#### Abstract

We analyze the destructive effects of mutual coupling and spatial correlation between the separate antenna elements on a combined diversity system consisting of multiple HAPs (High-Altitude Platforms) employing various compact MIMO (Multiple-Input Multiple-Output) antenna array configurations, in order to enhance the mutual information in HAP communication links. In addition, we assess the influence of the separation angle between HAPs on system performance, and determine the optimal separation angles that maximize the total mutual information of the system for various compact MIMO antennas. Simulation results show that although the mutual information is degraded by mutual coupling and spatial correlation, the proposed HAP diversity system still provides better performance compared to a nondiversity system for all tested scenarios.

#### 1. Introduction

High-Altitude Platforms (HAPs) are quasistationary aerial platforms operating in the stratosphere. This emerging technology is preserving many of the advantages of both satellite and terrestrial systems [1–6] and presently started to attract considerable attention in Europe through the European Union COST 297 Action [1]. Using narrow bandwidth repeaters on HAPs for high speed data traffic has several advantages compared to using satellites [1], especially when operating in a local geographical area. In this paper, we propose a multiple HAP diversity system in order to enhance the mutual information in HAP communication links. To further enhance the system performance, independent spatial or polarization channels can be acquired by means of multiple antenna elements at both the transmitter and the receiver, and thus this technique is referred to as Multiple-Input Multiple-Output (MIMO) communications. The MIMO wireless communication is also combined with OFDM (Orthogonal Frequency Division Multiplexing) to form a combined HAP MIMO-OFDM diversity system.

The proposed system consists of virtual MIMO spatial channels (created by multiple HAP diversity) in conjunction with the polarization and antenna pattern diversity (formed via special compact MIMO antenna arrays), as shown in Figure 1. Two compact MIMO antenna array configurations (the MIMO-Cube and MIMO-Tetrahedron) will be investigated, and their performance, in term has of mutual information, is analysed. The enhanced mutual information that is obtained through use of these compact MIMO antenna arrays, and of multiple HAPs, will degrade if the signals of the virtual MIMO spatial channels are correlated. Correlation of the signals can be caused by either low multipath richness in the radio propagation environment or by mutual coupling between the compact antenna array elements. Therefore, the effect of mutual coupling and spatial correlation will be taken into account when performing the simulations for this combined HAP diversity system. We will also analyse the influence of the separation angle between the multiple HAPs on system performance, and determine the optimal separation angles that maximize the total mutual information of the system for the various compact MIMO antennas.

The paper is organized as follows: in Section 2 the proposed combined MIMO-HAP diversity system is introduced. The mathematical analysis of the spatial and polarization channels is presented in Section 3. In Section 4, the various compact MIMO antenna array configurations, their characteristics, and modelling are introduced. The analysis of mutual coupling and spatial correlation effects on these channels is explained in Section 5. Simulation results and comparisons are presented in Section 6. Finally the conclusions are presented in Section 7.

#### 2. The Proposed Multiple HAP Diversity System

In this paper, we propose an application for high data rate transmissions using a system employing multiple HAPs. This system consists of a number of uncorrelated MIMO channels created in the space/polarization domain using HAP diversity in combination with a special type of compact MIMO antenna arrays [7–9] as discussed in Section 4. Figure 1 shows the location diversity setup for the case of three HAPs separated by the angle *θ*.

The propagation channel from the transmitter to the receiver can be seen as a simple transformation that contains the distance dependent-free space gain of the signal being transmitted where is the distance along the path between transmitting antenna element and receiving antenna element . There are no atmospheric interferences and the noise in the system is modelled as uncorrelated Gaussian noise. The total MIMO channel can then be assembled as where and are transformations of the transmitter and receiver antenna signals into and out of the mode domain as described in Section 3.

#### 3. The Space-Polarization Channel

The polarization and radiation pattern of the electromagnetic field can be expressed as a multipole expansion [10] of the spherical wave field emanating from a virtual sphere encompassing the antenna array that is being analyzed. This expansion consists of weighted orthogonal vector base functions on the surface of the virtual sphere that allow for a solution to Maxwell’s equations in spherical coordinates which can be written as where is the intrinsic impedance of vacuum and underscoring denotes a field vector. The base functions represent mode of the electromagnetic field in spherical coordinates when the far-field of the antenna is projected onto the virtual sphere. The functions and in (3) are spherical Hankel functions of order representing an outgoing (transmitted) or incoming (received) wave. The scalar weights and are the corresponding coefficients which will give the gain of each orthogonal function (mode) for a particular electromagnetic far-field pattern, as shown by

Using (4) we can calculate which modes are active on any arbitrary antenna enveloped by the virtual sphere by knowing the electric and magnetic field patterns of the antenna. These modes are theoretically orthogonal to each other and therefore represent independent ports of the antenna. The transmitting antenna-channel transform matrix is then assumed to be the linear transformation of the input signals into the mode domain according to and for the receiving channel-antenna, we have a similar transformation matrix from the mode domain into the output signals of the system following , where and are vectors containing the mode gains for a specific antenna type.

#### 4. The Compact MIMO Antenna Array Configurations

Each transmit and receive antenna arrangement of the proposed system consists of a special compact MIMO antenna array configuration, which possesses a structure of different designs and complexities as shown in Figure 2.

**(a)**

**(b)**

The first compact MIMO antenna array is known as the MIMO-Cube [10], which consists of twelve electric dipoles positioned on the twelve edges of a cube.It can be seen from Figure 3 that this antenna array theoretically has twelve independent antenna ports.

The second compact antenna that we have investigated is known as the MIMO-Tetrahedron, which has six electric dipoles placed at the six edges of a tetrahedron (see Figure 2),and so theoretically has six independent ports as can be seen in Figure 4.

In Figure 5 we show an example of the activation of the first 30 modes calculated by (4) for the MIMO-Tetrahedron antenna array.

In order to separate the modes in Figure 5 into independent MIMO channels, we employ a singular value decomposition (SVD) operation where and are unitary eigenvector matrices of the MIMO channel modes and comprises the subchannel gains of the independent ports of the antenna array.

The orthogonalization of the channel into the independent subchannels is accomplished by multiplying the signalto be transmitted with the matrix on the transmitter side of the channel, and multiply the received signal on the receiver side of the channel with the matrix according to where is a vector containing the received and decoded signal from the independent MIMO subchannels. Inserting (5) into (6) then orthogonalizes the channels into where is the received noise. Figure 6 shows the orthogonalized subchannels for the MIMO-Cube and MIMO-Tetrahedron antenna arrays, respectively. It can be clearly seen from this figure that the MIMO-Cube antenna provides superior channel gain levels compared to the MIMO-Tetrahedron antenna. This translates to higher system mutual information as will be shown in the results section.

**(a)**

**(b)**

Through the combination MIMO-OFDM (Orthogonal Frequency Division Multiplexing), we also achieve orthogonality over the frequency domain by using the Discrete Fourier Transform (DFT) with a Cyclic Prefix (CP). Assuming a cyclic prefix CP , where and is the symbol time, we can write the received signals as a circular convolution over the DFT-frame where is the size of the DFT, is receiving antennas, is transmitting antennas, and is AWGN. If we have a MIMO antenna system with transmitting antennas and receiving antennas, we can then write the signal for subchannel in the frequency domain between any pair of transmitting and receiving antennas as where denotes each separate subcarrier. If we write (9) in vector notation, we get where is an vector and is an vector of the received and transmitted signals, respectively. is the frequency response matrix of the channel between transmitters and receivers. The noise in the system is an vector assumed to be additive white Gaussian noise. The correlation matrix of the noise vector is thereby , where is the variance of the noise and is the identity matrix. Since we are using singular value decomposition (SVD) technique, we can now write the channel frequency response matrix as where is an matrix containing singular values that are larger than zero where is the rank of the matrix , and the matrix and the matrix are unitary and contain the corresponding eigenvectors as matrix column vectors. To obtain a diagonalized system, we then define where Since the MIMO-OFDM channels in (12) are uncorrelated and the correlation of the noise is then we can write the mutual information of the system [11] as where is the variance of the separate uncorrelated input signals in vector and is the rank of the system. The mutual information in (14) is constrained by the total radiated power from the transmitting antennas, defined as

To maximize the total sum of mutual information in all the subchannels, we use the so-called “*water-filling*” technique in which we allocate more power to the subchannels with high eigenvalues. The optimal “*water-filling*” solution [12] is then given by
where *, *, is the variance of the noise, and is a predefined threshold level of the signal-to-noise ratio in the system and it is dependent on the total transmitted power in (15). The average signal-to-noise ratio in the simulations is calculated as
where is the number of active channels used for which the variance of the input signals

#### 5. Mutual Coupling and Spatial Correlation

The interaction between the elements of the compact antenna array will cause mutual interference in the radiation patterns of the array elements, and this effect is known as “mutual coupling” [13–17]. The mutual coupling between the elements of the compact antenna array can be viewed as a multiport network on the transmitter and receiver sides, respectively, in which the relationships between the terminal voltages and currents of the different ports can be define as [13], where and are the impedance matrices of the antenna array. and have been calculated numerically by using the finite element method (FEM) for the antenna arrays shown in Figure 2. An infinite approximation of the array is assumed by not taking the edge effects of the finite array into account. The terminal voltages are then associated with the source voltages through the source impedances , as described in [13], where is the source voltage at the transmitter and is the source voltage at the receiver generated by the impinging electromagnetic wave. The impedance coupling matrices are denoted as and , and are defined by where and are the impedance matrices of the transmitting and receiving compact antenna arrays, and are diagonal matrices containing the source and load impedances which have been chosen as the complex conjugate of the self-impedance of the antenna given by the diagonal and of the impedance matrices and , and and are normalizing factors [13].

In (8) it was shown that the complex envelope of the received signal at the receiving compact array after matched filtering can be expressed in vector notation as
where **s** is the vector containing the signals from the transmitting antenna elements, is the zero-mean AWGN vector, and is the normalized channel matrix that, assuming separable transmit and receive spatial correlation matrices, is modeled using the Kronecker model [18, 19],
where is an i.i.d. channel response without spatial correlation and mutual coupling. and are the separable spatial correlation matrices on the receiving and transmitting sides, respectively.

The spatial correlation matrices and are calculated between the antenna element positions and polarization states according to [20] where element of matrix or is the spatial correlation between antenna elements and . and are the spherical coordinates expressing the spatial domain and and are the polarization angle and phase difference, respectively, that account for the effects in the polarization domain [20]. In (23), is the steering vector and is the joint probability distribution function of the parameter vector , where denotes a transpose operator. It is assumed that all the parameters are independent and that and are uniformly distributed.

#### 6. Simulation Results

In this section, we investigate the mutual information improvement resulting from the use of multiple HAPs system employing compact MIMO antenna arrays and compare the results to that obtained from a system without diversity. In addition, we will also show the effects of mutual coupling and spatial-polarization correlation (calculated by the Finite Element Method) on the mutual information of the system.The mutual information results shown in Figure 7 are obtained for a system of HAPs operating at an altitude of 22 km and with a separation angle of 20 degrees, and plotted against the average signal-to-noise ratio of the system. The elevation angle is held at 10 degrees throughout the simulations.

It is evident from Figure 7 that the HAP diversity system provides superior performance as compared to the single HAP case, and the MIMO-Cube antenna provides more mutual information than the MIMO-Tetrahedron antenna which is the result of the higher number of acquired HAP platforms. Simulation results also show that although the mutual information is degraded by correlation and mutual coupling, we still achieve significant gain compared to the single HAP case. From Figure 7 it is also clear that for example, for an average SNR of 20 dB we get a 110% more mutual information from the ideal MIMO-Cube antenna than from the ideal MIMO-Tetrahedron antenna, and a 228% more mutual information compared with the nondiversity single HAP system. If we take into account the correlation and the mutual coupling, the MIMO-Cube antenna still has a 108% more mutual information than the MIMO-Tetrahedron antenna and a 195% more mutual information than the ideal nondiversity single HAP system.

In Figure 8, we show the effect of separation angles between HAPs on the mutual information of the two compact MIMO antenna arrays. The separation angle (see Figure 2) is varied from 0 to 35 degrees in steps of 5 degrees (due to computational limitations) and plotted for an average SNR of 20 dB. It is clear from this figure that the separation angle between HAPs has a great impact on the mutual information of the system. It is also evident that the optimal separation angle that maximizes the total mutual information of the system at this SNR is found to be approximately 10 degrees for the MIMO-Tetrahedron antenna and approximately 15 degrees for the MIMO-Cube antenna. The degradation of mutual information performance at wider separation angles is due to the increase in distance between the HAPs and the ground stations. Figure 8 also suggests that the vector element antenna array is superior to the MIMO-Tetrahedron array at wide separation angles even though they access the same number of independent subchannels.

#### 7. Conclusions

In this paper, we have investigated the degradation in mutual information, due to spatial correlation and mutual coupling, when using different compact MIMO antenna arrays in a multiple HAP system employing multiple spatial-polarization modes. Simulation results show that the multiple HAP diversity systems utilizing compact MIMO antenna arrays provide significant mutual information enhancement, even though it can also be noticed that small degradation in mutual information is resulted due to the effects of spatial correlation and mutual coupling between the separate antenna elements of the MIMO-Cube and MIMO-Tetrahedron arrays. It was also shown that the performance of the compact MIMO-Cube antenna is superior to the MIMO-Tetrahedron antenna since it has twice the number of independent channels which will result in more system mutual information. Finally, we have also shown the effects of the separation angle between HAPs on system performance, and determined that the optimal separation angle that maximizes the total mutual information of the system is within a narrow sector of approximately 20 degrees for the compact MIMO antenna arrays investigated in this paper.