Intermodulation in Active Array Receive Antennas Klaus Solbach, Universität Duisburg, Hochfrequenztechnik, 47048 Duisburg, Tel. 00-79-86, Fax -498, Email: hft@uni-duisburg.de and Markus Böck, Antenna Technology ASE 7,European Aeronautic Defence and Space Company 89070 Ulm, Tel. 07-9-5650, Fax -580 E-Mail: markus.boeck@sysde.eads.net I. Introduction Large signals incident to the RF amplifier of a Radar receiver have been known to produce several deliterious effects, like compression of wanted signals, production of crossmodulations and harmonics and most problematic, the production of intermodulation signals. Intermodulation responses ( third-order-products ), in particular, can appear as false targets in the Radar signal processor, and thus degrade the Radar performance considerably. Such large signals may come from several sources, among them ground clutter, large targets, near-in targets and interfering radar systems and jammers. Modern solid state Radar systems are more prone to the appearance of large-signal effects from clutter returns etc., due to the use of relatively long pulses ( with overlapping of pulses from, e.g., clutter and targets) compared to high-peak power/ low-duty cycle and short-pulse tube systems. With the advent of active array technology in Radar antennas the additional problem appears of a multitude of RF amplifiers distributed across the array as compared with a single amplifier behind a passive antenna, Fig.. Several questions arise from this situation: What are the effects of a low gain and wide beamwidth element pattern compared to the high gain and narrow beam antenna pattern after beam forming in conventional array systems and what are the effects of the beam forming network behind the amplifiers on the generated spurious signals? At a fist glance, it seems obvious that the conventional system performs better due to the spatial filtering afforded by the antenna directivity pattern in front of the receiver and because of coherent combination of spurious signals in the combiner network of the active array. In the literature, active array versus passive array system aspects have been limited to Gain, ERP and G/T-figures, e.g. //, and large-signal intermodulation effects have been inspected for transmit active array antennas //. In //, the problem of spectral spreading of clutter responses due to third order intermodulation effects is discussed for active array Radar systems, however, assumptions concerning the beam former signal combining are over-pessimistic and no comparison is made between conventional and active array system. II. Array Model The investigation of above questions starts from a simple model of a two-element array, with distributed amplification (active array), Fig.. Two waves are assumed incident to the antenna from different directions and at different frequencies, producing phase increments of Φ and Θ of the received signals due to the spacing d of the elements. The amplifiers nonlinearity of the output / input relation is represented as a power series, which is truncated after the third order term. The beam forming network is assumed to perform ideal power combination without frequency dependence. A description both by algebraic equation is used and by the resulting antenna directivity pattern, which relates the angles of incidence (and thus the phase increments between the element signals) to the sum signal level at the output port.
II. Array Directivity Considering the plane wave signals incident to the two antenna elements, it is found that the two signals are combined in the beam forming network corresponding to their angle of incidence, i.e. corresponding to the phase increment between the elements; this means that the array directivity pattern may introduce different gains in the directions of the signals, as is well known from array theory. This result is no longer true for the third-order-products, the intermodulation signals, which are generated in the distributed amplifiers in front of the beam forming network: Here, the third order means the third power of the sum of the two signals at each amplifier with the well known creation of new frequencies (ω ω ) and (ω ω ). However, the complete arguments incorporate also the phase increments of the constituent signals, so that new increments Φ Θ and Θ Φ are created. Thus, the third-order sum signals after the beam forming network are attenuated as if received from totally different directions than the original signal directions. In other words, the beam forming network provides power combining also of the intermodulation signals, yet with different gain than effective for the received signals, depending on angles of incidence. II. Intermodulation Ratio The second major result concerns the resulting linear signal and intermodulation signal strengths for the worst case of broadside incidence for both signals ( α = β = 0 ): All signals behind the amplifiers are combined with the broadside directivity factor, since all phase increments are zero (two equal signals combine with a factor of ). Assuming identical amplifiers and equal amplitudes of both incident waves, A=B, it can be shown that the ratio of intermodulation signal amplitude and original signal amplitude is independent of the number of elements a A a A IM (active) = =. () 4a 4a A The passive -element array with a single amplifier for the sum-signal (increased by a factor of ) would produce a ( A) a A IM (passive) = =, () 4a A 4a while for an array of N-elements the factor of is replaced by the array directivity D=N, so that: IM (active) / IM (passive) = N = (0logN) db () It is clear from these results that the active array has an advantage over the passive array system with respect to the third-order intermodulation ratio, which improves as N, the number of elements / array directivity, e.g. by 0 db for a 0-element array or 0 db for a 000- element array. Note that the same improvement could only be achieved by inserting an attenuator of power loss equal to N in front of the amplifier in the passive array system, e.g. a 0 db-attenuator for a 0-element array! Keeping in mind that the noise level in both active and passive array variants will be the same, e.g. // and /4/, the advantage of the active array architecture vs. passive array architecture extends equally to the spurious free dynamic range (SFDR). III. Experiment
Both above results were demonstrated in an experiment using power dividers/combiners to simulate radiator elements and feeding signals from two RF sources to simulate incident waves. In order to afford a good match and tracking of all components and to allow the losses in the cables and in the power dividers/combiners to be negligible, the experiment was carried out at a frequency of GHz. Wilkinson power dividers were used for power division, combining and beam forming and amplifiers based on the MiniCircuits series GAL were used. Fig.(a) shows the setup representing a passive -element array with two waves incident from broadside and a single amplifier following the beam former-combiner. Amplitudes of linear and intermodulation components were measured using a spectrum analyzer. The active array representation is shown in Fig.(b), where the sequence of beam former and amplifier is reversed with respect to Fig.(a) and we use an extra amplifier which is adjusted to track the intermodulation level of the first amplifier to an accuracy better than db. Comparing the measured signal levels of both setups we find the IM-ratio of the active array model superior by about 6.5 db. Since eq.() expected 6 db improvement, this result provides satisfactory proof, considering the nonideal tracking of the two amplifiers. A third setup, Fig.(c), was used to test the dependence of the intermodulation products w.r.t. the angle of incidence of the waves. In the measurement setup, the signal of frequency f is distributed to the two antenna ports in-phase ( Φ =0), while the second signal of frequency f is split using one variable length cable, producing a phase shift Θ between the two antenna ports for this signal. From the observation of the two signals at the output of the active array model we find that the level of the first signal (f ) does not vary with Θ while the second signal (f) is attenuated as a function of its phase shift. In Fig. 4, bold symbols are entered for measured relative amplitudes of this signal as a function of phase shifts of 90, -45, 45, and 90, with 0 as a reference. It is seen that the attenuation is in approximate agreement with the beam former response (antenna pattern) which can be described by the cos( Θ /)- function. In each case the two intermodulation products at frequencies below and above the two principal carriers are also observed. Normalized levels of these products are entered into the diagram above their respective phase angles of Θ and - Θ. Again, the measured data fit approximately the beam former response; this result clearly demonstrates that the intermodulation products in an active array behave like signals incident from different angles compared to the actual signals. IV. Conclusion As a consequence of both above mentioned results, for most practical situations encountered in Radar system operation the active array antenna will not degrade the large signal handling of the system but, rather, it allows the use of amplifier circuits of inferior linearity, which can save power consumption and circuit complexity / cost. Literature // Kraft, U.R., Gain and G/T of Multielement Receive Antennas with Active Beamforming Networks, IEEE Trans. AP, vol.48, no., Dec.000, 88-89 // W.A. Sandrin, Spatial Distribution of Intermodulation Products in Active Phased Array Antennas, IEEE Trans. AP, Nov. 97, 864-867 // K.M.Harrington, Active Array Radar Nonlinearity Requirements-Spectral Analysis of Third Order Intermodulation Clutter, 996 IEEE Intern.Symp. on Phased Array Systems and Technology,5-8 October 996, Boston, Massachusetts, ISBN 0-780--6, -7 /4/ Solbach, K., Noise Signal Decorrelation in Broad-Band Active Array Systems, Frequenz, Band 55, Heft -, 00, 7-
Radiator Array Beam Forming Network Amplifier Fig. Principle concept of (a) passive array with single amplifier after beam forming and (b) active array with receive-amplification before beam forming Re{} = A cos( ω t) a α β Re{} = B cos( ω t) b Uniform Plane Wave Re d jφ jθ { a e + b e } Re{ a + b} = x = x Isotropic Radiator Elements Amplifier Phase Increments: π Φ = d sin α λ π Θ = d sin β λ y = a + a x + a x + a x 0 y y Antenna Pattern D D max D α D β 0 α -90 0 90 β = ( y + y ) s Fig. Two-element antenna concept of an active array (a)
(b) (c) Fig. Measurement setup to model -element array (a) passive array with broadside incident waves (b) active array with broadside incident waves (c) active array with one wave incident from broadside and one tilted incident wave Fig. 4 Measured response of linear and intermodulation signals in active array model with two incident signals as a function of phase shift Θ