An ANN-Based Model and Design of Single-Feed Cross-Slot Loaded Compact Circularly Polarized Microstrip Antenna

Transcription

1 An ANN-Based Model and Design of Single-Feed Cross-Slot Loaded Compact Circularly Polarized Microstrip Antenna Rakesh K. Maurya 1, Binod K. Kanaujia 2, A. K. Gautam 3, S. Chatterji 4, Sachin Kumar 5 1 Department of Electronics and Instrumentation Engineering, Faculty of Engineering and Technology, MJP Rohilkhand University, Uttar Pradesh-2436, India 2,5 Department of Electronics & Communication Engineering, Ambedkar Institute of Advanced Communication Technologies & Research, Delhi-1131, India 3 Department of EC Engineering, G.B. Pant Engineering College, Uttrakhand , India 4 Department of Electrical Engineering, NITTTR, Chandigarh-1619, India ABSTRACT An artificial neural network (ANN) based model and design is proposed for single-feed compact sized circularly polarized microstrip antenna (CPMA). The training data sets for ANN-based model are collected from Ansoft HFSS v.14 for varying the dimensions of slots and slits with the size of antenna at different relative permittivity and loss tangent with best axial ratio. To achieve an accurate ANN model, Levenberg-Marquardt (LM) algorithm and three hidden layered MLP network is trained. The design antenna consists of eight slits introduced at the boundary and corners of the radiating square patch with a cross-slot in the centre. The proposed antenna shows a compactness of 42% from conventional circularly polarised antenna with 3-dB axial ratio bandwidth of 1.9%. At last, the ANN model is validated by comparing its result with the electromagnetic simulation and measurement, which shows good agreement among them. The design model may be a good choice for small antenna requirement with circular polarization for indoor WLAN application and ANN model is enormously useful to design engineers for CAD models. Keywords: Artificial neural network, circularly polarized, compact microstrip antenna, synthesis model 1. INTRODUCTION Microstrip antenna are used for several communication applications due to their numerous advantages like simplicity, low manufacturing cost, light weight, low profile, reliable and ease in fabrication [1-2]. Currently, circularly polarized microstrip antennas are used in RFID systems, GPS, WLAN applications and mobile satellite communication due to their mobility and freedom in the orientation angle between a transmitter and receiver in comparison with a linearly polarized microstrip antenna. Normally, in CPMA dual-feed and single-feed structures are used. Dual-feed gives a wide axial ratio, but suffers with complicated structure and increase in antenna size, however single-feed CPMA has advantage of simple structure and compact size [3-6]. Nowadays, personal mobile communication and modern transceiver systems need convenient compact antenna to satisfy the severe constraints on the physical dimensions of the portable gadgets. A number of compact CPMA structures operating at fixed frequency have been reported in the literatures [7-16]. These includes circular and square patch loaded with slits and fed along with 45 at one of the slits [7-9], corner chopped square MSA with four slits [1-11], square ring MSA with truncated corners [13], truncated square patch with group of four bent slots [14] and diagonally asymmetric slotted MSA [16]. In recent few years artificial neural network based CAD models have been applied for analysis and synthesis of MSAs in various forms such as circular, rectangular, triangular patch antennas [17-2]. ANN based CAD model are much popular due to their accuracy in results, ability and adaptability to learn, least information requirement, quick real time response and simple implementation feature. A neural network model can be developed by learning through training. Training process minimize the training error between the target output and the output of ANN. This trained model can be used any time during such types of antenna design. In this paper, an ANN based model and design is proposed for single-feed compact sized circularly polarized microstrip antenna. The compactness is achieved by loading a cross-slot in the centre of the single-feed CPMA [8]. Further, compactness is achieved by introducing slits at the corners of the patch maintaining the circular polarization. The Volume 4, Issue 8, August 216 Page 1

2 proposed antenna with different design specification are synthesized, simulated and measured to validate the model. The ANN model shows good agreement with the simulated and experimental results. 2. ANTENNA STRUCTURE AND DESIGN Figure 1 shows the basic structure of single-feed circularly polarized square microstrip patch antenna and is named as antenna 1 [8]. Figure 2 shows the proposed structure of the cross-slot loaded circularly polarized microstrip antenna and is named antenna 2. Finally, proposed antenna structure with four slits inserted at the corners of the square patch is shown in Figure 3 and is known antenna 3. Figure 1 Geometry of single-feed circularly polarized square microstrip antenna (Antenna 1) Figure 1 Geometry of proposed cross slot loaded square microstrip antenna (Antenna 2) The antenna is fed at the diagonal of the patch using 5Ω coaxial probe. Finally proposed antenna 2 and antenna 3 are fabricated and printed on a 1.6 mm thick FR-4 epoxy substrate of relative permittivity 4.4 and loss tangent.12. The size of the ground plane for both the antenna is 46 X 46 mm 2. The detailed optimized dimensions are listed in Table 1 and Figure 4 shows the photograph of the fabricated patch. Figure 2 Geometry of proposed compact circularly polarized antenna with slits at the corner (Antenna 3) Volume 4, Issue 8, August 216 Page 2

3 Table 1. Optimized dimensions of the proposed structure Antenna Parameters mm Size of square patch, L*L 3*3 Length of slit, u 1 Length of slit, v 9.2 Width of slit, w 1 Length of cross slot 1 Width of cross slot 1 Length of corner slit along diagonal PR 12.5 Length of corner slit along diagonal QS 1.5 Width of corner slits 1 3. DEVELOPMENT OF ANN BASED MODEL Artificial neural network is one of the most popular and intelligent tool for synthesis of these types of problem. ANNs have gained attention as fast and flexible tool for antenna modeling having non-linear response, simulation and optimization [21-22]. In the course of developing an ANN model, the architecture of neural network and the learning algorithm are two most important factors. Earlier a number of structure and architectures have been used by many researchers [17-21]. The selection of structure and architecture for a particular model implementation depends on the problem to be solved by head and trial basis. After several experiments using different architecture coupled with different training algorithms, the Multilayer Perceptron (MLP) neural network architecture and Levenberg-Marquardt (LM) algorithm is found suitable to train the MLPs for obtaining high precision synthesis models [23-24]. An MLP consist of three layers: an input layer, one or more hidden layers and an output layer. The accuracy of ANN-based model depends on data sets used in training and testing. The data sets are collected from simulated structure on Ansoft HFSS simulator, for different cross-slot length/width and corner-slit length/width, relative permittivity and loss tangents. The data collection of proposed antenna 2 and antenna 3 for optimized dimensions shows the variation of S 11 with frequency for different cross-slot length and cross-slot width [25]. It is observed that the resonance frequency of antenna 2 decreases with increase in length and width of cross-slot. Hence, the resonance frequency of antenna 2 depends upon the dimensions of cross-slot. The variation of S 11 with frequency for antenna 3 at different length and width of corner-slits is shown [25]. Again it is found that the resonance frequency of antenna 3 is decreases with increase in length and width of corner-slits for entire band of operation in circular polarization region. Finally by these observations the optimized dimension for proposed antenna is found and is listed in Table 1 and collected data sets are used for training and testing of ANN model. Thus out of 4 data sets collected, 35 were used for training and rest were used to test the model. After several trials, by changing the epoch and neurons, it was found that the 5 layers MLP network (input layer, output layer and three hidden layer) achieved the task with high accuracy. The best network configuration for model is shown in Table 2 and Figure 5 represents an ANN-based synthesis model. Volume 4, Issue 8, August 216 Page 3

4 Table 2. Optimized Network Parameters of ANN Model for proposed structures ANN Parameters Value Network architecture MLP Model Network algorithm Levenberg-Marquardt Hidden layers 3 Neurons of hidden layers 5, 26 and 7 Activation function of hidden layers Tangent Sigmoid Activation function of input and output layer Linear Momentum coefficient.9 Learning rate.1 Training time and epoch with 2.5 GHz Intel(R) Core i5-421 second, 3935 epoch 321M CPU 4. RESULTS AND DISCUSSION In this paper ANN based model has been successfully introduced and designed the proposed antenna. The electrical measurement of design antenna is done on Agilent Network Analyzer of PNA L-Series. The compactness in proposed antenna 2 achieved is ~27% at resonance frequency 1.75 GHz and ~42% in antenna 3 at resonance frequency 1.7 GHz in comparison to conventional square patch antenna [26]. The comparison among the results of ANN-based model, simulation and measured for variation in S 11 with frequency is shown in Figure 6 and Figure 7. In the figures it can be observed that the resonance frequency is shifted towards its lower value because the path of excited patch surface current is increased due to the integration of slots and slits. It can also be seen that the ANN model shows good agreement with the HFSS simulation and experimental results. -5 S11 (db) Antenna2 (Simulated) Antenna2 (Measured) Antenna2 (ANN) Figure 6 ANN, simulated and measured reflection coefficient (S 11, db) versus frequency (GHz) of antenna 2 and antenna 1 Volume 4, Issue 8, August 216 Page 4

5 -5 Antenna3 (Measured) Antenna3 (Simulated) Antenna3 (ANN) -1 S11 (db) Figure 7 ANN, simulated and measured reflection coefficient (S 11, db) versus frequency (GHz) of antenna 3 and antenna 1 The axial ratio vs. frequency results for the same is shown in Figure 8 and Figure 9, and is been observed that both the proposed antennas (antenna 2 and antenna 3) have axial ratio below 3-dB, hence, circularly polarized. It can also be seen that a better axial ratio bandwidth is observed in antenna 3 as compared to antenna 1 and antenna 2. Here, also ANN model shows a good agreement with simulation and measured results. Axial Ratio (db) Antenna2 (Simulated) Antenna2 (Measured) Antenna2 (ANN) Figure 8 ANN, simulated and measured axial ratio (db) versus frequency (GHz) of antenna 2 and antenna 1 Axial Ratio (db) Antenna3 (Simulated) Antenna3 (Measured) Antenna3 (ANN) Figure 9 ANN, simulated and measured axial ratio (db) versus frequency (GHz) of antenna 3 and antenna 1 Volume 4, Issue 8, August 216 Page 5

6 Figure 1 shows the total radiated power of the designed antenna 3 in both H-plane and E-plane. The sense of rotation can be changed by simply changing the position of feed along another diagonal of the square patch. It is observed that total radiated power remains same with change in frequency. The antennas show stable radiation characteristics in both plane, and more than -3 db power is achieved within angle of 17 in E-Plane and 13 in H-Plane. Both the antennas have almost same radiation characteristics. 5. CONCLUSION The ANN-based modelling, design and fabrication have been done successfully for compact circularly polarized microstrip antennas. The proposed antenna is ~42% compact in size in comparison with conventional circularly polarized microstrip antennas. So the proposed compact antenna may be a best choice for indoor WLAN applications for circular polarization, mitigable to multipath fading in the field of wireless communication in addition to ANNbased model which would be a useful tool for the design engineers. References [1] J.R. James and P.S. Hall (Eds.), Handbook of microstrip antennas, Peter Peregrinus, London, UK, [2] Y.T. Lo, D. Solomon and W.F. Richards, Theory and experiment on microstrip antennas, IEEE Trans Antenna Propag 27 (1979), [3] S.L. Ma and J.S. Row, Design of single-feed dual frequency patch antenna for GPS and WLAN applications, IEEE Trans Antenna Propag 59 (211), [4] Z.B. Wang, S.J. Fang, S.Q. Fu and S.W. Lu, Dual-band probe-fed stacked patch antenna for GNSS applications, IEEE Antenna Wireless Propag Lett 8 (29), [5] C.H. Weng, H.W. Liu, C.H. Cu and C.F. Yang, Dual circular polarisation microstrip array antenna for WLAN/WiMAX applications, Electron Lett 46 (21), [6] Z.G. Wang, S.J. Fang, Q. Wang and H. Liu, An ANN-based synthesis model for the single-feed circularlypolarized square microstrip antenna with truncated corner, IEEE Trans Antenna Propag 6 (212), [7] K.L. Wong, W.H. Hsu and C.K. Wu, Single-feed circularly polarized antenna with a slit, Microwave Opt Technol Lett 18 (1998), [8] K.L. Wong and J.Y. Wu, Single-feed small circularly polarised square microstrip antenna, Electron Lett 33 (1997), [9] K.L. Wong and M.H. Chen, Single-feed small circular microstrip antenna with circular polarization, Microwave Opt Technol Lett 18 (1998), [1] W.S. Chen, C.K. Wu and K.L. Wong, Novel compact circularly polarized square microstrip antenna, IEEE Trans Antenna Propag 49 (21), [11] A.K. Gautam, P. Benjwal and B.K. Kanaujia, A compact square microstrip antenna for circular polarization, Microwave Opt Technol Lett 54 (212), [12] J.Y. Wu, C.Y. Huang and K.L. Wong, Compact broadband circularly polarized square microstrip antenna, Microwave Opt Technol Lett 21 (1999), Volume 4, Issue 8, August 216 Page 6

7 [13] W.S. Chen, C.K. Wu and K.L. Wong, Single-feed square-ring microstrip antenna with truncated corners for compact circular polarization operation, Electron Lett 34 (1998), [14] W.S. Chen, C.K. Wu, and K.L. Wong, Compact circularly polarized microstrip antenna with bent slots, Electron Lett 34 (1998), [15] W.S. Chen, C.K. Wu and K.L. Wong, Compact circularly polarized circular microstrip antenna with a cross slot and peripheral cuts, Electron Lett 34 (1998), [16] Nasimuddin, Z.N. Chen, and X. Qing, Compact circularly polarized asymmetric-slotted microstrip patch antennas, Microwave Opt Technol Lett 54 (212), [17] K. Guney and N. Sarikaya, A hybrid method based on combining artificial neural network and fuzzy inference system for simultaneous computation of resonant frequency of rectangular, circular and triangular microstrip antenna, IEEE Trans Antenna Propag 55 (27), [18] N. Turker, F. Gunes and T. Yildirim, Artificial neural design of microstrip antenns, Turk J Electr Eng Comput Sci 14 (26), [19] R.K. Mishra and A. Patnaik, Neural network-based CAD model for the design of square-patch antennas, IEEE Trans Antenna Propag 46 (1998), [2] R. K. Mishra and A. Patnaik, Design of circular microstrip antenna using neural network, IETE J Res 44 (1998), [21] Y. Kim, S. Keely, J. Ghosh and H. Ling, Application of artificial neural network to broadband antenna design based on a parametric frequency model, IEEE Trans Antenna Propag 55 (27), [22] M. Kara, The resonant frequency of rectangular microstrip antenna element with various substrate thickness, Microwave Opt Technol Lett 11 (1996), [23] E.D. Uubeyli and I. Guler, Multilayer perceptron neural network to compute quasistatic parameter of asymmetric coplanar waveguides, Neurocomputing 62 (24), [24] M.T. Hagan and M.B. Menhaj, Training feed-forward networks with the Marquardt algorithm, IEEE Trans Neural Network 5 (1994), [25] S. Kumar, B.K. Kanaujia, A. Sharma, M.K. Khandelwal and A.K. Gautam, Single-feed cross-slot loaded compact circularly polarized microstrip antenna for indoor WLAN applications, Microwave Opt Technol Lett 56 (214), [26] P.C. Sharma and K.C. Gupta, Analysis and optimized design of single feed circularly polarized microstrip antennas, IEEE Trans Antenna Propag 31 (1983), Volume 4, Issue 8, August 216 Page 7