07950927.pdf

Joshua M. Kovitz, Jean Paul Santos, Yahya Rahmat-Samii, Neil F. Chamberlain, and Richard E. Hodges

Enhancing Communications for Future Mars Rovers Using high-performance circularly polarized patch subarrays for a dual-band direct-to-Earth link.

T

he recent success of the Mars Science Laboratory (MSL) has garnered worldwide interest in further exploration of Mars. In this article, we discuss our latest developments in circularly polarized (CP) patch subarrays to be integrated into a larger 30-dB gain array. This larger array will facilitate an enhanced dual-band direct-to-Earth (DTE) link with NASA’s Deep Space Network at the X band. The frequency ratio between both bands is rather small, with a value of 1.17, and current dual-band or wide-band CP patch elements are not easily suited for X-band frequencies, antenna array requirements, or the harsh environments in space. Our array elements are single-feed and single-layer, easing fabrication requirements. Each element achieves circular polarization through an optimized CP half E-shaped patch design. More importantly, the subarray achieves the desired directivity, axial ratio (AR), and impedance matching performance with a compact design, allowing straightforward integration into the larger array for future Mars rovers. Our measurements of the 4 × 4 subarray prototype demonstrate the desired performance for creating an X-band DTE link.

The Path Toward an Enhanced DTE Mars Rover Link Space exploration has been an exciting enterprise in understanding our universe. The excitement comes not only from fundamental discoveries that strengthen our scientific knowledge but also from the evolution of the various technologies needed to support space exploration missions. NASA is no exception, cultivating spacecraft technologies to reliably gather valuable scientific knowledge for study. In interplanetary and deep space missions, NASA’s Deep Space Network is the primary conduit for communicating with spacecraft. For current Mars rovers, which include the Mars Exploration Rover Opportunity and the MSL Rover Curiosity, science data are returned to Earth using an ultrahigh-frequency band relay link to a spacecraft in orbit around Mars, such as the Mars Reconnaissance Orbiter depicted in Figure 1. The Mars orbiter subsequently relays the science data to Earth, transmitting from a Small Deep-Space Transponder (SDST) X-band radio and a high-gain antenna (HGA) [1], [2]. Opportunity and Curiosity are also equipped with a mechanically steerable X-band HGA and an SDST radio with a 15-W output. This system is used for commanding the rovers directly from Earth on the uplink and for sending mainly small amounts of telemetry directly to Earth on the downlink. Digital Object Identifier 10.1109/MAP.2017.2706651 Date of publication: 16 June 2017

50

1045-9243/17©2017IEEE

August 2017

IEEE Antennas & Propagation Magazine

mars rover—NASA/JPL-CALTECH, satellite, space, and earth—images licensed by ingram publishing

The current X-band HGA has a 25-dB gain [3]–[6]. Unfortunately, the relatively large space loss of the DTE link and the relatively low effective radiated power from the rover HGA preclude the practical transmission of science data directly to Earth. However, with the recent success of larger rovers, such IEEE Antennas & Propagation Magazine

August 2017

as Curiosity, high-performance DTE communication systems are now a viable option for future large-rover missions. The performance upgrade relies on a larger power amplifier and a larger antenna system. Developing a novel antenna—with higher gain and power handling than the current X-band HGA—would 51

Mars Reconnaissance Orbiter Indirect (Relay) Communications DTE Communications

Future Mars Rovers

FIGURE 1. Current Mars rover communications primarily use a satellite relay to communicate with Earth. Our vision in this article is to develop a high-performance antenna array significantly enhancing DTE links for Mars rovers. (Mars Rover and Mars Orbiter images courtesy of NASA/JPLCaltech. Earth image courtesy of Wikimedia Commons.)

Subarray Tiles

CP Elements

Waveguide to Stripline Connectors To Transceiver 1–16 Waveguide Power Divider

FIGURE 2. An exploded view of the complete 30-dBic array consisting of 4 × 4 subarrays. Each subarray is fed by the 1-to-16 waveguide power divider with high-power handling (100 W). Our goal is to develop a high-performance CP subarray for such an array implementation. enable greater flexibility and higher data rates. The big question is: which class of antenna would be suited for this goal? Based on a study performed at NASA’s Jet Propulsion Laboratory for enhanced DTE communications, the following key requirements were identified for a potential enhanced HGA system in future Mars rovers: ■■ right-hand circular polarization (RHCP) ■■ a boresight gain above 30 dBic at the transmit (Tx) band (8.425 GHz)—assuming that a circular polarization gain in units of dBic is defined as G RH (i, z) = ^4rU RH (i, z) h Pin, where U RH is the radiation intensity of the RHCP compot (i, z) | 2 /2h nent defined by U RH (i, z) = r 2 | E (r, i, z) ·R 52

and Pin is the input power to the antenna, with the RHCP t = (it - jzt ) / 2 polarization vector being defined as R ■■ operational frequency bands at a 70-MHz bandwidth centered at 8.425 GHz for transmit downlink (Tx: Mars to Earth) and at a 47-MHz bandwidth centered at 7.1675 GHz for receive uplink (Rx: Earth to Mars) ■■ power handling of 100 W continuous wave in the Martian atmosphere ■■ beam steerable in azimuth and elevation using a mechanical gimbal 3 ■■ an antenna volume less than 38 × 38 × 5 cm . The specific frequency bands were allocated to the Mars 2020 mission, and we use them to demonstrate our concept for future missions operating in similar bands. An important aspect of this enhanced novel design was the accommodation of the larger antenna on the rover, including its stowage during launch, cruise, entry, descent, and landing. This constrained the volume of the design (not including the gimbal system). Circular polarization was chosen for interoperability with Deep Space Network Earth stations. Using circular polarization removes the need for polarization alignment and reduces atmospheric effects, such as Faraday rotation. Meeting each of these individual requirements is difficult in itself. Simultaneously meeting all these requirements, however, is a complex task that requires a delicate antenna design balance. A wide variety of antenna techniques were considered in solving this design problem. Some examples include reflector antennas [7], waveguide slot arrays [8], and continuous transverse stub arrays [9]. The array architecture shown in Figure  2 was proposed as a compromise to meet mass, volume, and power handling requirements, which we found that none of the other antenna concepts met. This architecture consists of smaller patch subarray tiles fed by a waveguide power divider, as illustrated in Figure 2. The waveguide power divider enables high-power handling from the transceiver. For this array architecture, the development of the CP subarray tiles remains one of the most important steps, where the requirements of being dual-band, CP, and fabrication friendly for X-band frequencies must be satisfied.

DUAL-BAND CP PATCH ELEMENT DESIGN THE CURRENT STATE OF THE ART IN THE LITERATURE In any array design, the heart of the challenge lies with the proper design of the array elements. Since the element performance tends to dictate the array performance, choosing the right element is a crucial task. Finding a single-layer, single-feed element that supports circular polarization over two frequency bands is not straightforward, especially if the frequency ratio is low. Traditional CP patch antennas are fairly narrow band, where their combined |S11| # -10 dB and AR # 3 dB bandwidth is usually less than 2% [10]. Adding dual-band performance in both S11 and AR further compounds the difficulty. Researchers have investigated different techniques to add dual-band performance in CP patch antennas, including stacked patch antennas [11]–[14], August 2017

IEEE Antennas & Propagation Magazine

PROPOSED CP HALF E-SHAPED PATCH DESIGN One development to support the needed requirements above is through the use of the CP half E-shaped patch antenna shown in Figure 3, which has shown CP bandwidths similar to its counterparts. The design features a nearly 50% size reduction [24] from the CP E-shaped patch antenna [21]. The basic idea behind the CP half E-shaped patch antenna is to add a shorting bar across the slot of the half E-shaped patch antenna in [25]. Without the shorting bar, the half E-shaped patch antenna would be linearly polarized in the y direction, according to the coordinate system in Figure  3. This  shorting bar enables a new mode to radiate in the x direction and can be tuned to provide circular polarization, as shown in [26]. To test the performance of the CP half E-shaped patch antenna, the CP half E-shaped design was developed in the X band. The goal was to achieve good CP performance, defined as S 11 # −10 dB and AR # 3 dB, within both the Rx and Tx frequencies [27]. Previous S-band designs IEEE Antennas & Propagation Magazine

August 2017

50 mm

z

50 mm W

L

Ws B

yf

Ls

xf Ps

x

εr = 1.96

4.318 mm

slotted patch shapes [15], [16], and slotted ground planes [17], [18]. The issues with many of these designs is that they require multiple layers or very thin slots that could not be effectively scaled to X-band frequencies. Some designs (such as the slotted ground plane design) do not have those issues, but they do result in reduced antenna directivity, because the slot radiates behind the antenna. Other antennas managed to obtain nice dual-band features, but the frequency ratio between the two operational bands was quite large. Wide-band single-layer CP patch antennas offer another possibility in meeting the dual-band CP requirement. The bandwidth required to support both Tx and Rx would be a 16.1% AR-S11 bandwidth. Designs such as the CP E-shaped patch antenna [19]–[21], the CP U-slot patch antenna [22], the L-shaped probe [23], and the capacitively compensated traditional CP patch antenna [10] have shown increased bandwidth by using thick, low-permittivity substrates. The advantage of such patch elements is that they can be single-fed on a single layer, reducing various fabrication complexities and other mechanical issues that could be encountered in multilayer CP patch antennas. The CP E-shaped patch antenna has been able to achieve up to a 17% AR-S11 bandwidth, which would match  our design requirements. The only catch is that this wideband patch design is too large to place into a planar array design, considering that the width in [21] was 0.7m 0 at the upper frequency. Designs such as the L-shaped probe are difficult to fabricate and might be prone to mechanical failure in space applications. Our literature search did not produce any other possible wideband design suitable for our application. Since the CP E-shaped patch antenna showed a wide bandwidth with a simple fabrication procedure, we used it as the starting point for a deeper search. This began our investigation into a new shape that would offer dualband capabilities.

y

Ground Plane

FIGURE 3. The CP half E-shaped antenna geometry. The design exhibits dual-band CP support within a compact size. TABLE 1. THE DIMENSIONS OF THE CP HALF E-SHAPED PATCH IN THE X BAND (IN MILLIMETERS). Design

W

L

Ls

Ws

Ps

X band

15.1

11.5

9.8

3.0

3.3 1.6 0.9

yf

xf

, 2.3

of the CP half E-shaped antennas show a very wide-band S 11 response, yet only a single resonance in AR [24]. However, by increasing the substrate thickness, wide-band or dual-band AR-S11 designs may be possible. These previous S-band designs developed at the University of California, Los Angeles (UCLA), were frequency scaled to the X band and reoptimized for a Rogers RT Duroid 5880LZ substrate, which has a permittivity of e = 1.96. This substrate is lightweight and offers a low z-axis coefficient of thermal expansion to avoid via failure during the thermal cycle. The thickest available substrate is 4.318 mm. The ­d imensions of the CP half E-shaped patch antenna were tuned until good AR-S11 performance was obtained. The tuned X-band design parameters are listed in Table  1. As shown by the results in Figure 4, wide-band S 11 was achieved, and a dual-band AR was observed. Note that the measured AR  performance did not quite fall within the desired bands, but we fixed this through the use of ­optimization when incorporating the element into the subarray, as described later. The radiation patterns are also decent in both the Tx and Rx bands, with a slight beam squint in the Tx band shown in Figure 5. Beam squints are fairly common with thicksubstrate CP patch antennas and have been further investigated on the CP E-shaped patch in [28]. The results in [28] reveal that the beam tilt is likely due to higher-order modal distributions causing variations in the electric field phase at the antenna aperture. From these results, one might assert that the beam tilt in the CP half E-shaped patch in the Tx frequency is most likely due to similar phase variations at the aperture. More importantly, good array performance 53

0

10 Rx

–10

Tx

AR (dB)

|S11| (dB)

–5

–15 –20

8 6

CONFIGURATION OF THE 4 × 4 SUBARRAY

4

With the selection of the CP half E-shaped patch element for the array, a proper array configuration must be chosen to achieve the highest gain possible within a nominal footprint while avoiding any grating lobe effects. An important consideration for array planning in this context is beam steering. Mars rover systems utilize mechanical gimbals to steer the HGA beam [3], and it is likely that future rovers will use a similar gimbal. Mechanically steered systems avoid the complex feed networks needed in electronically steered phased arrays [29]. This reduces our requirements to a broadsidedirected array, simplifying the feed network design to obtain equal phase and magnitude for the element excitations. For the subarray, the physical area is limited to 9.5 × 9.5 cm2 , which is 1/16 of the total array area requirement of 38 × 38 cm2. With this requirement, the maximum directivity associated with this subarray area can be calculated using

2

–25 –30

was still demonstrated in spite of these beam squints. Within the array environment, the beam squint of the CP half E-shaped patch antenna was diminished due to the array factor. In summary, the CP half E-shaped patch antenna provides good dual-band performance with its compact size.

12

7

7.5

8

0 7

8.5

Rx 7.5

8

Tx 8.5

Frequency (GHz)

Frequency (GHz)

Simulation Measurement

Simulation Measurement

(a)

(b)

FIGURE 4. (a) The simulated (in HFSS) and (b) measured S 11 and AR performance of the CP half E-shaped patch element. A good dual-band performance can be observed at the Rx/Tx bands.

0 –30

0

30

–60

60 –10 0 dB

60

–60

D0 =

4 rA p m2

, (1)

FIGURE 5. The radiation pattern of the X band CP half E-shaped patch antenna at the Rx band: (a) z = 0c , (b) z = 90c , and at the Tx band, (c) z = 0c , and (d) z = 90c. LHCP: left-hand circular polarization.

where A p is the physical area of the subarray, m is the wavelength, and 100% aperture efficiency h ap is assumed. The maximum directivity of this 9.5 × 9.5-cm2 area is 19.52 dBic at the Tx frequency and 18.11 dBic at the Rx frequency. The key to realizing this maximum possible directivity is to choose an array configuration that utilizes the aperture area most efficiently and avoids grating lobe effects. Another important aspect in this investigation is to prevent the elements from touching each other. Even though the CP half E-shaped patch antenna has a compact size of 15.1 × 11.5 mm2, mutual coupling can still be strong because of the thick substrate. Several possible array configurations were examined. The representative configurations include 3 × 3, 3 × 3 without a center element, and 4 × 4, as depicted in Figure 6. For each array topology, the feed network simplicity, grating lobe effects, directivity, and mutual coupling were considered in detail. Among these three candidates, the 4 × 4 configuration was attractive in its feed network sim­plicity and

August 2017

IEEE Antennas & Propagation Magazine

–30

–90

–20

90

120

–120 –150

180

–30

–90

–30

0

–30

30

–20

90 120

180

180

90

150

0

30

–60

–10 0 dB

–120 –150

0 dB

(b)

60

–30

–10

120 –150

150

–60 –90

–20

–120

(a)

150

(c) RHCP–Simulation RHCP–Measurement

54

30

–30

–30

–90

–20

60 –10 0 dB 90

120

–120 –150

150

180 (d)

LHCP–Simulation LHCP–Measurement

where the beamwidths in the xz and yz planes would be controlled by the q x and q y values, respectively. Our calculations used q x = 1.7 and q y = 1.6 at the Tx band as well as q x = 1.4 and q y = 1.0 for the Rx band. The q values are based on a −10-dB beamwidth match with the CP half E-shaped patch element patterns for each frequency. The radiation pattern shows good sidelobe levels, with minimal radiation toward the horizon (i = 90c). This is observed at both the Tx and Rx frequencies, as observed in Figure 7. The maximum subarray directivity achieved by the 4 × 4 subarray is higher than the 3 × 3 and the 3 × 3 without center array configurations at a value of 19.89 dBic and 18.49 dBic for the Tx and Rx frequencies, respectively. The directivities of the full array having 16 × 16 elements were 31.6 dBic and 30.2 dBic for the Tx and Rx frequencies, respectively.

(b)

(c)

FIGURE 6. These subarray topologies were compared to choose the configuration that meets the requirements: (a) 3 × 3, (b) 3 × 3 without center, and (c) 4 × 4.

0

Rx

Tx

−10 −20 −30 −40

0 15 30 45 60 75 90 θ (°) (a)

Normalized Pattern (dB)



-jz f (i, z) = e ^it cos q x (i) - jzt cos q y (i) h, i # 90c, (2) 2

(a)

Normalized Pattern (dB)

radiation pattern performance. In particular, a corporate feed network could easily be manifested using four stages of 2-to-1 power dividers in a single layer. Also, 2-to-1 power dividers are the easiest divider to make broadband. The 3 × 3 configuration, in contrast, would be forced to use two stages of 3-to-1 power dividers, which are not as easily made broadband. Multiple stripline layers could be used to create the two stages, but this comes at the cost of further fabrication and mechanical complexity. The 4 × 4 subarray’s radiation performance was calculated based on array theory and showed desirable performance. The element pattern uses a cos q (i) pattern [30], which takes on the form

0

Rx

Tx

−10 −20 −30 −40

0 15 30 45 60 75 90 θ (°) (b)

FIGURE 7. The radiation pattern ( z = 0c cut) of the 4 × 4 subarray configuration with a center-to-center element spacing of 23.75 mm: (a) a subarray and (b) a full array with 16 × 16 elements.

DESIGN OF THE X-BAND OPTIMIZED SUBARRAY USING CP HALF E-SHAPED ELEMENTS The next challenge to overcome was integrating the CP half E-shaped element into a 4 × 4 subarray. Mutual coupling and other unwanted effects can deteriorate the desired performance. Using the CP half E-shaped element described in the previous section, a 4 × 4 array was created. It is further revealed that the AR performance for the 4 × 4 array was dramatically detuned when the element was placed in the array environment. Particle swarm optimization (PSO) [31], [32] was then used to reoptimize the elements in the array ­environment. The “4 × 4 Feed N ­ etwork Design” section details the stripline feed network design used to feed each ­element. The section concludes with the discussion of the integration of the elements and the feed network.

ELEMENT OPTIMIZATION USING A COAX FEED MODEL Modeling the array using currently available simulation tools can be accomplished in many ways. As a first step, we simulated the 4 × 4 subarray by explicitly placing 16 copies of the CP half E-shaped element in an ANSYS HFSS simulation model as shown in ­Figure 8. Each of the 16 elements was individually excited with a separate 50-Ω coaxial cable. This IEEE Antennas & Propagation Magazine

August 2017

FIGURE 8. The CP half E-shaped patch subarray is simulated under the 4 × 4 subarray configuration to evaluate its performance. Coaxial feeds are used to excite each element individually. coaxial excitation’s performance is similar to a via connected to a stripline feed network underneath the patch ground plane. The center-to-center element spacing was 23.75 mm (or 0.67 m 0 at the Tx). As shown in Figure 9 under “original design,” the impedance-matching performance (C) is adequate, achieving levels below −10 dB in both the Tx and Rx bands. The AR performance is decent in the Rx band but dramatically increases in the Tx band. This confirms that interactions between the elements have indeed worsened the performance of the previously tuned single CP half E-shaped element. This is expected because of the close spacing of the elements and the weak confinement of the fields underneath the elements. The weak field confinement can be explained by the use of thick substrates, where the fringe fields tend 55

to extend further than patch antennas with thin substrates. These effects ultimately result in poor AR performance at the Tx frequency, leaving us with only one choice: optimization. Nature-inspired optimization techniques offer an interesting approach to the antenna design problem [32]. Algorithms such as PSO mimic observed processes in nature to help guide the search for a good design [31]. PSO is employed to help improve the performance of the array. Our particular case requires good performance for both S 11 and AR at the Rx and Tx bands. Thus, we used the ­f itness function f (x) = max (VSWR (fRx), VSWR (fTx), 2 AR (fRx), 2 AR (fTx)),  (3)

0

10

–5

8

–10

AR (dB)

Γ (dB)

where x = [W, L, L s, W s, Ps, yf , xf , ,] is the vector of optimization parameters illustrated in Figure 3, VSWR is the voltage standing wave ratio and AR is the AR in magnitude (not in decibels). The frequencies were set to fRx = 7.1675 GHz and fTx = 8.425 GHz. We scaled the AR performance by 2 within the fitness function to equalize the goals for each parameter as VSWR = 2 and 2 AR = 2 for an AR of 3 dB. Notice that every element shares the same geometry, e.g., W, L, ,, and so forth. With the array explicitly implemented in HFSS, i.e., without periodic boundary conditions, the AR performance

–15 –20 –25 –30

Rx 7

4 2

Tx

0

7.5 8 8.5 Frequency (GHz) (a) Original Design

6

7

Rx Tx 7.5 8 8.5 Frequency (GHz) (b)

Optimized Design

FIGURE 9. The performance of the coax-fed array model before and after optimization: (a) the impedance matching performance C = V -in /V +in and (b) the broadside AR performance, which shows significant improvements from PSO.

Sp

Z0/M Vin+ V0

Vin–

1-to-M Power Divider V–

Z0 .. .

could be easily deduced by setting equimagnitude, equiphase excitation coefficients for the elements, and computing the respective AR at i = 0. The VSWR, however, proved to be a more challenging parameter to compute. The difficulty can be visualized by examining Figure  10. There are two components that should be included to compute the final array VSWR performance: the feed network (with M + 1 ports) and the antenna array (with M input ports). If we designate the left (input) port of the feed network as port 1, then the remaining ports 2, 3,…, M + 1 are connected to the antenna array. The active reflection coefficient available in HFSS can give a rough idea about the array input VSWR performance, but it does not include the feed network effects. There are two options to include the feed network effects. The first is to draw the stripline feed network explicitly in HFSS underneath the ground plane of the elements. Unfortunately, the simulation time for such a model can take an extremely long time because of the fine geometric features within the feed network. The second option is to use an approximate circuit model. Using a similar approach to that in [21], the circuit model assumes an ideal, lossless, reciprocal 1-to-M power divider providing equimagnitude, equiphase excitations to each of the elements. The source impedance is assumed to be Z 0 /M, and each port exciting the elements is assumed to have a characteristic impedance of Z 0 . In our case, M = 16. With this model, we are implicitly assuming that the impedance conversion is obtained. In other words, we are excluding the effects of the impedance transformers, assuming that a dual-band or wideband transformer can be integrated into the design without significant performance degradation. The impedance matching C can be found by computing V in for a given V + in . By writing the (M + 1) # (M + 1) S-parameter matrix S p for the feed network as

where S p, 11 is a scalar and s ps is a column vector defined as s Tps = 6S p, 21 ... S p,(M +1) 1@, the submatrix Slp is found by deleting the first row and column from S p as



SL M-Element Array V+

S p, 11 s Tps G, (4) Sp = = s ps Slp



R S p, 23 S S p, 22 S S p, 32 S p, 33 Slp = S h h S SS p,(M +1) 2 S p,(M +1) 3 T

V g S p, 2 (M +1) W g S p, 3 (M +1) W W .(5) j h W g S p,(M +1)(M +1)W X

From Figure 10, we have the equations V - = S L V +(6) VV+ ; in+E = S p ; in-E .(7) V V



By substituting (4) into these equations, one can finally obtain

FIGURE 10. A circuit model used to predict the impedance matching performance of the 4 × 4 subarray, where M = 16. 56



-1 1 in C = V+ = S p, 11 + s Tps ^S L - Slp h s ps . (8) V in

August 2017

IEEE Antennas & Propagation Magazine

In this equation, HFSS computes the values for the array S-parameter matrix S L, and the elements of S p can be derived based on our assumptions for the particular network configuration. In our case, a T-junction power divider was chosen for our feed network, whose ideal S-parameter matrix is given as



R S S S 1 = Sp M SS SS T

V 0 M M g M W M 1 -M 1 g 1 W W j j h W .(9) M 1 h j 1 -M 1 1 W W g M 1 1 1 - MW X

Once C is computed, the VSWR can be obtained: VSWR = (1 +| C |) / (1 -| C |). Note also that if the S-parameters are known versus the frequency for the 1-to-16 power ­divider network, these can be used for S p. In this article, we used the simplified, ideal model given by (9), since the feed network was being designed in parallel. The approximate circuit model reduced the computational time to find an optimized solution, because small features seen in the stripline network were not simulated. The optimization was performed on the whole 4 × 4 array environment, where each element’s dimensions were changed identically. As the results show in Figure 9 under “optimized design,” the optimization was successful not only in maintaining proper impedance matching performance in both the Tx and Rx frequencies but in significantly improving the AR performance. It should be emphasized that the impedance matching performance was slightly degraded for the “optimized design” case, but this sacrifice enabled a significant improvement in the AR at the Tx frequencies. The power of the PSO procedure is in systematically finding a design that balances the performance requirements.

the integrated array assembly. In practice, more vias could be placed to improve the isolation among the ports, but from our results, the four vias around each port were sufficient. In looking at both the S 11 and S 1n simulation results in Figure 12, adequate performance was achieved. The S 1n coefficients were acquired from HFSS and were used to predict the radiation pattern and the directivity

2

4

10

12

3

5

11

13

1 6

8

14

16

7

9

15

17

FIGURE 11. The 4 × 4 subarray feed network implementation in the stripline. Note that four grounding vias were placed around each port to improve isolation and minimize parallelplate mode excitation. In practice, more vias are typically added for improved performance.

0



m Tx

8

#s#

m Tx

2

–10

Rx

–20 –30

–10 –11 –12

Rx

Tx

–13 –14

, (10)

where m Tx is the wavelength at the Tx frequency in free space. This spacing was used to prevent a potential difference between the ground planes. Therefore, four vias were placed around each port in the feed network shown in Figure 11, with a spacing of m Tx /6 to ease the handmade fabrication process of IEEE Antennas & Propagation Magazine

August 2017

Tx

–40 7 7.5 8 8.5 Frequency (GHz) (a)

|Sm1| (dB)

The next step in the design process was the feed network. The feed point was placed at the center so that each branch to each element was equidistant to the other branches, establishing an equiphase design. The feed network is shown in Figure 11. We used a binomial impedance transformer and power dividers to equally split the power in the feed network while maintaining a broadband impedance matching. The feed network was implemented in stripline to avoid unwanted radiation. The 1-to-16 power divider was designed with coaxial outputs  to simplify the model and mimic probe-fed, impedance-matched patch antennas. Grounding vias along the transmission line were also recommended [33], [34], separated with a spacing s as

|S11| (dB)

4 × 4 FEED NETWORK DESIGN

–15 7 7.5 8 8.5 Frequency (GHz) (b)

m=2 m=4 m=6 m=8 m = 10 m = 12 m = 14 m = 16

m=3 m=5 m=7 m=9 m = 11 m = 13 m = 15 m = 17

FIGURE 12. The S-parameters of the feed network: (a) Good S11 performance was achieved and (b) nearly equal power was delivered to the loads within the Rx and Tx bands. 57

TABLE 2. THE ELEMENT DIMENSIONS AFTER OPTIMIZATION IN THE SUBARRAY ENVIRONMENT (IN MILLIMETERS). Design

W

After PSO

14.0 12.0 8.52

L

Ls

Ws

Ps

2.62 4.72

yf

xf

,

2.46 1.71 1.09

using array theory. The results shown in Figure 13 are compared to an ideal array having perfect uniform amplitude coefficients that achieves the desired broadside radiation pattern. The radiation patterns of the actual coefficients and the uniform amplitude coefficients match well, achieving good broadside radiation with minimal sidelobe levels. The directivity is also quite high, where 19.9 dBic and 18.5 dBic at the Tx and Rx frequencies, respectively, closely match the estimates given in the “Configuration of the 4 × 4 Subarray” section.

be integrated to form the final array assembly. The overall layer stack-up is shown in Figure 14. The SMA connector, stripline feed network, and antenna elements are integrated together in HFSS. To continue to find the best performance, the assembly dimensions were further manually tuned. The parameters include the line widths of the stripline, the dimensions of the patch elements, and the length of the quarter-wave transformers. We tuned the length of the quarter-wave transformers to the Tx frequency, since the impedance matching performance is more critical when transmitting from the rover. Looking at the results in Figure 15, this integrated assembly achieved a wide AR-S11 bandwidth. The radiation patterns of the assembly in Figure  16 also achieved very good broadside radiation. The simulation results show that the achieved directivity is 19.19 dBic and 18.09 dBic at the Tx and Rx frequencies, respectively.

SUBARRAY FABRICATION AND MEASUREMENTS

Normalized Pattern (dB)

The final integrated design was studied, and its performance adequately met the requirements; therefore, a prototype could be assembled. Each layer in the prototype was fabricated using INTEGRATION OF FEED NETWORK AND CP ELEMENTS photolithography. The insertion of vias and the combining of Now that each of the components for the array were develeach of the layers of the prototype were done by hand assemoped, the 4 × 4 CP half E-shaped element array and the bly through a systematic procedure. A vector network analyzer uniform amplitude and phase stripline feed network could was used to measure the S 11 performance, and the UCLA spherical near-field chamber was used to measure the AR, directivity, and radiation patterns. These measurements were 0 then compared to simulation. The radiation characteristics UA Coefficients shown in Figure 16 match well with those of simulations. The Actual Coefficients fabricated prototype and pattern coordinate system are shown –10 in Figure 17. The radiation patterns in both frequencies have low sidelobe levels and exhibit good broadside radiation, which –20 shows that the array factor greatly reduces the effect of the beam squints for this size of array. The measured AR in Figure 15 is much better than the –30 simulated AR, while the directivity is 18.74 dBic and 17.85 dBic for the Tx and Rx frequencies, respectively. The directivities –40 compare well within the simulation results, and the slight 0 20 40 60 80 drop is caused by slightly larger sidelobes and off-axis cross θ (°) polarization. The S11 discrepancy between simulation and measurement was studied intensively, and it was found that the FIGURE 13. The simulated radiation pattern z = 0° at the Tx reason for the difference was the fabrication tolerances created band shows that good broadside radiation was achieved. Similar results were observed in the Rx frequency. UA: through the chemical etching. When the stripline widths were uniform amplitude. measured after fabrication, the widths were not the same as those simulated in HFSS. When looking at the percent difference, noticeable Antenna Layer changes were observed. The dimen4.318-mm Rogers Duroid 5880LZ sions of the two designs are shown in Table  3. Even though the changes Ground Plane are not altogether drastic ( # 8 mils), 0.787-mm Rogers 5880LZ Stripline Layer the characteristic impedance Z 0 can 0.787-mm Rogers 5880LZ Ground Plane be dramatically changed as seen in SMA Connector Table 3. Although the array produced good S 11 performance, more precise fabrication methods could improve FIGURE 14. The previously developed components are integrated together to form performance to avoid these differences. the final subarray assembly. 58

August 2017

IEEE Antennas & Propagation Magazine

PERFORMANCE IN A 16 × 16 ARRAY CONFIGURATION

0

12

–10

10

–20

8

–30

6

–40

4

–50

2

–60

Rx 7

7.5

8

Tx 0 8.5

AR Simulated AR Measured

FIGURE 15. The simulated and measured AR-S11 performance

of the subarray assembly. Reasonably good impedance matching and broadside AR were obtained. IEEE Antennas & Propagation Magazine

0

30 60

–60 –90

–30

–20

August 2017

–10 0 dB 90

–120

120 –150

–30

180 (a) 0

150

30 60

–60 –30

–90

–20

–10 0 dB 90 120

–120 –150

–30

180 (b) 0

150

30 60

–60 –30

–90

–20

–10 0 dB 90 120

–120 –150

180 (c) 0

150

30 60

–60

Frequency (GHz) S11 Simulated S11 Measured

–30

–30

AR (dB)

|S11| (dB)

Future Mars rovers ultimately would use an array of these subarrays (Figure 2) to achieve the 30-dBic gain required to communicate back to Earth. Integrating 4 × 4 of these subarrays results in the required gain. Figure 18 plots the directivity along z = 0c for the array (using a similar coordinate system orientation as shown in Figure 17). The center-to-center spacing between each of the subarrays is 9.5 cm. With a center-to-center element spacing of 23.75 mm, the elements are uniformly spaced throughout the entire 16 × 16 array. The radiation patterns were found by taking the simulated subarray in HFSS and applying the respective array factor. Each subarray was excited with equal amplitudes, which is reflected in the . -13 dB sidelobes in the pattern. Mutual coupling within each 4 × 4 subarray was incorporated through full-wave analysis within HFSS, but the mutual coupling between each subarray was not incorporated into the simulation. Since the array is steered toward broadside and the subarray patterns account for their internal mutual coupling, the radiation pattern is well predicted by this approximation within the main beam and first few sidelobes, as observed by previous studies shown in [35]. Furthermore, these studies in [35] reveal that the directivity calculation is not significantly impacted. The half-power beamwidths are roughly 5.7c and 4.76c for the Rx and Tx bands, respectively. Similar performance was also observed in the z = 90c plane. Clearly, if lower sidelobes are desired, then a nonuniform amplitude taper can be implemented in the power dividing network. The gain to maintain the high data rate link is 28.8 dBic and 30 dBic for the Rx and Tx frequencies, respectively. Since waveguide power dividers and their respective components have been well characterized in the literature, we can do a simple calculation to estimate the gain of our entire system. Gain/ loss numbers are listed in Table 4. The simulated maximum directivity values are also provided.

–30

–90

–20

–10 0 dB

–120

90 120

–150

180 (d)

150

RHCP–Simulation LHCP–Simulation RHCP–Measurement LHCP–Measurement

FIGURE 16. The simulated and measured radiation patterns of the X-band CP half E-shaped array for the (a) Rx band at z = 0c, (b) Rx band at z = 90c, (c) Tx band at z = 0c, and (d) Tx band at z = 90c. 59

TABLE 3. THE FABRICATION TOLERANCES OF THE TRANSMISSION LINE.

TABLE 4. THE GAIN/LOSS BUDGET FOR THE PROPOSED ARRAY ARCHITECTURE.

Desired Z0

HFSS Width

Actual Width

Actual Z0

Quantity

Rx (7.1675 GHz)

Tx (8.425 GHz)

35 Ω

2.30 mm

2.18 mm

38.5 Ω

Simulated directivity

30.12 dBic

31.25 dBic

50 Ω

1.42 mm

1.23 mm

57.5 Ω

Simulated subarray loss

0.5 dB

0.59 dB

70.7 Ω

0.81 mm

0.63 mm

83.3 Ω

Waveguide feed loss

0.08 dB

0.07 dB

Waveguide probe loss

0.10 dB

0.10 dB

Mismatch loss

0.58 dB

0.38 dB

Total loss

1.26 dB

1.14 dB

Estimated gain

28.9 dBic

30.11 dBic

Gain requirement

28.8 dBic

30.0 dBic

9.5 cm

9.5 cm

y

x

30 20 10 0 –10 –20

Simulated Directivity (dBic)

Simulated Directivity (dBic)

FIGURE 17. The top view of the final prototyped CP half E-shaped patch array.

30 20 10 0 –10 –20

–20 –10 0 10 20 θ (°)

–20 –10 0 10 20 θ (°)

(a)

(b)

RHCP

LHCP

FIGURE 18. The simulated directivity of the full 30-dBic array at z = 0c for the (a) Rx band and (b) Tx band. The patterns were computed using the array factor for 4 × 4 subarrays spaced 9.5 cm with equal amplitudes. The losses are detailed in Table 4. The subarray loss is given by the HFSS simulation, assuming the Groiss model having 0.5-μm root-mean-square surface errors and a ­17.5-μm copper cladding thickness for all copper conductors [36]. Dielectric losses were included in the simulation as well. The interconnect was modeled in the simulation, and thus the subarray efficiency 60

includes the interconnect loss. The waveguide losses are computed for the waveguide length needed to feed the 16 subarrays as well as the waveguide-to-coax transition. The mismatch loss is also included, and the listing shows the mismatch losses associated with the measured subarray impedance matching performance. These values are a bit higher because of issues encountered  in handcrafting these multilayer boards. It is believed that fabricating the design with automated multilayer processing (versus handmaking the antenna) will lead to improved performance and improve the overall gain of the design at both frequencies. Despite this issue, the estimated gain shows that we achieved the overall gain requirement with a small margin to spare, an important outcome for this project.

CONCLUSIONs An antenna system with a 30-dBic gain and high-power handling has been conceptualized and designed with the aim to support the return of science data through the Mars rover DTE link. The antenna design requirements call for simultaneous dual-band, high-gain, compact, and CP performance. In this article, a CP half E-shaped patch antenna was used, featuring a compact, single-feed, single-layer design amenable to array applications. The design eases the fabrication complexity while still achieving good AR-S11 and radiation pattern performance. To support the high-gain requirement, a 4 × 4 subarray configuration featuring a stripline feed network with grounding vias was developed. The stripline feed network is designed so as to provide broadside radiation with minimal back radiation. These components are then integrated, first through a full-wave simulation in HFSS and finally through a fabricated prototype. The design was measured using the facilities at UCLA and demonstrated good performance. In summary, the developed 4 × 4 subarray proved to be a strong design candidate that can support the challenging requirements of such a mission without added fabrication complexity.

ACKNOWLEDGMENTS This work was supported in part by a Strategic University Research Partnerships grant through the Jet Propulsion Laboratory, California Institute of Technology, Pasadena, under contract with NASA. August 2017

IEEE Antennas & Propagation Magazine

AUTHOR INFORMATION Joshua M. Kovitz ([email protected]) received his Ph.D. degree in electrical engineering from the University of California, Los Angeles (UCLA), in 2015. He is currently a postdoctoral scholar at UCLA. His research focuses on open challenges within the emerging areas of reconfigurable antennas, cognitive radios, CubeSats, and millimeter-wave antennas. Jean Paul Santos ([email protected]) received his M.S. degree in electrical engineering from the University of California, Los Angeles (UCLA), in 2015. He is currently pursuing his Ph.D. degree at UCLA while working as an antenna/radio frequency engineer for the Naval Air Warfare Center Weapons Division at Point Mugu, California. His current research involves electrically small antenna design and array concepts. Yahya Rahmat-Samii ([email protected]) is a University of California, Los Angeles, distinguished professor. He is a holder of the Northrop Grumman chair in electromagnetics and a member of the U.S. National Academy of Engineering. He won the 2011 IEEE Electromagnetics Field Award and is the recipient of the 2016 IEEE Antennas and Propagation Society John Kraus Antenna Award. Neil F. Chamberlain ([email protected]) is a senior engineer in the Flight Communications Systems Section at the NASA Jet Propulsion Laboratory, Pasadena, California. Richard E. Hodges ([email protected]) is the supervisor of the Spacecraft Antennas Group and principal engineer at the NASA Jet Propulsion Laboratory, Pasadena, California. His current research interests include spaceborne deployable reflectarray antennas, deployable reflectors, and waveguide slot array antennas.

REFERENCES

[1] NASA. (2014, Nov.). Technologies of broad benefit: telecommunications, [Online]. Available: http://mars.jpl.nasa.gov/mer/technology/bb_telecommunications .html [2] D. Bell, S. Allen, N. Chamberlain, M. Danos, C. Edwards, R. Gladden, D. Herman, S. Huh, P. Ilott, T. Jedrey, T. Khanampornpan, A. Kwok, R. Mendoza, K. Peters, S. Sburlan, M. Shihabi, and R. Thomas, “MRO relay telecom support of Mars science laboratory surface operations,” in Proc. IEEE Aerospace Conf., 2014, pp. 1–10. [3] A. Olea, A. Montesano, C. Montesano, and S. Arenas, “X-band high gain antenna qualified for Mars atmosphere,” in Proc. 2010 4th European Conf. Antennas Propagation (EuCAP), 2010, pp. 1–5. [4] P. R. Brown, R. E. Hodges, and J. C. Chen, Antennas for Deep Space Applications. Hoboken, NJ: Wiley, 2012, pp. 664–694. [5] J. Taylor, A. Makovsky, A. Barbieri, R. Tung, P. Estabrook, and A. G. Thomas. (2005, Oct.). Mars Exploration Rover telecommunications. [Online]. Available: http://descanso.jpl.nasa.gov/DPSummary/MER_article_ cmp20051028.pdf [6] A. Makovsky, P. Ilott, and J. Taylor. (2009, Nov.). Mars Science Laboratory telecommunications system design. [Online]. Available: http://descanso.jpl.nasa .gov/DPSummary/Descanso14_MSL_Telecom.pdf [7] Y. Rahmat-Samii and R. Haupt, “Reflector antenna developments: A perspective on the past, present, and future,” IEEE Antennas Propag. Mag., vol. 57, pp. 85–95, Apr. 2015. [8] R. Elliott, “The design of waveguide-fed slot arrays,” in Antenna Handbook, Y. Lo and S. Lee, Eds. New York: Springer-Verlag, 1988, pp. 805–842. [9] W. W. Milroy, “Continuous transverse stub element devices and methods of making same,” U.S. Patent 5,266,961, Nov. 1993. [10] J. M. Kovitz and Y. Rahmat-Samii, “Using thick substrates and capacitive probe compensation to enhance the bandwidth of traditional CP patch antennas,” IEEE Trans. Antennas Propag., vol. 62, pp. 4970–4979, Oct. 2014. [11] C. Su and K. Wong, “A dual-band GPS microstrip antenna,” Microwave Opt. Technol. Lett., vol. 33, no. 4, pp. 238–240, 2002.

IEEE Antennas & Propagation Magazine

August 2017

[12] P. Nayeri, K.-F. Lee, A. Z. Elsherbeni, and F. Yang, “Dual-band circularly polarized antennas using stacked patches with asymmetric U-slots,” IEEE Antennas Wireless Propag. Lett., vol. 10, pp. 492–495, 2011. [13] “A dual-band circularly-polarized stacked patch antenna with a small frequency-ratio,” in Proc. IEEE Antennas Propagation Soc. Int. Symp., Orlando, FL, 2013, pp. 942–943. [14] Nasimuddin, X. Qing, and Z. N. Chen, “A wideband circularly polarized stacked slotted microstrip patch antenna,” IEEE Antennas Propag. Mag., vol. 55, no. 6, pp. 84–99, 2013. [15] K.-P. Yang and K.-L. Wong, “Dual-band circularly-polarized square microstrip antenna,” IEEE Trans. Antennas Propag., vol. 49, no. 3, pp. 377–382, 2001. [16] W. Liao and Q.-X. Chu, “Dual-band circularly polarized microstrip antenna with small frequency ratio,” Prog. Electromagnet. Res. Lett., vol. 15, pp. 145–152, 2010. [17] Nasimuddin, Z. N. Chen, X. Qing, “Dual-band circularly polarized S-shaped slotted patch antenna with a small frequency-ratio,” IEEE Trans. Antennas Propag., vol. 58, no. 6, pp. 2112–2115, 2010. [18] X. Bao and M. Ammann, “Dual-frequency circularly-polarized patch antenna with compact size and small frequency ratio,” IEEE Trans. Antennas Propag., vol. 55, no. 7, pp. 2104–2107, 2007. [19] F. Yang, X. Zhang, X. Ye, and Y. Rahmat-Samii, “Wide-band E-shaped patch antennas for wireless communications,” IEEE Trans. Antennas Propag., vol. 49, no. 7, pp. 1094–1100, 2001. [20] A. Khidre, K. Lee, F. Yang, and A. Elsherbeni, “Wideband circularly polarized E-shaped patch antenna for wireless applications,” IEEE Antennas Propag. Mag., vol. 52, no. 5, pp. 219–229, 2010. [21] J. M. Kovitz, H. Rajagopalan, and Y. Rahmat-Samii, “Design and implementation of broadband MEMS RHCP/LHCP reconfigurable arrays using rotated E-shaped patch elements,” IEEE Trans. Antennas Propag., vol. 63, pp. 2497–2507, June 2015. [22] K.-F. Tong and T.-P. Wong, “Circularly polarized U-slot antenna,” IEEE Trans. Antennas Propag., vol. 55, pp. 2382–2385, Aug. 2007. [23] S. S. Yang, K. Lee, A. A. Kishk, and K. Luk, “Design and study of wideband single feed circularly polarized microstrip antennas,” Prog. Electromagnet. Res., vol. 80, pp. 45–61, 2008. [24] J. M. Kovitz, J. P. Santos, and Y. Rahmat-Samii, “The CP half E-shaped patch: Evolving from linear polarization to compact single feed circularly polarized antennas,” in Proc. IEEE Antennas Propagation Soc. Int. Symp., Vancouver, BC, Canada, 2015, pp. 1918–1919. [25] R. Chair, C. L. Mak, K.-F. Lee, K.-M. Luk, and A. A. Kishk, “Miniature wide-band half U-slot and half E-shaped patch antennas,” IEEE Trans. Antennas Propag., vol. 53, pp. 2645–2652, Aug. 2005. [26] J. Kovitz, H. Rajagopalan, and Y. Rahmat-Samii, “Circularly polarised half E-shaped patch antenna: A compact and fabrication-friendly design,” IET Microw. Antennas Propag., vol. 10, no. 9, pp. 932–938, 2016. [27] W. Langston and D. Jackson, “Impedance, axial-ratio, and receive-power bandwidths of microstrip antennas,” IEEE Trans. Antennas Propag., vol. 52, pp. 2769–2774, Oct. 2004. [28] J. M. Kovitz and Y. Rahmat-Samii, “Understanding the radiation pattern anomalies in the broadband CP reconfigurable E-shaped patch,” in Proc. IEEE Antennas Propagation Soc. Int. Symp., Memphis, TN, 2014, pp. 1580–1581. [29] R. Haupt and Y. Rahmat-Samii, “Antenna array developments: A perspective on the past, present and future,” IEEE Antennas Propag. Mag., vol. 57, pp. 86–96, Feb. 2015. [30] Y. Rahmat-Samii and S.-W. Lee, “Directivity of planar array feeds for satellite reflector applications,” IEEE Trans. Antennas Propag., vol. 31, no. 3, pp. 463–470, 1983. [31] J. Robinson and Y. Rahmat-Samii, “Particle swarm optimization in electromagnetics,” IEEE Trans. Antennas Propag., vol. 52, no. 2, pp. 397–407, 2004. [32] Y. Rahmat-Samii, J. M. Kovitz, and H. Rajagopalan, “Nature-inspired optimization techniques in communication antenna designs,” Proc. IEEE, vol. 100, pp. 2132–2144, July 2012. [33] D. Pozar, Microwave Engineering, 3rd ed. Hoboken, NJ: Wiley, 2005. [34] I. Rosu. Microstrip, stripline, and CPW design. [Online]. Available: http:// www.qsl.net/va3iul/Microstrip_Stripline_CPW_Design/Microstrip_Stripline_ and_CPW_Design.pdf [35] A. A. Kishk, “Prediction of large array characteristics from small array parameters,” in Proc. 2007 2nd European Conf. Antennas Propagation (EuCAP), Nov. 2007. doi: 10.1049/ic.2007.1111. [36] S. Groiss, I. Bardi, O. Biro, K. Preis, and K. Richter, “Parameters of lossy cavity resonators calculated by the finite element method,” IEEE Trans. Magn., vol. 32, no. 3, pp. 894–897, 1996. 

61