Personal tools

Resource Scaling for Antenna Placement Modeling on a SAAB JAS-39 Gripen Aircraft

This white paper demonstrates how resource requirements scale for the modelling of a modern fighter aircraft when the frequency increases. It also demonstrates how difference simulation methods may be applied and how they scale relative to each other.

A model of a SAAB JAS-39 Gripen aircraft is used to demonstrate the modelling of coupling between two L-band blade antennas on an aircraft, one on top and behind the cockpit, the other below the nose cone on the front of the aircraft. This same model is then used to demonstrate how computational resource requirements scale for antenna placement problems as the solution frequency increases.  As frequency increases full wave solutions (e.g. MLFMM) will run into memory constrains and this work also demonstrates that in such cases high frequency asymptotic methods in FEKO (e.g. PO) are well suited to continue simulation work.

Antenna Coupling Simulation

Antenna placement and coupling problems are typically characterised by S-parameter computations. S11 demonstrates input impedance of the antenna and may be used to investigate whether the antenna detunes when mounted on a platform, as opposed to operation in free space.  S21 is used to demonstrate the attenuation between transmit and receive antennas on the same platform.  In the following case such results clearly indicate good impedance matching for the antenna in the frequency band of interest, while coupling between the two antenna positions was very small (S21 < -50dB).

Figure 1: L-band blade antenna mounting positions for coupling simulation


Figure 2: L-band antenna matching and coupling parameters on Gripen aircraft


Computational Resource Scaling for Antenna Placement Studies of Increasing Frequency

The Gripen model was simulated with a hypothetical antenna located roughly in the middle of the spine of the aircraft.  Frequency of operation (and simulation) was increased to demonstrate how resource requirements scale for the MLFMM and PO modelling methods.  The same simulation platform was used for all simulations:

  • 2x Quad core Intel Xeon E5606 CPU, 2.13 GHz (i.e. 8 processes in total)
  • 72 GByte RAM

All simulations were run with multi-threaded parallel processing on all 8 available cores.  This setup ensured efficient memory management while running parallel simulations and maximal use of the available CPU cores.

The following figure demonstrates how electrical size increase for doubling simulation frequency.  The instantaneous current distributions on the airframe demonstrate how the aircraft doubles in size in wavelength terms and thus becomes a significantly bigger EM problem each time that simulation frequency is doubled.

Figure 3: Current distribution on the airframe at different simulation frequencies

300 MHz600 MHz1.2 GHz
POSTFEKO_SAAB_EDS_Grippen_model_300MHz.png POSTFEKO_SAAB_EDS_Grippen_model_600MHz.png POSTFEKO_SAAB_EDS_Grippen_model_1G2Hz.png
Click on individual images for full-size version

The following graphs demonstrates numerically how solution complexity increases with doubling of frequency.  It also demonstrates that in cases where the MLFMM requires very large amounts of memory and CPU-time the PO requires very little time and memory to solve the same problem.  Comparisons of the radiation patterns for MLFMM and PO at 1.2 GHz and 2.0 GHz demonstrate that both methods give good results.  PO does not predict radiation patterns as accurately as MLFMM in the shadow region of the problem (below the airframe) and users of the PO should keep this effect in mind when solving antenna placement problems.

Figure 4: Scaling of memory for increasing simulation frequency

All memory requirements scaled relative to MLFMM 300 MHz

Memory scaling.png

Figure 5: Scaling of run-time for increasing simulation frequency

All run-time requirements scaled relative to MLFMM 300 MHz

Run-time scaling.png

Figure 6: Comparison of far field radiation patterns computed with MLFMM and PO

1.2 GHz2 GHz