THE THEORY AND ARCHITECTURE OF A PLANEWAVE GENERATOR Clifton C. Courtney and Donald E. Voss. Voss Scientific, 418 Washington St., SE Albuquerque, NM 87108 Randy Haupt Utah State University, 4120 Old Main Hill Logan, UT 84322-4120 Larry LeDuc 412 TW / EWD, 30 Hoglan Ave. Edwards Air Force Base, CA 93524-8210 ABSTRACT The radiation properties of an antenna are defined in the far field, since this is the environment that they will operate. Creating far field conditions when testing a large aperture antenna is quite challenging. This is particularly true if testing occurs within the confines of an anechoic chamber, or if other complicating field characteristics (like angle-of-arrival simulation) are desired. Rather than attempt to generate a true planewave in the usual manner, we propose an instrument that creates a field distribution in the near field of a transmit array that is planewave-like in nature only over specified regions of interest (a region occupied by an antenna under test, for example); we do not require that the incident field be a true planewave at other locations. In these other locations the field is free to assume any value demanded by the governing equations of electromagnetics. By relaxing the requirement on the electromagnetic field in the test volume, we considerably reduce the complexity of the problem and define a tractable problem with a potential engineering solution. Keywords: planewave generator, far field, near field, antenna, SUT, genetic algorithm, T-factor, test fidelity 1. INTRODUCTION Planewave stimulus and the measured response of integrated aircraft electronic systems are critical components of aircraft test and certification programs. To conduct a test, a planewave environment in the volume occupied by the system under test (SUT) is typically required (see Figure 1a.). Unfortunately, for large SUTs (e.g. aircraft), the ability to generate a planewave over the entire extent of the test volume is limited, to a large extent, by the operating frequency and size of the test facility. Traditionally, to improve the quality of the incident field, the SUT is located at a Lines of constant magnitude, constant phase. (a) True plane wave illumination over the entire asset. Locally, lines of constant magnitude, constant phase. (b) Plane wave illumination over local areas of the asset. Figure 1 – True planewave and pseudo-planewave. position far from the transmitting antenna (far field), but the size of the test facility limits the available separation distance, and for frequencies below 1 GHz this separation is practically unachievable in indoor facilities for even moderately sized SUTs. This paper describes an approach to the development of a radio frequency Planewave Generator1 for the measurement of integrated aircraft electronic systems coupled to the environment through electromagnetic (EM) aperture(s). Instead of demanding that a planewave environment exist over an entire volume, we suggest that to measure an electronic systems response, it is only necessary to create an incident field that is planewave in nature only over local points and volumes on the SUT as shown in Figure 1b. In practice, the test engineer would identify the locations on an asset where the apertures of interest are located. A Planewave Generator (PWG) would then produce an electromagnetic field distribution where the incident field has planewavelike qualities only over these locations. By relaxing the requirement on the EM field over the entire test volume, we have greatly reduced the complexity of the problem. Such a PWG system could be comprised of a set (one, or more) of transmitting stations, each an array of broadband elements with individual amplitude and phase control. These stations could then be distributed about the test facility. The excitation of each element (phase and amplitude) is determined using a Genetic Algorithm (GA) 1 Patent Pending optimizing procedure. The exact aperture location(s) on the asset, and the direction and polarization of the desired planewave are input to the GA by the user. The GA then searches the large number of possible excitation vectors for a near-optimal solution, one that produces a “best-fit” field distribution with “local” planewave behavior over the SUT’s apertures. This presentation will begin with a brief description of the concept. Next, an explanation of the use of a GA will be presented for this specific application, and will include a description of the GA process and definition of the GA “fitness” criteria used to evaluate the quality of the electromagnetic field associated with a particular excitation vector. A PWG hardware architecture and typical microwave channel design is then described, and finally, a simulated result of PWG operation is given. 2. PWG CONCEPT DESCRIPTION Based on our studies we have concluded that the ability to produce planewave regions over entire large assets in confined areas typical of an anechoic chamber (even a large one) is difficult, if not impossible. This is particularly true if other more sophisticated characteristics, like angle of arrival simulation, are desired. Rather than attempt the generation of a planewave with operation in the far field of a radiator we propose a system that uses a transmit phased array, with operation in the nearfield region of the transmit array, but in the farfield of the elemental radiators of the array. Although the field at large distances from a radiator takes on propagation characteristics that are impossible to tailor, the nearfield can be fashioned to exhibit planewave properties over limited extents. The technique we describe here employs an optimizing algorithm to iteratively seek an array excitation vector that will generate planewave-like regions only over limited extents of a large asset. In other words, the effort will be to produce a field distribution with planewave amplitude, phase, and polarization relationships only over the locations occupied by, for example, distributed apertures of the SUT. The field distribution is planar and coherent only over the distributed apertures; we do not require that the incident field be planewave-like at locations not occupied by apertures of interest. The field may assume any value in other areas of the test volume and on the asset where no specific value is demanded (i.e. between two distributed apertures). By relaxing the requirement on the EM field in the entire test volume occupied by the SUT, we considerably reduce the complexity of the field the planewave generator must produce, and define a tractable problem with a potential engineering solution. But this sort of field tailoring can only occur in the nearfield of the transmitting station of the PWG. We assert that the suggested approach can produce the desired local field behavior, i.e. the resulting EM field of the PWG will be an approximation to the desired planewave field behavior over limited extents of the volume occupied by the SUT. Clearly, the departures from planewave behavior will produce some second-order effects, and these effects may limit the capability and fidelity of the PWG in some respects. The more important second order, non-ideal effects include: 1. Nearfield scattering produced by true planewave illumination of the SUT will not be reproduced correctly by the local planewave illumination. Typically, however, this scattered energy affects the response of distributed apertures on the SUT only minimally. These effects can even be accounted for by including multiple locations in the system vector optimization procedure. 2. The local planewave field will not properly produce farfield planewave scattering needed to measure radar cross section of the entire asset, but many other mission-relevant test configurations are now possible with the proposed PWG system. 3. Field components or polarizations may exist in the local planewave case, especially extremely close (in the reactive nearfield region of each element of the array distances of order λ or less) that would not exist in the true planewave environment. Here again, the relative magnitude of the unwanted components should be small with proper placement of the asset, and the effects on the response of the distributed apertures would be minimal. Even with these limitations, the proposed system will be extremely capable. The proposed “local planewave” concept will permit the in-situ test and evaluation of avionics systems with coupling apertures that occupy limited extents ( 10λ × 10λ regions and greater) on large test assets (100’s of feet in extent). The proposed system will also permit the evaluation of their response to a wide variety of EM field engagement scenarios (simulate angle of arrival to ±5o or greater), in both static and dynamic test (computer controlled simulation of fly-by) modes. In addition, due to the modular and distributed nature of the proposed PWG, we believe it will be possible to simultaneously evaluate the response of multiple aperture locations spread across the extent of the asset to coherent stimulation of a pre-described incident field. 3. GENETIC ALGORITHM OPTIMIZATION Genetic Algorithms find or uncover optimal, or near optimal, solution(s) from members of a population. In essence the GA is a search technique. In an abstract sense, a GA optimizing procedure is based on the mechanics of natural selection and genetics. They combine survival of the fittest and reproduction concepts with random but structured search methodologies to locate and identify solutions which best satisfy a given problem. While many optimization problems can be succinctly stated (mathematically in many cases), there is a class of problems for which the problem definition and identification of all parameters is impossible (the economy, for example). For these types of problems the GA technique has found much success [2, 3]. The major actions of the GA procedure are to: (1) form the initial population. This means to choose a set of possible solutions (often at random) for the initial population. Then (2) evaluate the “fitness” of each population member, which is the way in which the GA determines how well a specific member of the population produces a field that satisfies the requirements. Next (3) choose a reproduction scheme using a pair of “parents” from the population to form “offspring” or new members of the population, and implement a scheme to reduce the number of members in the current population. And finally (4) determine whether a termination condition has been reached (either a satisfactory solution has been found, or a maximum number of iterations has been reached). For the GA to operate, the characteristics of the individual members of the population (possible solutions) need to be expressed in a distinguishable way. The population for this application consists of the value of the magnitude and phase of the drive signal of each element in the PWG transmit array, and we have chosen to represent these values as a concatenated binary number. For example, we could let both the magnitude and phase of the excitation of each array element be represented as 4-bit quantities as shown in Figure 2. The complete genetic code for each possible solution then would be a collection of 8-bit strings appended to one another to form the “chromosome” of a population member. Reproduction is the term used to describe the mechanism by which the GA “evolves” the current generation into the next generation. Many types of reproduction schemes have been studied, but the one employed in this work is 4-bit Magnitude code for the 1st array element 0 1 1 0 4-bit Phase code for the nth array element 4-bit Phase code for the 1st array element 0 1 0 1 0 1 1 0 1 Figure 2. - The genetic code of an n-element array. simple. First, all members of the present population are ranked according to their fitness. A probability is then assigned to each member with the most fit members assigned a higher probability. This probability is then used to select two members of the population who will be used to create a population member of the nextgeneration. Once the two “mating” members are selected another random number, p , is generated. This random number specifies the crossover point, the location in the chromosome where the parent is split. For example, if the chromosome length is 480 bits long, and p = 0.27 , then the chromosomes of both parents will be split into two sections that are p × 480 bits and (1 − p ) × 480 bits in length (rounding to the nearest integer, and preservation of the chromosome length is maintained). The four partial chromosomes are then paired to form two next generation offspring. The next generation is filled in this way until the maximum population number is reached. In our approach, the worst half of the members of the previous generation with respect to the fitness function are then discarded, and replaced by the offspring to form the next generation. Other procedures are also used to introduce random processes in the formation of the next generation. The process repeats for the generation next until a terminating condition is reached. A terminating condition is required since the size of the solution space can be quite large. For example, if we consider an array with 16 elements, each with 15 bits of amplitude and phase, then the size of the solution space is: 216 elements ×2×15 bits = 2 480 = 3.1217 × 10144 . With a population size of 400 and for a simulation that evolved through 300 generations, the total number of possible solutions examined is < 400 ( pop size) × 300 ( gen ) = 120k which represents a ratio of solutions evaluated / total solutions = 3.84 × 10 −140 . In spite of this relatively small sampling of the solution space, the GA procedure typically will converge to a near-optimum solution. 4. T-FACTOR FIGURE OF MERIT The GA procedure must assess the fitness of each member of the population. The fitness for this application is defined as a measure of the ability of the PWG transmit array to satisfy a desired electromagnetic field distribution over predefined spatial locations. We use a modified Tfactor [1] to evaluate the fitness of a potential excitation vector candidate. The T-factor is defined as & & & E (r ) − E (r ; rref ) & & & E (r ; r ) + E (r ) & & T (r ; rref ) = 1 − ref * where E (r& ) is the radiated field of the array at the * position r& , E (r&ref ) is the radiated field at the reference position, r&ref , E (r&ref ) = ª ( kˆ × E (r&ref ) ) × kˆ º pˆ pˆ is the value ¬ ¼ { } of the reference field at the reference position, & & & & & − jk ( r − r ) is the value of E (r ; rref ) = ª( kˆ × E (rref ) ) × kˆ º pˆ pˆ e {¬ ¼ } ref the reference field translated to the evaluation position, and the quantity T ( r&; r&ref ) is the T-factor, a scalar function of vector position variables. The geometrical quantities associated with this confusing amalgam of definitions are depicted in Figure 3. The figure shows the origin (location) of the radiating elements of the PWG, the reference position, and the evaluation location (field point). The reference position is the location taken as the “perfect” planewave value since we are free to choose one position as perfect. Fitness optimization points (FOP) are defined as locations where the fitness of the radiated field is evaluated. These FOPs are typically distributed over the extent of the desired planewave regions, as shown in Figure 6, and the T-factor is computed at each FOP location. The fitness for a candidate array excitation vector is defined as the minimum T-factor value for all FOPs. k̂ = Unit vector in desired direction of propagation p̂ = Unit vector in direction of desired polarization & rref = & r= Reference position Evaluation point Planewave Polarization p̂ z Evaluation Position & r x & rref k̂ & & r − rref Planewave Reference Direction Planewave Reference Position Figure 3 – Geometry for T-factor definition. The T-factor has a number of valuable and useful features. First, it is bound by zero and one, 0 ≤ T ( r&; r&ref ) ≤ 1 , with & & T ( r ; rref ) = 1 indicating a perfect planewave-like field condition. Second, its definition can be altered slightly to consider just amplitude or just phase when computing the fitness of a potential solution. Values of the T-factor for various amplitude and phase differences are given in the tables below. Note that the values of T-factor bound the deviation for both amplitude and phase. For example, T = 0.9 implies that the amplitude and phase variation is both less than 1-dB and 15-degrees respectively. The radiated electromagnetic field of the PWG transmit array is computed in the usual way [4]. However, we have found that certain simplifying assumptions can be used to accelerate the field calculations. One of these methods is to use an omni-directional, scalar radiator model for the array elements. This technique is described in more detail in [6]. Properties of the T-factor (amplitude only) & & & & E / ERef T (r , rRef ) E / ERef T (r , rRef ) dB dB 1 0.9 0.8 0.7 0.6 0.5 0.25 0.01 0 -0.91 -1.94 -3.09 -4.43 -6.02 -12 -20 0 0.947 0.889 0.823 0.75 0.667 0.4 0.182 1 1.1 1.2 1.42 1.5 2 5 10 0 +0.828 +1.58 +3.04 +3.52 +6.02 +13.97 +20 Properties of the T-factor (phase only) 0 0.952 0.910 0.826 0.8 0.667 0.333 0.182 Δϕ & & T (r , rRef ) 0 1 5 10 15 20 25 30 1. 0.991 0.956 0.912 0.869 0.826 0.783 0.741 Δϕ & & T (r , rRef ) 45 60 75 90 135 0.617 0.5 0.391 0.292 0.076 180 0. 5. HARDWARE DESIGN A hardware design concept for a single transmit array of the PWG is shown in Figure 4 (sized for L-band operation). A standard PC controls the input RF source, and the characteristics of each RF channel of the array. No permanent reference sensors are required for feedback in this system, but feedback would improve the robustness of the PWG field. In general the PWG transmit array and SUT can be arbitrarily positioned in the chamber. The positioning information is used as input to the GA to calculate an optimal excitation vector (element drive magnitude and phase) solution for the test at hand. Once the solution is calculated, the computer converts the desired vector information into actual control settings from a calibration table residing in local memory. The calibration table is a characterization of each channel’s properties including cable effects, vector modulator properties, etc. These excitation settings then are applied to each of the sub-array’s transmitter control components to produce to the desired radiated RF signal. This system is designed to provide frequency agility and limited beam steering. Complex scenarios may be calculated in advance, then “played back” sequentially to simulate, for example, a source that is changing its relative location with respect to the SUT over time. The PWG may also utilize multiple transmit arrays in unison to accomplish a desired field distribution, or the system may operate each PWG transmit array independently. Considerable system testing and operating experience will be required to determine the optimum system configuration for a given SUT and test field requirement. Figure 4 depicts a PWG transmit array with 16 radiators arranged in a square matrix. This number and configuration of radiators was found through analysis to produce acceptable results, and the size associated with this number of radiators in the frequency band of interest (0.5 – 2.0 GHz) was considered manageable. The 500 MHz circular waveguide radiators are also shown in this figure. The design’s modular nature is indicated; reconfiguring for operation at the upper frequency bands is simply a matter of exchanging the transmit modules. A functional block diagram for each channel of the PWG is depicted in Figure 5. Each PWG transmit array will have: • • a ÷ N signal divider, where N = number of elements of the PWG transmit array; Fixed Wideband Cell Components 3dB the proper number of fixed, independent microwave channels; • removable radiating assemblies designed to operate at the center frequency of the center band; • a compact computer controller for local instrument control, calibration file storage, etc.; • a GPIB link between the user (master) computer and the PWG transmit array; and • fixed fiducial markings to permit accurate surveying of each transmit station. Interchangeable Narrowband Cell Components O.5-2.0 GHz Vector Modulator I ÷16 RF Source To Verification Circuits To Identical Circuits CPU +20dB Q Isolator Directional Coupler Low Pass Filter Open Ended Circular Waveguide Interchangable Bands: 0.5-0.8 GHz 0.8-1.3 GHz 1.3-2.0 GHz 16 bit DAC Figure 5 – PWG: channel hardware architecture. 6. PWG SIMULATION Extensive simulations of the performance of the PWG have been conducted. First, the desired planewave region (test zone) is described mathematically, and then the required planewave fidelity is specified. Specification of the desired test zone is accomplished by defining the central point of a rectangular surface (in this case) over which constant phase and amplitude is desired (planewave at normal incidence). The relationship between the transmit linear array, the desired planewave region (test zone), and the fitness optimization points is depicted in Figure 6. z kth Scalar Isotropic Radiator Figure 4 – L-band Planewave Generator concept. The key control component of each channel of the PWG is a vector modulator, it provides the needed phase and amplitude control for each channel. These modulators are standard commercial-off-the-shelf units, which can be found in typical phased array systems [5]. The nominal operating specification for the modulator is typically insufficient for the current application. Consequently a component calibration will be required. It has been shown [7] that much greater accuracy can be achieved with proper calibration. The calibration and operational verification of the PWG system will involve the comprehensive compiling of its output characteristics as a function of many parameters (frequency, power level, I and Q settings of the vector modulators, etc.). These characteristics will be assembled in calibration files stored on the system computer that the PWG will access during operation. Calibration of the vector modulators is a timeconsuming operation, and the optimal approach is still being developed. & rk′ Fitness optimization Points & r y d x Fitness optimization Points Reference Position Figure 6 – Geometry of fitness points. Next, the GA conducts an optimization to determine the array excitation vector. Finally, using the value of the optimum array excitation vector, the radiated field is computed for various spatial regions of interest, and other figures of merit are calculated. A 1-D simulation was conducted with the following parameters: 25 transmitters aligned in a linear array, frequency = 3 GHz, transmitter element-to-element separation = 4 inches, separation between transmit array and planewave region = 2.5 meters, size of planewave region = 1.5 meter in width ( or 15 wavelengths wide). The GA algorithm of the PWG simulation computed a transmit array vector (magnitude and phase for each element in the transmit array). The magnitude and phase of the radiated field were then computed over and beyond the bounds of the desired planewave regions, these values are shown in Figure 7a (field amplitude) and Figure 7b (field phase). The field T-factor, shown in Figure 7c, was also computed and is > 0.8 over the extent of the desired planewave region. Returning to the table of T-factor values, one sees this means that the magnitude of the field variation will be < 3-dB, and the phase variation < 20degrees over the planewave region. The figures below show the resulting field satisfies both conditions. 1.5 phase - degrees field magnitude 200 1.0 0.5 -200 -600 Specified Planewave Region -1000 Specified Planewave Region -1400 0.0 -15 -10 -5 0 5 10 15 20 z/l field T-factor (a) -15 -10 -5 (b) 0 5 10 15 20 1.0 8. REFERENCES 0.8 [1] Spherical Near-Field Antenna Measurements, ed. J. Hansen, Peter Peregrinus Ltd., UK, 1988. [2] R. Haupt, “An Introduction to Genetic Algorithms for Electromagnetics,” IEEE Antennas Propagat. Mag, vol. 37, pp. 7-15, Apr. 1995. [3] Practical Genetic Algorithms, R. Haupt and S. Haupt, John Wiley and Sons, Inc., NY, 1998. [4] Antenna Theory, W. Stutzman and G. Thiele, J. Wiley and Sons, Inc., 2nd edition, NY, 1997. [5] http://www.generalmicrowave.com [6] M. Timmons, C. Courtney and D. Voss, “Practical Architecture and Simulated Performance of a PseudoPlane Wave Generator,” EURO-Electromagnetics 2000, Edinburgh, Scotland, 2000. [7] “Development of a dynamic precision phase / amplitude controller: Final Report,” Herley Micro. Products, General Microwave, 425 Smith St., Farmingdale, NY 11735-1198. 0.6 Specified Planewave Region 0.4 0.2 0.0 -15 (c) Figure 8 – Future Planewave Generator Concept z/l -10 -5 0 5 10 15 20 z/l Figure 7 – PWG simulation results. 7. SUMMARY This paper has described a system approach to the generation of planewave-like EM fields over limited regions for the purpose of testing the in-situ response of integrated aircraft electronics coupled to the external environment though distributed, spatially limited, EM apertures. The approach proposes the use of a transmit phased array, with operation in the nearfield region of the transmit array, but in the farfield of the radiators of the array. It employs an optimizing algorithm to determine the transmit array’s excitation vector, and seeks to generate planewave-like regions only over limited extents of a large asset. Clearly, the departures from true planewave behavior of the transmit array’s radiated field will produce some second-order effects, and these may limit the capability of the PWG in some respects. But, the proposed concept will permit the in-situ test and evaluation of avionics systems that occupy limited extents on large SUTS, and will permit the simulation of their responses to a wide variety of EM field engagement scenarios in both static and dynamic test modes. In addition, due to the modular and distributed nature of the proposed PWG, the system should exhibit scaling properties that will permit the system to expand in concert with the desired performance parameters (size of planewave zone, maximum angle of arrival simulation, etc.). A highly integrated PWG concept is shown in Figure 8. 9. ACKNOWLEDGMENTS This work was supported in part by the United States Air Force under the Small Business (SBIR) Innovation Research program. The authors wish to thank the following individuals for their contributions to the development of this technology: Mr. Abraham Atachbarian (EAFB SBIR Program Manager), Ms. Joanne M. Eldredge (EAFB Contracting Office), Mr. Mark Timmons and Mr. Ged Bluzas (EAFB, former Project Officers), Mr. David Slemp (Agilent Technologies), and Dr. Bill Swartz, Mr. Frank Elwood, and Mr. Philip Critelli (Voss Scientific).