A Dual-Band Patch Antenna Employing a Folded Probe Feed for Nonlinear Radar Applications

In this work, we design a novel, integrated transmit (TX)/receive (RX) dual-band folded-probe-fed patch (DFPFP) antenna to improve nonlinear radar (NLR) system performance. Designed for down-looking, airborne NLR applications, the DFPFP antenna occupies an improved form factor ( $0.9\lambda \times 0.9\lambda \times 0.1\lambda $ , 1.8 kg) over existing commercial-off-the-shelf (COTS) designs and additionally offers minimal drag in airborne applications due to its low profile. The DFPFP antenna is simulated, fabricated, and tested, and excellent gain (9 dBi) and bandwidth (15%) performance are demonstrated. Iterations of the design from single, standalone TX and RX antennas to the final, integrated design are presented and discussed. The final DFPFP is compared to a pair of COTS antennas that are currently used for NLR system testing. Evaluated in the context of an NLR system, the custom DFPFP antenna significantly outperforms the COTS antennas, with signal-to-noise ratio (SNR) improvements as high as 17 dB, while offering significantly improved size, weight, and power (SWaP) performance.


I. INTRODUCTION
I N a conventional radar system, a target is illuminated with a signal at a given frequency f 0 .Depending on the properties of the target, the transmitted wave is scattered and reflected back toward the radar receiver.Observation of this reflected signal allows for the detection and ranging of various targets such as weather formations, asteroids, and aircraft [1]- [3].More recently, the concept of nonlinear radar (NLR), or nonlinear junction detection, has been explored.Many man-made devices contain nonlinear sources in the form of semiconductors, bi-metal junctions, transmission lines, and a number of other potential sources [4]- [8].When illuminated with a radar signal, these devices accordingly reflect signals both at the fundamental frequency, f 0 , as well as at its harmonic frequencies, 2f 0 , 3f 0 , etc. [9].
By exploiting this phenomenon of nonlinearity, it is possible to detect the presence of certain man-made electronic devices while rejecting the clutter associated with the presence of linear scatterers [10].While NLR presents promising possibilities for expanding tactical warfighting capabilities, there remain several challenges associated with its use, particularly in the area of antenna design.One such challenge is the diminished power level associated with the harmonic response of a target, which results in a significantly lowered SNR compared to conventional radar systems.In order to mitigate this issue, relatively high-gain antennas are generally required to deliver significant amounts of power to the target.Due to their flexibility, wideband antennas are often used in practical NLR systems.However, in the presence of strong interferers, use of these wideband antennas may result in a degraded signal-to-interference-plus-noise ratio (SINR), as compared to a more narrowband antenna.NLR antennas must exhibit good linearity, as antennas themselves can introduce nonlinearities into a radar system due to the presence of bi-metal junctions, connectors, or transmission lines [11], [12].Lastly, NLR antennas must exhibit good SWaP performance, as they must be practical for field operation.
The concept of NLR has been demonstrated in several prior works.Previous NLR research has typically focused on one of two application areas: detection of electronics, or insect/animal tracking.In insect/animal tracking, NLR is often selected for use over conventional radar due to the high clutter rejection associated with its use [10].In [13], the authors designed a NLR system to track the flight of the Asian yellowlegged hornet.In the paper, the authors designed a nonlinear target consisting of a Schottky diode in the center of two copper wires oriented in the shape of loops.This target was attached to the hornets, which were then tracked using the NLR system.The authors used a 1.8 m-long commercial-offthe-shelf (COTS) slotted waveguide antenna for transmitting, and a 0.51 m-long microstrip patch array for receiving.At the chosen frequencies, these dimensions correspond to 56λ and 31λ, respectively.A similar concept was demonstrated in [14] with improved range.In the experiments in [14], the authors made use of a large COTS dish antenna for both transmit and receive.The authors in [15] similarly demonstrated the use of NLR for tracking of small frogs.In the experiments in [15], the authors used large COTS horn antennas.Unlike insect/animal tracking, in which targets are fabricated to produce a strong nonlinear response, NLR can also be used for the detection of extant electronic devices.In [16], the authors demonstrated the use of NLR to detect two different COTS two-way radios.In the work, the authors make use of multitone harmonic radar, in which two signals at different frequencies are transmitted toward a target, and both the 2f 0 harmonic as well as the intermodulation products of the two transmitted signals are observed.By using this method, they demonstrated the ability to differentiate between the two different electronic devices by analyzing the characteristics of the received nonlinear signals.
In their experiments, the authors used ETS Lindgren 3164-03 horn antennas for both transmitting and receiving, which measure 0.51 × 0.33 × 0.33 m 3 , or approximately 1.3λ × 0.8λ × 0.8λ at their TX center frequency and 2.6λ × 1.6λ × 1.6λ at their RX center frequency, and weigh 9.1 kg each.In [17], a synthetic aperture radar (SAR) using large COTS horn antennas is developed for NLR which allows for detailed target imaging.In all of the preceding works, large COTS antennas were used for system testing.While some smaller, custom antenna designs utilizing circular polarization were explored in [18] and [19], the reduced gain of these antennas is generally detrimental to the overall performance of an NLR system.In order to enable practical field use of NLR, it is critical to improve the SWaP of the antennas used in these systems, while maintaining strong NLR performance.Army Research Lab's RF Signal Processing and Modeling Branch has recently explored the use of more portable COTS designs such as the RFSPACE LPDA-MAX and RFSPACE TSA-600 [20], [21].However, the SWaP improvements offered by these COTS antennas come at the cost of degraded NLR system performance.
In this work, we design two custom antennas to meet the design requirements of an NLR system, while maintaining small size, low weight, and high power handling.More specifically, our antennas are designed to achieve a detection range comparable to that of [16] using approximately the same transmit power, with a form factor comparable to those of the COTS RFSPACE antennas.Each of our custom antennas is designed to cover a range of frequencies for its corresponding application: 600-700 MHz for transmit and 1200-1400 MHz for receive (approximately 15% bandwidth).At their respective operating frequencies, the dimensions of both antennas are approximately 0.9λ × 0.9λ × 0.1λ.The operating frequencies were selected as a result of previous works which suggested that the strongest response from nonlinear devices of interest typically falls somewhere within this range of transmit frequencies [10].The antennas were designed to have a gain similar to the larger COTS antennas used in previous works, on the order of 10 dBi.Because there exists a fundamental trade-off between the gain of an antenna and its relative size [22], [23], and because NLR has been demonstrated effectively using antennas with a gain of approximately 10 dBi, this value was selected as the target goal for the custom antennas.Lastly, in the design of these antennas, special consideration is given to preservation of linearity.Accordingly, designs with the inclusion of bi-metal junctions or external matching circuits are avoided, and the inclusion of transmission lines and connectors is minimized.Overall, these design efforts directly support the improved portability of NLR systems, with the possibility that a future system could not only be carried by a person, but piloted by an unmanned aerial vehicle (UAV).
In Section II, we discuss the design, fabrication, and testing of the custom FPFP antennas.Gain and S 11 measurements are provided for both the custom antennas as well as the COTS antennas.In Section III, both sets of antennas are tested using the same NLR setup.A surrogate target is constructed, and system SNR levels are provided for varying power levels and target distances.In Section IV, we consider a design that combines the two FPFP into a single, integrated, dualband folded-probe-fed patch (DFPFP) design.The design methodology is detailed, and the measurement results of the fabricated prototype are provided.In Section V, the DFPFP is tested in the same NLR system described in Section III, and results are provided.Finally, Section VI provides some final comments on the presented work.

II. ANTENNA DESIGN I
To meet the requirements described in Section I, two folded-probe-fed patch (FPFP) antennas, one each for transmit and receive, were designed.A fully annotated model of the FPFP design is presented in Fig. 1, while all corresponding parameter values for both the transmit and the receive antennas are presented in Table I.The design of these antennas was inspired by the design described in [24].This design was selected as a starting point for our nonlinear radar antennas due to its excellent gain, around 10.0 dBi, and bandwidth, around 25%, relative to its form factor, about 1.7λ × 1.7λ × 0.1λ.Additionally, this design is capable of matching a wide range of impedance values without the need for an external matching network, simply by varying the length and height of its feed arms.This is a critical consideration for a nonlinear radar antenna, as inclusion of external matching networks should largely be avoided due to the likely generation of harmonics [18].In order to tailor the antenna for nonlinear radar applications, several changes were made to the original design.It was determined that to minimize potential antennagenerated nonlinearities, our antenna should be constructed solely of copper.Accordingly, we opted to use a solid copper ground plane, as opposed to the printed circuit board (PCB) ground plane used in the original design.By making this substitution, one of the primary power-handling limitations was removed, as breakdown of the dielectric substrate of the PCB would be the primary mechanism of power limitation at high power levels [25].This change also necessitated a change to the feed mechanism.In the original design, an SMA edge connector was used to feed a microstrip line, which fed two L-shaped probes that were soldered into vias.By contrast, our design utilizes a direct SMA feed, which feeds the folded copper arms that excite the patch.This feed mechanism, which utilizes 0.8 mm-thick solid copper sheets, is more robust than the microstrip lines of the original design.PCB traces are typically on the order of tens of micrometers in thickness [26].By contrast, the relatively thick feed arms of the FPFP provide much greater power handling capability.Lastly, we re-evaluated the required size of the antenna ground plane.In this work, SWaP is a critical consideration, and the primary contributor to the size and weight of the FPFP antenna is its ground plane.Accordingly, the effect of procedurally reducing the size of the ground plane was analyzed.Reducing the size of the ground plane primarily has two effects.Initially, as the size is reduced, the antenna continues to behave as expected of a patch antenna, albeit with increasingly reduced bandwidth.As the size of the reduced ground plane approaches the size of the patch, the gain, bandwidth, and polarization purity of the  antenna all begin to drop off sharply, and the behavior of the antenna can no longer be accurately predicted based on the expected behavior of a patch antenna.Based on this analysis, the size of our ground plane was set around 0.9λ, as shown in Fig. 1.At this size, the antenna still behaves similarly to the original design, with a reduced bandwidth of about 15%.Because our application requires only approximately 15% bandwidth with a preference for high out-of-band rejection (to reject would-be interferers), this empirical analysis allows us to reduce the occupied area of the antenna ground plane by approximately 72% while improving the performance of the antenna for its intended application.
In order to set the antenna input impedance, the parameters a, b, and d, representing the height and length of the feed arms, and the offset of the feed from the edge of the patch, respectively, are iterated over to modify the capacitance, inductance, and resistance of the port.These parameters do not strongly impact the gain or other features of the antenna, and are used only to set the input impedance.Note from Table I that some parameter values are common between the TX and RX antenna (in terms of the wavelength), while others are not.In general, most of the values shown in the table are fixed relative to the wavelength, and are based on previous probe-fed and U-slot patch antenna designs [24], [27]- [29].In the case of W and L, the minor discrepancy between the two sets of values is a result of manufacturing tolerances, as these components were ordered cut to size by the manufacturer.The h f value is the same for both antennas as a result of the dimensions of the panel-mount SMA socket that was selected for fabrication.Lastly, observe that the values for a, b, and d vary significantly between the two antennas, as these dimensions are used only to modulate the input impedance of each antenna and do not necessarily scale with frequency.
Each of the antennas was fabricated using 0.8 mm-thick (1/32") copper sheets.The copper for the ground planes and patch of each antenna was ordered cut to size.By contrast, the feed arms for each antenna were cut to size from a larger sheet of copper using metal shears.The arms were then folded into shape, and a hole was drilled in the center of the arms to accept an SMA feed.Holes were drilled into the ground plane of each antenna in the appropriate location to allow for the SMA feed.This feed was mechanically fixed in place using machine screws.Dielectric screws were used in order to prevent the creation of a bi-metal junction, as would be the case if metal screws were used.The folded feed arms were soldered to the protruding end of the SMA connector to securely attach them in place.Although this solder joint does result in the creation of a bi-metal junction, the antenna was tested in this configuration to verify that its behavior remained linear for the range of power levels that we planned to test with.In future designs, this solder joint could be easily omitted given more refined fabrication techniques that allow for use of a secure press fit.Styrofoam was cut to size and used to suspend each patch above its ground plane, and to support the ends of each feed arm.The fabricated transmit and receive antennas are shown in Fig. 2(b) and Fig. 3(b), respectively.As fabricated, the transmit antenna measures 0.4 m × 0.4 m × 0.05 m (about 0.9λ × 0.9λ × 0.1λ at 650 MHz) and weighs roughly 1.8 kg, while the receive antenna measures 0.2 m × 0.2 m × 0.03 m (about 0.9λ × 0.9λ × 0.1λ at 1300 MHz) and weighs roughly 0.5 kg.To provide a point of comparison for the FPFP antennas, two antennas that are currently used by Army Research Lab for NLR system testing were also evaluated: the RFSPACE LPDA-MAX for transmitting and the RFSPACE TSA600 for receiving.These antennas are shown in Fig. 2(a) and Fig. 3(a), respectively.The LPDA-MAX measures 0.5 m × 0.4 m (1.1λ × 0.9λ at 650 MHz), and weighs approximately 0.5 kg, while the TSA600 measures 0.33 m × 0.24 m (1.4λ × 1.0λ at 1300 MHz) and weighs approximately 0.2 kg.The measured VSWR of the four antennas is presented in Fig. 2(c) and Fig. 3(c), while the measured gain of each is presented in Figs. 4 and 5.In each case, each antenna is oriented so that its direction of maximum radiation is oriented along the z-axis, with its E-plane in the φ = 90 • cut plane, and its H-plane in the φ = 0 • cut plane.
Observe from Figs. 2(c) and 3(c) that for both TX and RX, the FPFP antennas are generally better-matched within the operating frequency range, while simultaneously having greater rejection outside of the operating frequency range.In the case of the transmit antennas, the FPFP-TX demonstrates better matching than the LPDA-MAX from approximately 550 to 675 MHz, while in the case of the receive antennas, the FPFP-RX demonstrates better matching than the TSA600 from 1170 to 1375 MHz.This behavior improves the performance of a NLR system by allowing more energy from a desired signal to pass through the antenna, while providing higher rejection for would-be out-of-band interferers.Next, observe from Figs. 4 and 5 that the maximum gain values of the FPFP antennas are greater than those of the COTS antennas.This gain improvement similarly improves the performance of a NLR system by allowing for more directed energy transmitted towards the target (on the TX side), and by acting as a spatial filter to receive more energy from the target (on the RX side).While it is true that the fabricated FPFP antennas weigh significantly more than the COTS antennas, this is not a fundamental requirement of the FPFP antennas.The largest contributing factor to the weight of these antennas is the thickness of the sheet copper, which we selected as 0.8 mm for its combination of pliability and rigidity to allow simple prototyping with our limited in-house fabrication means.However, given slightly more advanced fabrication techniques, future iterations of this design could make further use of lightweight support materials (e.g., precision-cut foam) that would easily allow for a reduction in the thickness of the sheet copper by a factor of four, which would result in weight performance that exceeds those of the COTS antennas considered.Initial simulations suggest no detrimental impact on antenna performance resulting from this reduction of the

III. EXPERIMENTAL RESULTS I
Next, the two sets of antennas were separately tested using the same NLR system.For transmitting, a 650 MHz continuous wave (CW) signal was sent from an Agilent E4400B signal generator into a Mini-Circuits ZHL-20W-13+ amplifier which fed the transmit antenna.Because each of these devices is capable of generating its own harmonic content, filtering between each transmit stage is critical to ensure that a received harmonic signal is generated by the target and not by components of the transmit chain itself.Accordingly, Mini-Circuits VLF-630+ low-pass filters were placed after Fig. 6.System architecture block diagram: the TX chain consisted of an Agilent E4400B signal generator that was fed into a Mini-Circuits power amplifier, which was bracketed by low pass filters to eliminate harmonic signals produced by the signal generator and the power amplifier.The RX chain consisted only of a high-pass filter which fed into a Tektronix RSA306 spectrum analyzer.In this setup, the surrogate target was comprised of a Mini-Circuits ZX60-V63+ amplifier connected to an RFSPACE TSA600 antenna, which provides a nonlinear junction (NLJ) that emulates a radio-frequency front-end.
the signal generator and after the amplifier to eliminate the harmonics generated by each of the transmit chain devices.To construct a surrogate nonlinear target, a Mini-Circuits ZX60-V63+ amplifier was connected to an RFSPACE TSA600 wideband antenna.This surrogate target configuration provides a resonant-nonlinear circuit with characteristics similar to a radio-frequency front end, similar to the targets constructed in [17].For receiving, the receive antenna was fed into a Tektronix RSA306 spectrum analyzer (SA).Two Mini-Circuits VHF-1320+ high-pass filters were placed at the SA input to fully attenuate the f 0 signal reflected by the target, which can produce harmonics in the SA front-end if left unfiltered.The block diagram of the complete system architecture is presented in Fig. 6.
The two sets of antennas were connected to the system, and measurements were taken with varying power levels.For both antenna sets, the transmit and receive antennas were spaced approximately 1.5 m apart from one another.To not exceed the power limitations of the Mini-Circuits filters, power levels output by the signal generator were limited to a maximum of -10 dBm.Accordingly, power levels ranging from -25 to -10 dBm were considered, which equates to a range of roughly 10 dBm to 25 dBm delivered to the transmit antenna, after accounting for transmit amplifier gain and filter insertion losses.At each considered power level, the surrogate target Fig. 7. SNR comparison of the 2f 0 signal obtained by the NLR system using the two sets of antennas.The system was evaluated with the target spaced from 2-5 m away from the baseline of the radar array, while the two antennas were separated 1.5 m from one another.A range of input power levels delivered to the antenna from 10 dBm to 25 dBm are depicted.
was placed at a distance from the baseline of the two receivers varying from 2-5 m.The complete results of this experiment with the two sets of antennas are presented in Fig. 7.In the figure, SNR values are reported with respect to the system noise floor over a bandwidth of 4 MHz.From the figure, observe that use of the custom FPFP antennas improved the performance of the NLR system at every considered power level and range of target distances.In some cases, the SNR of the NLR system was improved by as much as 20 dB.While the target distances considered are not particularly long from an operational perspective, the presented results are expected to scale with power as demonstrated in [13] and [15].

IV. ANTENNA DESIGN II
Next, to further improve the SWaP performance of the FPFP antennas, a design that combines the transmit and receive antennas into a single, integrated, dual-band folded-probe-fed patch (DFPFP) antenna was considered.To achieve this goal, a candidate design was explored in which the receive feed arms were added to the transmit FPFP, opposite the transmit feed arms.For the probe-fed patch design to function as intended, the patch must be at the appropriate height, H, relative to its ground plane.Accordingly, the center portion of the transmit patch was lowered to accommodate operation with the receive feed arms.The original width of the receive patch, W , was used to determine the width of the lowered section of the transmit patch.Finally, the remaining length of the lowered portion in excess of the original receive patch length dimension L was removed.In this way, the original dimensions of the radiating edges of the receive patch are maintained, while the original dimensions of the radiating edges of the transmit patch are also somewhat maintained.This preliminary design is presented in Fig. 8(a).For this initial design, the transmit gain is significantly diminished, reaching a gain (IEEE) at broadside of only 4.3 dBi.By contrast, the receive gain is still relatively high and only reduced to 9.1 dBi at broadside.In addition to the reduced gain, this initial design also suffers from poor matching.While the original FPFP design is capable of being matched to a wide range of input impedance values by varying the height, length, and depth of the feed arms, this mechanism is no longer functional for either the transmit or receive ports in this initial design iteration.
To correct these issues, an additional change was made, shown in Fig. 8(b).In this iteration, the width of the arms that provide support between the receive patch and transmit patch was reduced.Based on the properties of patch antennas, the electric field distribution of a non-radiating edge of a patch antenna is at a maximum at its edges, but at a minimum at its center [30].Accordingly, the support arms were shrunk down to only occupy the electric field null at the center of the nonradiating edge of the receive patch.By making this change, the interference between the natural electric field distributions of the receive and transmit portions of the patch were reduced.In this design iteration, the transmit gain has been largely restored, now reaching a gain of 8.4 dBi.Additionally, the receive gain is also increased, now reaching a value of 10.1 dBi at broadside.Finally, the matching mechanism for the receive port has been restored.However, the matching for the transmit port remains to be addressed.
A final change was made to correct the transmit matching issue.In this iteration, a strip was added to the center of the radiating edge of the receive patch, as shown in Fig. 8(c).The strip was oriented so that its height coincided with the height of the original transmit patch H. Inclusion of this feature provides support for the natural electric field distribution of the transmit patch antenna along its radiating edge, which previously only had a large gap so as not to interfere with operation of the receive patch.Inclusion of this strip results in a further increase to the transmit gain, which now reaches 9.2 dBi.As expected, inclusion of the strip also results in a reduction of the receive gain, which now reaches a value of 8.7 dBi at broadside.Although these gain values are reduced from the original FPFP designs, they still offer significant improvement over the COTS designs.Lastly, both the transmit and the receive port matching mechanisms are working as intended and are capable of matching a variety of input impedance values without the need for an external matching network.
The final, annotated DFPFP design is presented in Fig. 9, while pictures of the fabricated design are provided in Fig. 10.The parameters of the final design as shown in Fig. 9 are presented in Table II.The fabricated DFPFP design measures 0.4 m × 0.4 m × 0.05 m (about 0.9λ × 0.9λ × 0.1λ at 650 MHz) and weighs approximately 1.8 kg.Note that in the final design, the strip protruding from the receive patch has been offset to one side of the antenna.This was done strictly for fabrication reasons and largely does not affect the antenna performance.Instead, by shifting the strip from the center to either side, the receive gain is increased while the transmit gain is simultaneously decreased in roughly equal proportion.Accordingly, shifting the strip results in no net effect on the Friis transmission equation [30].As previously described, the DFPFP antenna was fabricated in-house from 0.8 mm-thick sheet copper.In this case, the copper sheets were cut to size using a rotary tool, and folded into the required shape.
The input reflection coefficients for the transmit and receive antenna ports, S 11 and S 22 , respectively, are presented in Fig.  11.Special attention should be paid to the behavior of the reflection coefficient for the transmit port from 600-700 MHz, and for the receive port from 1200-1400 MHz, as these are the desired bands of operation for each port.Observe from the figure that matching across these ranges for the respective ports is generally better than 10 dB.In general, we observe a good agreement between the measurement and simulation results, with some discrepancies appearing as frequency increases and fabrication imperfections begin to have a more pronounced effect.This effect leads to a somewhat larger receive bandwidth than intended, but there is otherwise generally good agreement, especially within the desired operating frequency ranges.Although coupling between ports is generally a critical consideration for dual-band antennas, it is less relevant for nonlinear radar applications which utilize disparate frequency ranges.Accordingly, external filters can simply be used to  minimize the coupling between ports.Nonetheless, DFPFP S 21 is presented alongside the S 11 and S 22 in Fig. 11.For the DFPFP design, coupling as high as 8 dB is observed at the transmit center frequency.However, the coupled signal is effectively eliminated by use of external filters.The transmit and receive gain is presented in Figs. 12 and 13, respectively.The antenna is again oriented so that its direction of maximum radiation is oriented along the z-axis, with its E-plane in the φ = 90 • cut plane, and its H-plane in the φ = 0 • cut plane.As previously mentioned, offsetting the strip to the side has resulted in a somewhat decreased transmit gain, now around 8.3 dBi at broadside, and a somewhat increased receive gain, now around 9.4 dBi at broadside.In both cases, we observe reasonably good agreement between measurement and simulation results, and in all cases, polarization purity of over 30 dB is maintained for the vertically-polarized radiation pattern.

V. EXPERIMENTAL RESULTS II
Next, the integrated DFPFP was tested in a NLR system using the same setup described in Section III.Corresponding results are presented in Fig. 14.The previous results for the COTS antennas are also once again presented.First, we note that the performance of the DFPFP is somewhat diminished as compared to the performance of the separate FPFP antennas as shown in Fig. 7.This is primarily due to the somewhat reduced gain of the antenna, as well as some insertion loss of the additional inline filters required to mitigate the mutual coupling between ports.The authors note, however, that a Fig. 14.SNR comparison of the 2f 0 signal obtained by the NLR system using two sets of antennas.The system is evaluated with the target spaced from 2-5 m away from the baseline of the radar array, while the two COTS antennas were separated 1.5 m from one another.A range of input power levels delivered to the antenna from 10 dBm to 25 dBm are depicted.significant portion of this mutual coupling is due simply to the co-location of the transmit and receive antennas.Previously, the two antennas were separated by a distance of 1.5 m.In the case of the DFPFP, however, this distance has been reduced to zero.If the two separate transmit/receive antennas were also co-located, they would similarly experience relatively high mutual coupling.Although the performance is somewhat reduced as compared to the separate FPFP antennas, the integrated DFPFP still significantly outperforms the COTS antennas, as depicted in Fig. 14.Furthermore, due to the co-location of its transmit and receive antennas, the DFPFP occupies significantly less volume than the separate transmit and receive antennas.This is a critical feature that allows for the integration of the DFPFP antenna on smaller platforms where space may be tightly constrained.

VI. CONCLUSIONS
In this work, the performance of an NLR system is shown to be significantly improved by replacing general-purpose COTS antennas with custom-designed FPFP antennas.To achieve improved performance, the custom antennas provide higher gain and reduced bandwidth, while still maintaining good linearity and similar SWaP performance.By switching to the custom antennas, SNR improvements as high as 20 dB are observed in the tested NLR system.Next, by leveraging the properties of patch antennas, the two FPFP antennas were integrated into a single, dual-band design in order to further improve the SWaP performance.In this configuration, the DFPFP antenna maintains similar gain, linearity, polarization purity, and bandwidth performance as the individual FPFP antennas in isolation, albeit with a considerably reduced footprint.As compared to two standalone antennas with significant spacing, the integrated DFPFP is capable of being mounted on much smaller platforms, enabling potential use of NLR for new modalities (e.g., small airborne platforms).When tested using the same NLR system, the DFPFP antenna performance is somewhat diminished compared to the two FPFP antennas, but it still significantly outperforms the COTS antennas with improved SWaP performance.Overall, this work directly supports the improved portability of NLR systems, which may enable future systems to be carried by a person or piloted by a UAV.

Fig. 1 .
Fig. 1.Annotated CAD model of the custom FPFP design.Corresponding parameter values are presented in TableI.

Fig. 4 .
Fig. 4. Gain patterns for the custom transmit folded-probe-fed patch and COTS RFSPACE LDPA-MAX antennas in (a) E-plane and (b) H-plane.

Fig. 5 .
Fig. 5. Gain patterns for the custom receive folded-probe-fed patch and COTS RFSPACE TSA600 antennas in (a) E-plane and (b) H-plane.

Fig. 8 .
Fig. 8. Intermediate design steps for the dual-band folded-probe-fed patch antenna (a) initial step resulting in poor TX gain, poor TX and RX matching, but good RX gain.(b) Second step which features reduced-width support arms for the RX patch, which results in good TX and RX gain, good RX matching, but poor TX matching.(c) Final design with a strip added in the center of the RX patch that results in good TX and RX gain and matching.

Fig. 9 .
Fig. 9. Annotated CAD model of the integrated DFPFP design.Corresponding parameter values are presented in TableII.

Fig. 10 .
Fig.10.Isometric view of the fabricated DFPFP antenna.The antenna was fabricated in house using 0.8 mm-thick sheet copper, which was cut to size using a rotary tool and folded into the shape depicted.Inset: angled front view of the antenna.

Fig. 11 .
Fig. 11.Reflection coefficients for the transmit and receive ports of the DFPFP antenna, S 11 and S 22 , and the port coupling, S 21 , shown across both the TX frequency range (600-700 MHz) and RX frequency range (1200-1400 MHz).

Fig. 12 .
Fig. 12. Measured and simulated transmit gain patterns for the dual-band folded-probe-fed patch antenna in (a) E-plane and (b) H-plane.

Fig. 13 .
Fig. 13.Measured and simulated receive gain patterns for the dual-band folded-probe-fed patch antenna in (a) E-plane and (b) H-plane.

TABLE I FINAL
DIMENSIONS FOR TX AND RX FPFP ANTENNAS.

TABLE II FINAL
DIMENSIONS FOR THE INTEGRATED DUAL-BAND FOLDED-PROBE-FED-PATCH ANTENNA PRESENTED IN SEC.IV.