Micro-algorithme génétique couplé à la méthode TLM pour la synthèse de formes d antennes non-intuitive

Transcription

1 Micro-algorithme génétique couplé à la méthode TLM pour la synthèse de formes d antennes non-intuitive TLM method coupled with a Genetic Algorithm for non-intuitive synthesis of antenna shapes Robert STARAJ, Jean-Lou DUBARD GDR Ondes _ Journée Thématique Méthodes d optimisation de formes d antennes _ ENST _ Lundi 10 Juillet 2006

2 Motivation Objective Tools Requirements of new wireless systems multi-standards : WLAN, HiperLAN, GSM, DCS, UMTS. small size ( Lengths < λ/4 ) Low manufacture cost Antennas design Difficulty to perform intuitive design of small for multi-band operations

3 Motivation Objective Tools Introduction Coupling TLM-GA Antennas design Conclusion E θ E φ Amplitude (db) WAVE LENGTHS TOWARD GENERATOR ---> 10 <--- WAV ELENGTHS TOWARD LOAD θ (Deg) Specific characteristics + constraints : small size low manufacture cost

4 Motivation Objective Tools Introduction Coupling TLM-GA Antennas design Conclusion E θ E φ Amplitude (db) WA VE LENGTHS TOWARD GE NERATOR ---> 10 <--- WAV ELENGTHS TOWA RD LOAD θ (Deg) Specific characteristics + constraints : small size low manufacture cost Unintuitive design process Efficient with original shapes

5 Motivation Objective Tools 3D-TLM electromagnetic solver Modelisation of with arbitrary shapes and materials PML layers for efficient free space simulation Time-Domain simulation + Prony-Pisarenko method Wide frequency band analysis Parallel computing Genetic Algorithms optimizer Global maxima Optimal solution Multidimensional & Multimodal function search domain

6 Motivation Objective Computational resources CINES Tools IBM-SP 528 processors SGI 512 processors 592 Go memory 256 Go memory 1 Tflop/s 512 Gflop/s IDRIS IBM-SP 256 processors 832 Go memory 1,33 Tflop/s

7 fundamental GA key points Optimization process Antenna parameters Binary coding Chromosome µga Actual performance N FB fitness = min i = 1 1 NFB Stochastic operations Fitness function N i S11 n i n = 1 ( ) db i N i Desired performance setting S11 ( n i ) to 10dB db when S11 ( n i ) 10dB db (Selection, reproduction, crossover and mutation) Population size (From 100 to individuals)

8 fundamental Optimization process µga µga : the process restarts when prematurely convergence occurs ( difference in bit < 5%)

9 fundamental Optimization process µga performances Optimization of the coaxial probe position for a rectangular patch antenna (1024 combinations) 1.00 µga Fonction coût normalisée Normalized fitness function AG_pop5 GA_pop5 AG_pop64 GA_pop64 AG_pop32 GA_pop32 µga_pop5 microag_pop Nombre de simulations Number of TLM simulations

10 Broadband optimization of a rectangular patch Original antenna S 11 db 0dB -10dB -20dB Frequency (GHz) BP = 2.6% at VSWR 2

11 Binary coding of the patch The patch is divided into N square pixels Metallic Pixel Nonmetallic Pixel 1 0 Antenna N bits chromosome unlimited shapes of antenna Slow convergence

12 Broadband optimization of a rectangular patch Original antenna S 11 db 0dB -10dB -20dB Optimized antenna Frequency (GHz) BP = 2.6% at VSWR 2 BP = 5.2% at VSWR TLM simulations CPU time=220 mn 12 processors (IBM SP3)

13 Dual-band optimization of a rectangular patch Original antenna S 11 db 0dB -10dB -20dB Optimized antenna Frequency (GHz) Operating at 4.5GHz Operating at 4GHz and 5GHz 400 TLM simulations CPU time=220 mn 12 processors (IBM SP3)

14 Design of a microstrip antenna for ISM bands Desired frequency band at VSWR 2 WLAN 2.45GHz±50MHz HIPERLAN2 indoor 5.25GHz±100MHz HIPERLAN2 outdoor 5.6GHz±127.5MHz Parameters to be optimized 3 x Binary coding Antenna 100 bits chromosome

15 Design of a microstrip antenna for ISM bands L c W c W 1c L s L 1c W 2c L 2c W 3c L 3c measurement TLM 2000 TLM simulations 12 processors 28 hours CPU time S11 (db) Fréquence Frequency (GHz)

16 Design of a microstrip antenna for ISM bands co-polarization and cross-polarization radiation pattern 0 E plane 2.45 GHz 0 H plane Pattern Amplitude (db) Pattern Amplitude (db) Theta (degrés) Theta (degrees) Theta (degrés) Theta (degrees) Measured ( ) and TLM simulation (---) 0 6-III E planee plane 0 H plane Amplitude Pattern (db) Pattern (db) Amplitude (db) Theta (degrés) Theta (degrees) 5.25 GHz Theta (degrés) Theta (degrees)

17 Optimization of a Planar Inverted-F Antenna for UMTS/WLAN application Goal >>> UMTS/WLAN (1.92 : 2.5GHz) Optimized parameters : Position of the coaxial probe Position & dimension of the vertical short-circuit Number & length of slots inserted in the upper plate

18 Optimization of a Planar Inverted-F Antenna for UMTS/WLAN application L w c W 2 L 3 L 1 l L c 2 h W 1 W Dimension in mm : L=W=35 L =27,5 L =6,3 L =20 W =21, W =7,5 w =9,4 l =4,4 h=9 2 c c Slot width : 1,25mm Ground plane : 100x100mm CPU time: 18h / 12 processors

19 Optimization of a Planar Inverted-F Antenna for UMTS/WLAN application 0-5 S11 (db) Measurement Optimization Frequency (GHz)

20 Optimization of a Planar Inverted-F Antenna for GSM/DCS application L1 L2 Side view 0 feed W3 L3 SS W Wm W1 W2 S11 (db) Shorting wall Top view Lm Lm=100mm, Wm=33.75mm, W=28.75mm, L1=10.625mm, L2=24.375mm, L3=22mm, W1=14.375mm, W2=11.25mm, W3=5mm, S=1.25mm Frequency (GHz) 440 TLM simulations CPU time: 10h / 12 processors

21 Optimization of a Planar Inverted-F Antenna for GSM/DCS/WLAN application 0 (db) اS11 ا Frequency (GHz)

22 Efficient automatic process for nonintuitive antenna design New multi-standards with original shapes CPU time consuming Future works Modelization of arbitrary oriented narrow slots with 1 cell Use of diakoptics