This is a text version of the presentation given on the above paper at the International Radar Symposium, Munich September 1998. The full reference which expands on some aspects of the performance of existing radars is
P R Foster and D J Browning, `Dependence of Ground Clutter upon Airborne Radar Performance', IRS 1998, Munich, p167-176
The purpose of this paper is to examine the problems of modelling the performance of an airborne radar antenna, that is, an antenna installed on an aircraft for an AEW radar and subsequently to examine the effect of any changes in the antenna performance on the ground clutter present in the radar system. Clutter is defined as any unwanted radar echo. In the case of an airborne radar, most clutter effects arise from reflections from the earth's surface, hence the term `ground clutter'.
The characteristics of ground clutter in range-velocity space are well-known [1] . Away from the regions obscured by main beam clutter and by altitude line effects (specular reflection from the earth's surface immediately below the radar antenna), the magnitude of the contributions to the radar return due to ground clutter is determined primarily by the antenna sidelobe performance.
Table 1 shows typical RF parameters for an airborne radar antenna of high and low gain. An aperture efficiency of 50% has been assumed. In addition to these RF parameters, low sidelobes are required which should be lower in Azimuth than Elevation.
| Antenna Type | Peak Gain (dBi) | Beamwidth (degrees) | Aperture diameter (l) |
|---|---|---|---|
| High Gain | 37.0 | 2.0 | >30 |
| Low Gain | 30.0 | 5.0 | >13 |
These dimensions are difficult to achieve with the radar antenna installed within the fuselage. Antennas sited above and below the fuselage have been used [2]. Such a position subjects the antenna to far more degradation from the aircraft structure than a nose-mounted antenna. A detailed examination of installed antenna performance needs to take into account the interaction between the antenna and the aircraft structure itself. Moreover, these interactions will vary with beam position, whether the beam is scanned electronically or mechanically and also with frequency.
The modelling techniques, which can be used to compute the radiation characteristics of antennas installed on structures, are determined primarily by the dimensions of the structure relative to the operating wavelength. For structures which are small in comparison to the wavelength, the Method of Moments technique (MoM) may be used [3]; however, this becomes computationally impractical for structures greater than a few wavelengths. For structures with dimensions in excess of 3 to 5 wavelengths, diffraction techniques may be employed [4]. Diffraction techniques have the advantage that they are based on ray tracing and runtimes are independent of the size of the structure in wavelengths. However, the underlying theory of diffraction assumes that a well-formed wavefront from the antenna strikes the diffracting object and this is not the case with an antenna having a gain in excess of 30 dBi and illuminating an aircraft structure. The radiating aperture must be modelled in terms of a set of sub-apertures, each of which is of sufficiently low gain to produce a well-formed wavefront striking the aircraft.
The aperture of the high gain antenna must be sub-divided in small sub-apertures whose effective gain is low enough to meet the ray tracing criteria. Each sub-aperture is then treated separately for ray tracing and the results integrated. For an array antenna, the subdivision is straightforward in that either individual elements or subarrays containing groups of elements can be used as the smallest unit in the diffraction calculations. The choice of subarray must be made with care since the element radiation patterns are added assuming a mean position for the phase centre of the subarray. It is possible to introduce severe and erroneous grating lobes by incorrect subdivision of the antenna aperture. Each antenna must be dealt with separately and few generalisations can be made; on the whole, an antenna aperture must be broken down into elements which are less than about 0.65 wavelengths apart in order to avoid these spurious grating lobes. Division into aperture elements is more difficult for reflectors since not only must the radiating aperture be included but also the spillover of the feed system past the reflector or subreflector and any additional internal diffraction within the reflector system must be taken into account.
If the radar antenna is divided into N sub-apertures or elements, the total runtime will be N times that for a single element. In addition, the half-power beamwidth for a single element will be quite wide (of the order of 30 to 60 degrees) and the angular resolution of the calculation need not be smaller than 2 to 5 degrees. However, since the results for the AEW antenna have to be summed and as the final radiation pattern has a beamwidth of a few degrees only, the angular resolution for every element in such a calculation has to be much greater and the time required to compute the installed antenna performance becomes proportional to N * N. This dependence results in prohibitive computation times for the large airborne radar antennas currently used for AEW applications.
For the calculation of the effects of ground clutter, the antenna far-field radiation characteristics are required at angular intervals which allow accurate interpolation between points. There must therefore be at least 10 points per 3dB beamwidth around the main beam and the angular zone covered must be large enough to include sidelobes down to the level of interest which we will set as -50 dB.
For a given antenna, radar system parameters and operational scenario, the effects of ground clutter upon airborne radar performance may be modelled by calculating and summing the contributions from the earth's surface, as a function of range, Doppler frequency shift, antenna radiation pattern, ground clutter reflectivity et cetera, in order to obtain the distribution of clutter in range-velocity space. The dependence of the ground clutter returns on antenna performance may then be assessed in terms of obscuration analysis [5], that is, the proportion of range-velocity space obscured as a function of clutter power level above receiver noise. The computer program CASPAR [7] calculates the ground clutter in range-velocity space using a computationally fast technique. For monostatic radar geometries and pulse-Doppler radar systems, the computation time required to calculate the obscuration in range-velocity space for a single main beam look angle and a single set of radar parameters varies typically from 3 to 30 minutes depending primarily upon the amount of the earth's surface which must be considered. This in turn depends upon the platform altitude, the pulse width and the PRF.
This example consists of an S-band array antenna utilising 16 and 4 vertically polarised rectangular waveguide elements in the azimuth and elevation planes respectively, a total of 64 elements. The antenna has a modified Taylor amplitude taper [8] applied to provide a first sidelobe level in azimuth of -35 dB and -20 dB in elevation. The azimuth and elevation plane half-power beamwidths are 7.5 and 23.0 degrees respectively. The array is mounted as a dorsal fin on a large aircraft with broadside direction normal to the aircraft fuselage (Figure 1). The modelled aircraft has an ellipsoidal fuselage of total length of 36.5 m and vertical and horizontal semi-axes of 1.95 m. The radiation patterns were computed in free space (Figure 2) and then installed on the aircraft (Figure 3). The array elements were phased to give the main beam an elevation of -2.0 degrees.
The azimuth sidelobes are hardly affected but the elevation pattern is considerably degraded. The nulls are filled in and there is a major area or sidelobe disruption in the lower hemisphere.
Calculations have been performed for a typical medium PRF monostatic pulse-Doppler radar operating at S-Band, using the calculated free-space and installed azimuth and elevation plane patterns and with the following additional primary parameters:
Figure 4 shows the clutter range-velocity contour plot in free space where there is a sharp `spike' at a range of 7 km. This is the altitude line which occurs when specular reflection from immediately below the antenna occurs. The bulge to the right between 8 and 16 km in range is due to the elevation sidelobes. Figure 5 shows the installed performance. The altitude line has been removed and the zone between 8 and 16 km is essentially unchanged.
This example consists of an S-Band array antenna utilising 40 and 20 horizontally polarised rectangular waveguide elements in the azimuth and elevation planes respectively, a total of 800 elements. The antenna has modified Taylor amplitude tapers applied to provide first sidelobe levels of -35 and -25 dB in the azimuth and elevation planes respectively with half-power beamwidths of 3 (azimuth) and 5.5 degrees (elevation). This array is also mounted as a dorsal fin on the medium-sized aircraft (see Figure 1) with the broadside direction normal to the aircraft fuselage.
Figure 6 shows the free space radiation pattern over an angular zone of +/- 45.0 degrees about the broadside position. Figure 7 shows the installed pattern over the same angular zone. In this case, the runtime was over 110 minutes on a Pentium Pro 266 MHz. In order to speed up the production of results, the problem was sub-divided and run on several machines simultaneously. Ray tracing is ideal for this process as the results for each angle are independent.
The system assumptions are as for example 1 above. Figure 8 shows the clutter results for the antenna in free space which is dominated by the main beam. Figure 9 shows the ground clutter results for the installed antenna and there is a significant spike at the 1 and 2 dB level.
The larger of the two examples with 800 elements is quite small compared to some existing AEW radars which have over 4000 elements. We have shown that, although technically feasible, in general, the detailed modelling of the installed performance of high gain antennas suitable for AEW radar operations involves prohibitive computation times for currently available computer hardware. If the calculations are restricted to the principal planes or to limited regions about the main beam, the computation times become tractable. However the aperture must be divided into a sufficiently small number of sub-apertures. For the aircraft fuselage dimensions considered and for the types of array antenna considered, the difference between free space and installed antenna performance is significant and these differences affect the clutter levels. However, the significance of the effects depends upon the particular antenna being modelled and the combination of radar parameters selected. A full characterisation of the dependence of clutter upon antenna performance requires knowledge of the full antenna radiation pattern characteristics as a function of azimuth and elevation look angles and antenna position upon the aircraft. Given sufficient computational resources, the combination of modelling techniques discussed in this paper could be used to provide quantitative indications of the dependence of clutter upon installed antenna performance.
An improvement factor in computation times of between 3 and 10 could be achieved by utilising a high-performance work-station or a super-computer. However, a possible alternative would be to employ a parallel processor architecture. Separate processors could be employed to calculate the radiation patterns of separate sub-apertures. Alternatively, the computation task could be divided into separate angular regions and these sub-tasks could then be deployed on separate processors as was done for the example of 800 elements.
In the work described above, the computer programs, ALDAS [6] and CASPAR [7], were used.
