Low Sidelobe Pattern Synthesis Using Iterative Fourier Techniques Coded in MATLAB Will P.M.N.

Keizer Clinckenburgh 32, 2343JH, Oegstgeest, Netherlands Tel: +31 713011249; E-mail: willkeizer@ieee.org

Abstract - The iterative Fourier technique for the synthesis of low sidelobe patterns for linear arrays with uniform element spacing is described. The method uses the property that for a linear array with uniform element spacing, an inverse Fourier transform relationship exists between the array factor and the element excitations. This property is used in an iterative way to derive the array element excitations from the prescribed array factor. A brief outline of the iterative Fourier technique for the synthesis of low sidelobe patterns for linear arrays method will be given. The effectiveness of this method to realize low sidelobe sum and difference patterns will be demonstrated for linear arrays equipped with 50 and 80 elements. This demonstration of effectivity involves also the recovery of the original low sidelobe patterns as close as possible in case of element failures. Included is a program listing of this synthesis method coded in MATLAB. With a few minor modifications/additions the included MATLAB program can also be used for the design of thinned linear arrays having a periodic element spacing [6]. Since the computational part of the included MATLAB program is coded as vector/matrix operations, this program can easily be extended for the synthesis of low sidelobe patterns of planar arrays with a periodic element spacing including pattern recovery in case of defective elements [1]-[2]. Keywords: antenna arrays, low sidelobe synthesis, FFT, MATLAB

1. INTRODUCTION Recently several papers have been published about the iterative Fourier technique (IFT) used for synthesis of low side patterns [1[-[2]. The results presented in both papers demonstrated that this method is particular suited for the large arrays with periodic spacing of the array elements and also that the IFT method is quite able to handle various design constraints related to both the radiation domain as well as the aperture domain. The design examples given in [1] showed that sum patterns and difference patterns featuring ultra low sidelobes can be combined with multiple ultra low level sector nulling. These examples covered amplitudeonly weighting and complex weighting of the aperture illumination. In [2] synthesis results based on the IFT method were presented for a large planar array corrupted by defective elements. It was shown that for each of the three monopulse patterns a very acceptable recovery of the original very low sidelobe performance was possible when the array was corrupted by a large number of defective elements. This paper deals with the synthesis of low sidelobe sum and difference patterns for periodic linear arrays using the IFT method. A description of the IFT pattern synthesis method will be given including a listing of the MATLAB code that was used to get the results presented in this paper. It will be demonstrated that using the IFT method a partial or even complete recovery of the original sidelobe performance in case the array suffers from defective elements, is feasible.

The IFT approach uses the property that for an array having a uniform spacing of the elements, an inverse Fourier transform relationship exists between the array factor (AF) and the element excitations. Because of this relationship, a direct Fourier transform performed on AF will yield the element excitations. The underlying approach relies on the repeatedly use of both types of Fourier transforms. At each iteration, the newly calculated AF is adapted to the sidelobe requirements, which then is used to derive a new set of excitation coefficients. Only those excitation coefficients belonging to the array are used to calculate a new AF. A key characteristic of this iterative synthesis method is that the algorithm itself is very simple, highly robust, and very easy to implement in software requiring only a few lines of code when programmed in MATLAB as will be shown. The computational speed is very high because the core calculations are based on direct and inverse fast Fourier transforms (FFTs). The presented results are related to linear arrays consisting of 50 elements and 80 elements located in periodic grid with half-wavelength inter-element spacing and comprise sum and difference patterns.

2. FORMULATION OF THE IFT METHOD FOR LOW SIDELOBE PATTERN SYNTHESIS The far-field F(u) of a linear array with M elements arranged along a periodic grid at distance d apart, can be written as the product of the embedded element pattern EF and the array factor AF
=
1

(1) (2)

=
=0

where Am is the complex excitation of the mth element, k is wavenumber (2/), is the wavelength, u = sin and angular coordinate measured between far-field direction and the array normal. Equation (2) forms a finite Fourier series that relates the element excitation coefficients Am of the array to its AF through a discrete inverse Fourier transform. AF is periodic in u-dimension over the interval d/. Since AF is related to the to the element excitations through a discrete inverse Fourier transform, a discrete direct Fourier transform applied on AF over the period /d will yield the element excitations Am. These Fourier transform relationships are used in an iterative way to synthesize low sidelobe pattern for arrays with a periodic element arrangement. The synthesis procedure starts with the calculation of AF using an initial set for the M element excitation coefficients. The calculation of AF is carried out with a K-point inverse FFT with K>M and using zero padding. This is followed by an adaptation of the sidelobe region of AF to the sidelobe requirements. Only the sidelobe samples of AF that exceed the sidelobe thresholds are corrected, the other samples of AF are left unchanged. After this correction, a direct K-point FFT is performed on the adapted AF to get an updated set of excitation coefficients. From those K excitation coefficients only the M samples belonging to the array are retained. When an amplitude-only synthesis has to be performed, the phases of the retained M excitation coefficients are made equal to the phases of the initial element excitations at the start of the synthesis after which a new updated AF will be calculated. This process is repeated until the new updated AF will satisfy the sidelobe requirements. Constraints for the element excitations can be quite easily incorporated in the IFT synthesis method. Dynamic range constraints for the array illumination are included by setting a lower 2

limit to the amplitude values of the updated element excitation coefficients at each iteration. The occurrence of defective elements across the aperture can be included by setting the associated excitation coefficients to zero. The inclusion of defective elements in the synthesis provides the possibility to restore the original sidelobe performance as close as possible. Both element coefficient constraints are executed near the end of each iteration just before the next iteration starts with the calculation of a new array factor. Implementation of the IFT algorithm for the synthesis of low sidelobe patterns for linear arrays using amplitude-only element weighting proceeds as follows. 1. Start the synthesis using a uniform excitation for M elements in case of the sum pattern and an odd linear taper when the synthesis involves the difference pattern. 2. Compute AF from {Am} using a K-point inverse FFT with K>M. 3. Adapt AF to the prescribed sidelobe constraints. 4. Compute {Am} for the adapted AF using a K-point direct FFT. 5. Truncate {Am} from K samples to M samples by making zero all samples outside the array. 6. Make the phase of the M samples of {Am} equal to the phase of initial excitation at Step 1. 7. Set the magnitude of the excitations violating the amplitude dynamic range constraint to the lowest permissible value. 8. Enforce the optional defective element constraint. Take element failures into account by setting their excitation values to zero. 9. Repeat Steps 2-9 until the prescribed sidelobe requirements for AF are satisfied or the allowed number of iterations is reached. The above algorithm refers to the amplitude-only low sidelobe synthesis. In order to apply phase-only synthesis the present Step 6 has to be replaced by: "Make the amplitude of the M samples of {Am} equal to one". When Step 6 is deleted, the synthesis is of the complex weight type. Any low sidelobe AF consisting of K samples can be realized with K element excitations. The objective of the IFT low sidelobe synthesis method is to arrange that the M elements of the array take completely over the contribution to AF of the excitations of the (K-M) elements located outside the aperture. This means that when the synthesis is successful when the excitations of (K-M) elements outside the aperture have become zero. The IFT method is implemented in the MATLAB program SidelobeSynthesis. This program determines for a uniform linear array the element excitation coefficients producing an array factor that matches the user defined peak sidelobe requirements. The capabilities of SidelobeSynthesis are demonstrated by the results obtained for an 80-element array and a 50element array. These results include both sum and difference patterns. Also pattern recovery in case of element failures is addressed for both types of far-field patterns. The pattern recovery examples refer to the 50-element array having 8 failed elements.

3. EXAMPLES The first example concerns a linear array consisting of 80 elements spaced 0.5 wavelength apart and characterized by an isotropic embedded element pattern. For this array a sum and difference pattern was synthesized using SidelobeSynthesis both featuring a -45 dB maximum peak sidelobe level. The dynamic range of the amplitude of element excitations, defined as 3

the ratio of its maximum value to its minimum value |Amax|/|Amin|, was not allowed to exceed 21 dB. Fig.1 shows the results of the synthesis for the sum pattern and Fig. 2 those of the difference pattern. The synthesis was carried out with 4096-point direct and inverse FFTs. Fig. 1a depicts the low sidelobe sum pattern with a maximum peak sidelobe level (SLL) of 45.02 dB. The peak levels of the sidelobes are not uniform but fall off for the far-out sidelobes. The result of Fig. 1a was obtained after 73 iterations as can be seen from Fig. 1b.The same figure shows how the maximum peak SLL decreases during the iteration process. Fig. 1c depicts the sum taper that is responsible for the pattern of Fig. 1a. In this figure one can see that the element excitations obey the dynamic range requirement of 21 dB. Fig. 1d illustrates how the number of far-field directions of the sidelobe region exceeding the SLL requirement of -45 dB, decreases with increasing iteration number. This figure shows also the widening of the main beam due to the drop in maximum peak SLL as the synthesis progresses. The number of far-field directions contained in the main lobe region, Fig. 1d, is a direct measure of its width. As can be seen from Fig. 2a the maximum peak SLL of the difference pattern is also -45.02 dB. To obtain this result, 94 iterations were required. Information about the decrease of the maximum peak SLL as function of the iteration number is shown in Fig. 2b. Fig. 2a provides also numerical information about the taper efficiency and the relative angle sensitivity Kr [3]. The relative angle sensitivity qualifies the difference slope and has a maximum value of one for the odd linear taper, which features the steepest difference slope but suffers from the highest sidelobe level. When the dynamic amplitude range of the element excitations is increased from 21 dB to 24 dB, then the low sidelobe synthesis reveals uniform -45 dB peak sidelobes for the sum pattern, see Fig. 3a, a result that was obtained after 50 iterations, Fig. 3b. The larger dynamic range for the taper has no influence on taper efficiency; it provides equal level peak sidelobes to the array factor. The 3 dB larger dynamic range for the difference taper has only a marginal effect on the difference pattern as can be noted from Fig. 4a and on the number of iterations. The second array to be considered is a 50-element linear array having a 0.5 uniform interelement spacing and an isotropic embedded element pattern for the elements. The magnitude of the taper for both the sum and difference pattern was allowed to vary over range less than 6 dB. The peak SLL requirement for the amplitude-only low sidelobe synthesis is -21 dB for both patterns. Fig. 5 shows the synthesized results for the sum pattern and Fig. 6 those for the difference pattern. The sum pattern with maxim peak SLL of -21.19 dB matches the SLL requirement of -21 dB after only three iterations. For the difference pattern 10 iterations were needed to get a maximum peak SLL of -21.26 dB. The same 50-element linear array is described in [4]. But instead of using amplitude-only synthesis, the authors applied complex weighting to get a maximum peak SLL of -21 dB. A second objective was that the width of the main beam of the synthesized pattern must be equal to that of the pattern produced by the uniform sum taper. There were no requirements for the dynamic range of taper. In [4] only the sum pattern was considered and the synthesis of this pattern required about 1000 iterations to get a maximum peak SLL of -21 dB. The synthesis method in [4], based on vector-space projections, has strong similarities with the IFT method since both methods rely on the use of forward and backward Fourier transformations between the radiation domain and the aperture domain. However the way element constraints are implemented in the method of [4] is different from that of the IFT method. Reference [4] describes the recovery of the original sidelobe performance of the sum pattern when the array is corrupted by defective elements. In [4] the elements with the indices 8, 20, 4

23, 24, 25, 26, 27, 29 were made inoperable by setting the amplitudes of their excitation coefficients to zero. Due to the occurrence of these defective elements the maximum peak SLL was degraded from -21 dB to -11dB. Pattern recovery in [4] revealed a maximum peak SLL performance of -14 dB. Fig. 7a shows the effect of the same 8 defective elements on the synthesized sum pattern of Fig. 5a. As can be seen the original maximum peak SLL has increased to -11.20 dB and the taper efficiency has decreased to 0.804. Fig. 7b shows the corresponding taper corrupted by the 8 defective elements. Fig. 8 shows the effect of the 8 defective elements on the difference pattern results of Fig. 6. The rise in maximum peak SLL due to the occurrence of defective elements is for the difference pattern smaller as for the sum pattern resulting in maximum peak SLL of -17.33 dB. Pattern recovery in case of element failures has also been performed with the IFT method using the same 8 defective elements as in [4]. Fig. 9 shows the result of the recovery for the sum pattern of the 50-element array. A better maximum peak SLL could only be obtained by allowing a larger dynamic range for the amplitude of the element excitation coefficients. Fig. 9a demonstrates that a maximum peak SLL of -16.63 dB is feasible for the sum pattern when the dynamic range of 6 dB was raised to 18 dB. This recovered sum pattern was obtained after 32 iterations using amplitude-only weighting. The achieved maximum peak SLL of -16.63 dB is more than 2 dB lower than the corresponding result of -14 dB realized for the recovered pattern in [4]. Complete recovery of the difference pattern could be achieved by raising the dynamic range from 6 dB to 8 dB. Fig. 10a shows that a maximum peak SLL of -21.29 dB is obtained after only 8 iterations. The data for the taper efficiency and relative angle sensitivity listed in Figs. 1-10 are calculated using the equations given in [5] for these parameters. The computer time to calculate the results was quite modest. The synthesis of the sum pattern of Fig. 1a required only 0.11 sec CPU time and 0.14 sec was needed for the difference pattern of Fig. 2a. The computation time of the implemented IFT method is directly proportional to the number of iterations. The computations were performed on a PC equipped with an Intel Core 2 Quad Q6600 processor running at 2.40 GHz and using 2 GB of RAM memory. The coding of the method was done in MATLAB R2008b.

4. CODING IN MATLAB Fig. 11 lists the coding in MATLAB of the program SidelobeSynthesis that is used to obtain the results of Figs. 1-10. The computational part of the program is coded as matrix or vector operations instead of for-loops to minimize computational times. The IFT method is contained in the inner function lowSLLift, Lines 61-129 of the listing in Fig. 11. Input data used by the program is contained in the Lines 5-11. Lines 5-6 specify the array, Line 7 the maximum peak SLL that may not to be exceeded and the Line 8 the dynamic range of amplitude taper defined as the ratio of maximum amplitude and the minimum amplitude of the array excitation coefficients. The program assumes an even number of radiating elements.

From Line 113 it can be noted that the adaptation of AF samples to the peak SLL requirement uses a value that is 40 dB lower than the required peak SLL. Experimentally , it was found that an adaptation with a much lower value than the requirement provides a substantial reduction of the number of iterations of the IFT synthesis and furthermore it creates also substantial deeper nulls for the sidelobes. At each iteration, the program computes the width of the main beam to determine which farfield directions of AF are contained in the sidelobe region. This computation is required since the main beam width widens as the sidelobes are reduced. Due to this widening less sidelobes samples of AF have to be checked and adapted to the sidelobe requirements (Step 3). The main beam width computation involves the angular location of the first null at both sides of the main lobe region, Lines 79-93. The indices of all AF samples located in the sidelobe region and violating the peak level sidelobe requirement, are contained in the vector indSynth, Line 100. The synthesis is successful when the number of sidelobes that exceed the sidelobe requirement becomes zero. Checking for the zero value of this number occurs at Line 109. The dynamic range requirement applied on the amplitude of the element excitations is arranged by the vector operation of Line 123. The defective element constraint responsible for the recovery of the original pattern as close as possible, involves also one vector operation, Line 125. To be become effective, this line has to be uncommented (= deleting the comment symbol %). The statements of Lines 135-149 are responsible for the computation of the taper efficiency and the relative angle sensitivity. It must be noted that these calculations are only valid when the element spacing is 0.5 wavelengths.

5. CONCLUSION With the examples presented in this paper is demonstrated that the iterative Fourier technique is ideally suited for the synthesis of low sidelobe pattern for arrays with periodic element spacing. The IFT method can deal with a wide range of constraints both in the radiation domain as well as in the aperture domain as has been proved by the results of this paper and earlier by those of [1]-[2]. The IFT method features a very high computational speed, is robust and can be quite simply implemented in MATLAB. With a few minor modifications the IFT method can also be used for the design of thinned arrays having periodic element spacing [6]. The IFT method is not a new method; it was already described more than 20 years ago in [7]. This paper came to the attention of this author near the end of the year 2007 when he was investigating the possibility to use non-uniform FFTs for low sidelobe pattern synthesis of aperiodic arrays. Reference [8] that deals with non-uniform FFTs, provides a description of the IFT method for the synthesis antenna patterns developed by C. Carroll and B. Vijaya Kumar [7]. A few years after the publication of [7] Bucci et al. [9] presented the intersection approach for array synthesis, an iterative method that uses forward and backward transformation applied to the aperture domain and aperture domain. These forward and backward transformations, which can be FFTs, are operated in the same way as the FFTs with the IFT method. The method used in [4] is based on the approach described in [9] and makes use of forward and backward FFTs.

However, even Carroll and Vijaya Kumar were not the first to have proposed the IFT method for antenna applications. The origin of the IFT method goes back to the near-field alignment of phased array antennas which technique was introduced around 1980. The way the IFT method is used for low-sidelobe synthesis is comparable to the near-field alignment of phased array antennas by which the computed far-field derived from planar a near-field measurement is transformed back to the aperture plane. This method is used in an iterative way to negate the uncertainty of the back projected field that is caused by the truncation (Gibb's phenomenon) of the far-field plane wave spectrum to the part located in visible space. This truncation is necessary to filter out the contribution of evanescent modes, located in invisible space, to the aperture field. Using the reconstructed radiating aperture phase and amplitude data from the backward transformation, the phase and amplitude settings of the T/R modules are adjusted followed by a new near-field measurement resulting in a new far-field computation. This new computed far-field is anew subjected to a backward transformation to get more precise phase and amplitude settings for the T/R modules. By repeating this process a few times the remaining phase and amplitude errors of the array elements at the end of the phased array alignment are then of the order of the residual uncertainty caused by the backward transformation of the incomplete plane wave spectrum. This type of phased array alignment makes us of inverse and direct FFTs in an identical way as is done with the IFT method. The IFT method proposed in this paper is derived from the near-field phased array alignment technique.

6. REFERENCES [1] W.P.M.N. Keizer, "Fast low sidelobe synthesis for large planar array antennas utilizing successive fast Fourier transforms of the array factor," IEEE Trans. Antennas Propagat., vol. 55, no. 3, pp. 715-722, March 2007. W.P.M.N. Keizer, "Element failure correction for a large monopulse phased array antenna with active amplitude weighting," IEEE Trans. Antennas Propagat., vol. 55, no. 8, pp. 2211-2218, August 2007. E.T. Bayliss, "Design of monopulse antenna difference patterns with low sidelobes," Bell Syst. Tech. J., pp. 623-650, May-June 1968. Y. Yang and H. Stark, "Design of self-healing arrays using vector-space projections," IEEE Trans. Antennas Propagat., vol. 49, no. 4, pp. 526-534, April 2001. D.K. Barton and H.E. Ward, "Handbook of radar measurement," Prentice-Hall Inc, 1969. W.P.M.N. Keizer, "Linear array thinning using iterative Fourier techniques," IEEE Trans. Antennas Propagat., vol. 56, no. 8, pp. 2211-2218, August 2008. C.W. Carroll and B.V.K. Vijaya Kumar, "Iterative Fourier transform phased array radar pattern synthesis," Proc. of SPIE, vol. 827, pp. 73-84, 1987. S. Bagchi and S.K. Mitra, "The nonuniform discrete Fourier transform and its applications in signal processing," Kluwer Academic Publishers, Boston, 1999, pp. 153154. O.M. Bucci, G. Franceschetti, G. Mazzarella and G. Panariello, "Intersection approach to array pattern synthesis," IEE Proc., vol. 137, Pt. H, no. 6, Dec. 1990, pp. 349-357.

[2]

[3] [4] [5] [6] [7] [8]

[9]

Fig. 1 Synthesized -45 dB SLL sum pattern realized for an 80 element linear array with dynamic range of Amax/Amin = 21 dB for the taper. The result is obtained by amplitude-only synthesis. (a) Normalized array factor. (b) Behaviour of the maximum peak SLL during the iteration process. (c) Amplitude taper responsible for the array factor of (a). (d) Number of far-field directions violating the -45 dB SLL requirement and number of far-field directions contained in the mainlobe region during the iteration process.

Fig. 2 Synthesized -45 dB SLL difference pattern realized for an 80 element linear array with dynamic range of Amax/Amin = 21 dB for the taper. The result is obtained by amplitude-only synthesis. (a) Normalized array factor. (b) Behaviour of the maximum peak SLL during the iteration process. (c) Amplitude taper responsible for the array factor of (a). (d) Number of far-field directions violating the -45 dB SLL requirement and number of far-field directions contained in the two mainlobes during the iteration process.

Fig. 3 Synthesized -45 dB SLL sum pattern realized for an 80 element linear array with dynamic range of Amax/Amin = 24 dB for the taper. The result is obtained by amplitude-only synthesis. (a) Normalized array factor. (b) Behaviour of the maximum peak SLL during the iteration process. (c) Amplitude taper responsible for the array factor of (a). (d) Number of far-field directions violating the -45 dB SLL requirement and number of far-field directions contained in the mainlobe region during the iteration process.

10

Fig. 4 Synthesized -45 dB SLL difference pattern realized for an 80 element linear array with dynamic range of Amax/Amin = 24 dB for the taper. The result is obtained by amplitude-only synthesis. (a) Normalized array factor. (b) Behaviour of the maximum peak SLL during the iteration process. (c) Amplitude taper responsible for the array factor of (a). (d) Number of far-field directions violating the -45 dB SLL requirement and number of far-field directions contained in the two mainlobes during the iteration process.

11

Fig. 5 Synthesized -21 dB SLL sum pattern realized for a 50 element linear array with dynamic range of Amax/Amin = 6 dB for the taper. The result is obtained by amplitude-only synthesis. (a) Normalized array factor. (b) Behaviour of the maximum peak SLL during the iteration process. (c) Amplitude taper responsible for the array factor of (a). (d) Number of far-field directions violating the -21 dB SLL requirement and number of far-field directions contained in the mainlobe region during the iteration process.

12

Fig. 6 Synthesized -21 dB SLL difference pattern realized for a 50 element linear array with dynamic range of Amax/Amin = 6 dB for the taper. The result is obtained by amplitude-only synthesis. (a) Normalized array factor. (b) Behaviour of the maximum peak SLL during the iteration process. (c) Amplitude taper responsible for the array factor of (a). (d)

13

Fig. 7 Sum pattern of Fig. 1 when corrupted by 8 defective elements with indices 8, 20, 23, 24, 25, 26, 27, 29. (a) Normalized array factor. (b) Sum taper with 8 defective (non-radiating) elements.

Fig. 8 Difference pattern of Fig. 2 when corrupted by 8 defective elements indices 8, 20, 23, 24, 25, 26, 27, 29. (a) Normalized array factor. (b) Difference taper with 8 defective (nonradiating) elements.

14

Fig. 9 Recovery of the sum pattern of Fig. 7 corrupted by 8 defective elements. The dynamic range of the taper Amax/Amin was raised from 6 dB to 18 dB. (a) Normalized array factor. (b) Behaviour of the maximum peak SLL during the iterative synthesis. (c) Amplitude taper responsible for the array factor of (a). (d) Number of far-field directions violating the -16.50 dB SLL requirement and number of far-field directions contained in the mainlobe region during the iteration process.

15

Fig. 10 Recovery of the difference pattern of Fig. 8 corrupted by defective 8 elements. The dynamic range of the taper Amax/Amin was increased from 6 to 8dB. (a) Normalized array factor. (b) Behaviour of the maximum peak SLL during the iteration process. (c) Amplitude taper responsible for the array factor of (a). (d) Number of far-field directions violating the -21 dB SLL requirement and number of far-field directions contained in the two mainlobes during the iteration process.

16

function SidelobeSynthesis

2
% ------------------ Start Input Parameters -------------------------------

4 6 8 10 12
% ------------------ End Input Parameters --------------------------------noEl dx specSLL ratioIllum noIter noFF = = = = 80; 0.5; -45; 21; % % % % # of array elements, only even number! Normalized element spacing Required SLL (dB) Ratio maximum/minimum value Illumination (dB)

= 130; = 4096;

% Maximum # of iterations % # of FF directions

14
u = (-noFF/2:noFF/2-1)/dx/noFF; % u-coordinates FF

16
patternType = {'Sum Pattern';'Difference Pattern'};

18 20 22
monResults(1:noIter,1:4) = 0; IllumEven IllumOdd minIllum = ones(1,noEl); = 2*(0:noEl-1)/(noEl-1) - 1; = 10^(-ratioIllum/20); % Uniform illumination % Odd linear illumination % Minimum value illumination % Container for storing % intermediate synthesis results

24
maxAF = 1;

26
for imk = 1:2

28
if imk==1

30
IllumInit = IllumEven; % Initial illumination sum pattern at start

32
indPat = 0;

34
else

36
IllumInit = IllumOdd; % Initial illumination difference pattern

38
indPat = 1;

40
end

42
IllumS = IllumInit;

44
Phase = exp(1i*(angle(IllumInit)));

46
tic

48
[Illum, AFabs] = lowSLLift;

50
toc

52
plotGraphics(Illum./max(abs(Illum)))

54
monResults(1:iCount,1:end) = 0; % Restore initial state

56
end

58
% -----Low sidelobe synthesis using iterative FTT technique ---------

60 62 64

function [Illum, AFabs]= lowSLLift for ikk = 1:noIter Illum = IllumS;

Fig. 11 Listing of the MATLAB function SidelobeSynthesis

66
AF = ifftshift(ifft(Illum,noFF)); % AF computed by FFT % Absolute value AF

68
AFabs = abs(AF);

70
maxAF = max(AFabs);

72
AFabs = AFabs/maxAF; % Normalize |AF| % Normalize AF (complex)

74 76 78

AF

= AF/maxAF;

% ------- Find all FF nulls -------------------------------------minVal = sign(diff([inf AFabs inf]));

80
indMin = find(diff(minVal+(minVal==0))==2); % Indices FF nulls

82
% -------- Find all FF peaks -------------------------------------

84
indPeaks = find(diff(minVal)<0); % Indices FF peaks

86
[peakLevel indP] = sort(AFabs(indPeaks),'descend');

88
indPeakMB = sort(indPeaks(indP(1:1+indPat)),'ascend');

90
indMax = indPeakMB(1); = find(indMin<indMax,1,'last'); % Index first null mainbeam

92
indNullL

94
% ------ Find indices all SLL directions ----------------------

96
indSLL

98 100 102 104 106 108 110

indSynth max_SLL

= [1:(indMin(indNullL)-1) ... (indMin(indNullL+1+indPat)+1):noFF]; = AFabs(indSLL)>10^(specSLL/20); = 20*log10(peakLevel(2+indPat)); % Maximum value SLL

noSLLdir = sum(indSynth); % # of SLL directions > SLL requirement monResults(ikk,1:4) = [ikk noSLLdir max_SLL ... (indMin(indNullL+1+indPat)-indMin(indNullL))]; if noSLLdir==0, return, end % ----- Adapt AF to SLL constraints -----------------------------

112
AF(indSLL(indSynth)) = 10^((specSLL-40)/20);

114
IllumS = fft(ifftshift(AF)); % Illumination derived from AF

116
% ---- Adapt Illumination to aperture domain constraints ---------

118
IllumS = abs(IllumS(1:noEl)); % Truncate # elements illumation

120
IllumS = IllumS/max(IllumS);

122
IllumS(IllumS<minIllum) = minIllum; % Dynamic range constraint % Defective elements

124
%IllumS([8,20,23,24,25,26,27,29]) = 0;

126
IllumS = IllumS.*Phase;

128
end

130

Fig. 11 Listing of the MATLAB function SidelobeSynthesis

end

132
function plotGraphics(Illum)

134
AF = 20*log10(AFabs);

136
maxAFunif = max(abs(ifftshift(ifft(IllumEven,noFF))));

138
taperEff

140
Kr = 0;

= ((maxAF/maxAFunif)^2)*sum(abs(IllumEven).^2)/... sum(abs(Illum).^2); % Difference pattern

142
if indPat ==1

144
coordX = ((0:noEl-1) - (noEl-1)/2)/(noEl-1); % Position array elements

146
% Relative angle sensitivity

148
Kr

150
end

= abs(sum(coordX.*Illum))/sqrt(sum(abs(Illum).* ... abs(Illum))*sum(coordX.*coordX));

152
fonts = 11;

154
iCount = max(monResults(:,1));

156
hFig = figure('Tag','Array');

158 160 162 164 166 168 170 172 174 176 178 180 182 184 186 188 190 192 194
aX2 = axes('position',[.07 .1 .35 .35],'Units','Normalized'); plot(20*log10(abs(Illum)),'Parent',aX2); set(aX2,'XLim',[1 noEl]) ylabel('Magnitude (dB)','Fontsize',fonts+1) xlabel('Element Index','Fontsize',fonts+1) title(['Illumination ',patternType{1+indPat}, ' ',num2str(noEl),... ' Elements Linear Array'],'Fontsize',fonts+1) aX3 = axes('position',[.53 .58 .35 .35],'Units','Normalized'); plot(monResults(1:iCount,1),monResults(1:iCount,3),'Parent',aX3); ylabel('Max. SLL (dB)','Fontsize',fonts+1) xlabel('# of Iterations','Fontsize',fonts+1) set(aX3,'XLim',[1 iCount]) aX1 = axes('position',[.07 .58 .35 .35],'Units','Normalized'); plot(u,AF,'Parent',aX1); set(aX1,'Ylim',[specSLL-20 0]) xlabel('u','Fontsize',fonts+1); ylabel('Normalized Far Field (dB)','Fontsize',fonts+1) title([patternType{1+indPat},' ', num2str(noEl), ... ' Elements Linear Array'],'Fontsize',fonts+1) txtStr{1} = {['element spacing = ',sprintf('%3.1f',dx), ' \lambda'];... 'isotropic element pattern';... ['taper efficiency = ',sprintf('%5.3f',taperEff)];... ['max SLL = ',sprintf('%6.2f',monResults(iCount,3)),' dB'];... ['SLL Req. = ',sprintf('%6.2f',specSLL),' dB']}; txtStr{2} = {['element spacing = ',sprintf('%3.1f',dx), ' \lambda'];... 'isotropic element pattern';... ['taper efficiency = ',sprintf('%5.3f',taperEff)];... ['K_{r} = ',sprintf('%5.3f',Kr)];... ['max SLL = ',sprintf('%6.2f',monResults(iCount,3)),' dB']}; text(0.66,0.84,txtStr{1+indPat},'Parent',aX1,'Units','Normalized',... 'Fontsize',fonts-1);

Fig. 11 Listing of the MATLAB function SidelobeSynthesis

196 198 200 202 204 206 208 210

title('Progress max SLL during Synthesis','Fontsize',fonts+1) text(0.66,0.90,['# of iterations = ',sprintf('%5.0f',iCount)], ... 'Parent',aX3,'Units','Normalized','Fontsize',fonts-1); axes('position',[.53 .1 .35 .35],'Parent',hFig); Hndl = plotyy(monResults(1:iCount,1),monResults(1:iCount,2), ... monResults(1:iCount,1),monResults(1:iCount,4)); ylabel(Hndl(1),'# of SL Directions violating SL Requirements', ... 'Fontsize',fonts+1) xlabel(Hndl(1),'# of Iterations','Fontsize',fonts+1) set(Hndl(1),'XLim',[1 iCount]) ylabel(Hndl(2),'# of FF Directions contained in Mainlobe', ... 'Fontsize',fonts+1) set(Hndl(2),'XLim',[1 iCount]) end end

Fig. 11 Listing of the MATLAB function SidelobeSynthesis