ma.

E. C. Jordan and K. O. Balmain, Electromagnetic waves and Radiating systems, Englewood Cliffs, NJ: Prentice Hall, 1968.

H. Nakano, "Helical and Spiral Antennas-A Numerical Approach," Letchworth, Hertfordshire, England, 1987.

R. O. Schmidt, " Multiple Emitter Location and Signal Parameter Estimation," PROC. RADC Spectral Estimation Workshop, pp. 243-258, Oct. 1979.

6.

7.

8.

9.

Dr. Krishna M. Pasala is a Professor in the Department of Electrlcal and Computer Engineering at the University of Dayton, Dayton, Ohio. He received the B.E. degree from Andhra University in 1970 and the Ph.D. in in Electronics and Communications from the Indian Institute of Science in 1975.

Dr. Robert Penno is an Assistant Professor in the Department of Electrical and Computer Engineering at the University of Dayton, Dayton. Ohio. He received a B.S.M.E. degree from Rose Polytechnic Institute in 1971, the M.S.E.E. from Rose-Hulman Institute of Technology in 1984, and the Ph.D. from the University of Dayton in 1987.

207