Transmitarray antenna

Transmitarray Antenna also see: Phase-Shifting Surfaces

A transmitarray is a phase-shifting surfaces (PSS), a structure capable of focusing electromagnetic radiation from a source antenna to produce a high-gain beam. Transmitarrays consist of an array of unit cells placed above a source (feeding) antenna . Phase shifts are applied to the unit cells, between elements on the receive and transmit surfaces, to focus the incident wavefronts from the feeding antenna . These thin surfaces can be used instead of a dielectric lens. Unlike phased arrays, transmitarrays do not require a feed network, so losses can be greatly reduced . Similarly, they have an advantage over reflectarrays in that feed blockage is avoided . It is worth clarifying that transmitarrays can be used in both transmit and receive modes: the waves are transmitted through the structure in either direction.

An Overview of Transmitarray Techniques

Transmitarrays can be split into two types: fixed and reconfigurable. First, consider fixed transmitarrays. At each location on the surface of the structure, the unit cells are physically scaled or rotated in order to obtain the required amplitude and phase distribution. Thus, only one focusing direction is available. High aperture efficiency (55%) was achieved in at oblique angles of incidence using precision-machined double split ring slot unit cells. A switched-beam transmitarray covering the 57 – 66 GHz band was reported in . Three different types of unit cells were used, based on patches and coupling slots. Similarly, a 60 GHz design used unit cells with a 2-bit phase resolution and selected an optimal $F/D$ ratio to widen the bandwidth . When $F/D$ = 0.5, a scan loss of 2.2 dB was achieved at a 30° steering angle.

• where $F$ is the focal length and $D$ is the diameter of the transmitarray.

In a reconfigurable transmitarray, the focusing direction is determined by electronically controlling the phase shift through each unit cell . This enables the beam to be steered towards the user. PIN diodes can be used to enable fast phase reconfiguration with an insertion loss below 1 dB . However, a large number of components is typically required, which increases the cost.

It has been shown that transmitarray implementation can be divided into two approaches: layered-scatterer and guided-wave . The first approach uses multiple coupled layers to achieve a phase shift, but has poor sidelobe level (SLL) performance when steering due to higher-order Floquet modes. The second approach enables wider steering, at the expense of increased hardware cost and complexity.

A reconfigurable transmitarray, operating at 29 GHz with circular polarisation, was demonstrated as a beamformer in . A boresight gain of 20.8 dBi was achieved, and the scan loss was 2.5 dB at 40°. Another implementation example is an active Fresnel reflectarray with control circuitry for the PIN diodes . Although the unit cells were optimised, the scan loss was 3.4 dB at 30°. Reconfigurable near-field focusing can be implemented using slots containing PIN diodes . By adjusting the phase compared to a reference wave, holographic principles enabled the use of a compact, planar feeding structure and suppression of undesired lobes. This was extended in to an implementation of a Mills cross based on PIN diodes, in which an aperture was synthesised for imaging applications. Radial stubs were used to isolate the bias lines from the RF. By switching combinations of meta-elements on or off, the scan loss was 0 dB for steering angles of ±30°, but the total efficiency was only 35%.

Different types of unit cells have been used within the same transmitarray. In , slot elements were placed near the centre of the transmitarray, as their polarisation performance is better at normal incidence, whereas double square ring slot elements were used at the edges, as they perform better at oblique incidence angles. This enabled the subtended (flare) angle of the feed horn to be increased, and hence the length of the horn, and the overall antenna size, to be reduced. Unit cells were not required at the centre of the transmitarray, where the phase shift was 0°. This reduced the insertion loss to around 1 dB at 105 GHz, as the majority of the beam amplitude was in the central region. In a different design, SIW aperture coupling was employed to reduce insertion losses and widen the bandwidth of a transmitarray operating at 140 GHz . Due to the large number of vias required, this performance improvement was at the expense of a more complex and costly fabrication.

Recently, a transmitarray was fed by a planar phased array operating at 10 GHz, in order to achieve a high beam crossover gain level whilst maintaining an aperture efficiency of 57.5% . The scan loss was 3.13 dB at ±30°. Similarly, a lens-enhanced phased array antenna is shown in Fig. \ref{ch2_fig_abbaspour} . By combining the beam steering capabilities of phased arrays and the focusing properties of transmitarrays, this hybrid antenna has a smaller form factor , and steers to ±45° in both planes with a 3.2 dB increase in directivity at this angle. Its reconfigurable phase-shifting surface (PSS) contained micro-electro-mechanical (MEMS) switches to change the length of resonators, sandwiched within an antenna-filter-antenna structure. The PSS created the optimal 2D phase distribution needed to achieve high-gain beam focusing, but the MEMS fabrication process was complex and costly, requiring a large number of control lines. MEMS and other mechanical switching methods can achieve a relatively low insertion loss (2.5 dB) and an excellent linearity, but are prone to stiction and reliability issues .

A key challenge in transmitarray design is that the insertion loss increases with the number of conductor layers within the unit cell. In , it was shown that the optimal number of layers to maximise the gain (directivity vs. loss) is 3 layers. This is corroborated by the analysis of cascaded sheet admittances formulated in . However, for scenarios when cost and efficiency are more important, a low-cost two-layer transmitarray may be preferred . Alternatively, the efficiency can be improved by integrating the antenna used to feed the transmitarray within a monolithic chip, as recently demonstrated in the D-band frequency range (114 – 144 GHz) . Another high-gain transmitarray was demonstrated, operating at D-band (110 – 170 GHz) . The $F/D$ was optimised to maximise the aperture efficiency. The antenna was connected to an integrated frequency multiplier to demonstrate a communication link. A data rate of 1 Gbps was achieved over a distance of 2.5 m, with an error vector magnitude (EVM) of 25% .

As explained in Section \ref{impairments}, the gain of a planar structure falls when steering to wide angles, due to having a finite projected area. One possible solution is to use conformal transmitarrays , whose side panels are aligned with wide steering angles. In , a 400-element 3-facet transmitarray achieved a simulated boresight directivity of 27.4 dBi, and a scan loss of 1.9 dB at 50° for a panel tilt angle of 70°. For large values of tilt angle, the fractional bandwidth increased , and the maximum steering angle was extended to 80°. A conformal transmitarray design has also been proposed for aerodynamic antennas in aircraft , with a relatively high boresight SLL of -13 dB. This was demonstrated experimentally in , with a steering range of ±15° and an insertion loss of up to 3.6 dB. Another example achieved mechanical beam steering to ±75° with a simulated gain of 28.9 dBi, albeit with a scan loss of 3 dB .

A multifaceted active deflector operating at 60 GHz was proposed in . This is one of the first examples of a reconfigurable conformal PSS. A 4-faceted pyramid (frustrum) was investigated, along with a half-truncated icosahedral (buckyball) shape. The advantage of these topologies is a wide angular coverage up to ±90°, but the disadvantage is a 3 dB reduction in boresight gain. Each unit cell contained two patches separated by a crossed-slot aperture, and a CMOS power amplifier (PA) integrated circuit (IC). A beam crossover criterion was proposed to achieve a quadratic phase distribution at the join point between panels. A buckyball lens with a negative refractive index using a split-ring resonator (SRR) metamaterial has been patented . The negative refractive index increased the steering range, but the total efficiency was below 50%.

Reconfigurable materials have shown promise for enabling a low-loss beam steering transmitarray. A vanadium dioxide reconfigurable metasurface operating at 100 GHz was presented in . In , a fishnet metamaterial lens was designed, using liquid crystal to achieve a 360° electronically-controlled phase range. The 5 dB unit cell insertion loss could be reduced by controlling the Bloch impedance (both $\epsilon _{r}$ and $\mu _{r}$) of each unit cell . The advantage of liquid crystal is that its loss tangent reduces with frequency, however it suffers from a slow switching time of around 100 ms and fabrication difficulties.

The bifocal technique has been proposed to extend the steering of both reflectarrays and transmitarrays to wider angles . In , the phase distribution was calculated as a Gaussian-weighted mean of the phase distributions for each focus. This was used in .

Another challenge is the high cost of millimetre wave components. Within a transmitarray, each unit cell contains e.g. at least one PIN diode. Thinning algorithms have been used to reduce the number of elements in phased arrays , but these have not yet been applied to transmitarrays. Particle swarm optimisation enables the synthesis of transmitarrays with limited phase range, whilst still maintaining a low SLL . However, this algorithm required 4000 iterations to converge to a solution, and cannot be expressed as a simple rule.

The main finding of this section is that transmitarrays can achieve more flexible reconfiguration compared to fixed lenses, but currently lack the ability to prevent beamwidth broadening and can suffer from large insertion losses.

• could discuss their history?

• moved later

It is useful to compare the performance of a PSS against a switched design with amplitude-only control at the unit cells. In , switchable transparency via plasmon resonances was studied, but as half of the Fresnel zones were blocked, only half of the maximum focusing gain was available, and the SLL was high (-3 dB). Oh et al. reported a phase-shifting surface (PSS) lens fed by a phased array. A boresight gain of 19 dBi was achieved, over a narrow steering range of ±20°. When fed by a dipole in , a focusing gain of 11.2 dB was achieved, with a scan loss of 2.2 dB.

A similar argument applies to transmitarrays. The projected area of the feeding antenna determines the illumination efficiency of a transmitarray panel. Provided that the insertion loss of each unit cell is minimised, an aperture area appropriate to the feed radiation pattern can efficiently focus the wavefronts from the feed. If the panel is tilted to apply focusing at wide steering angles, it could reduce the scan loss.

Thirdly, transmitarrays have excellent focusing properties and a thin profile, but to date, there have been few attempts to mitigate scan loss using a transmitarray. Reconfigurable transmitarrays enable beam steering, but are costly, as they require a large number of components and conductor layers in order to achieve a 360° phase range. Reducing the number of layers and the number of reconfigurable unit cells could simplify the manufacturing process and reduce the cost of a dense deployment of access points.

This chapter presents the design of a conformal transmitarray with thinned reconfiguration, operating at 28 GHz. The transmitarray consists of several panels, which are rotated to align with different beam directions, increasing the gain and thus reducing the scan loss. Within each panel, the unit cells can be reconfigured between two states. Unlike a conventional transmitarray, not all of the unit cells are electronically controlled. A unit cell placement rule is proposed to significantly reduce the required number of reconfigurable unit cells. Reconfigurable unit cells are only placed in regions where a phase reversal is required, hence reducing the biasing complexity and component cost, at the expense of reduced directivity. Additionally, only two layers of substrate are used, which reduces the insertion loss and cost. The design reported in this chapter has a smaller form factor than the design reported in Chapter 4, and provides an alternative method for scan loss mitigation.

As described in Chapter 2, a transmitarray is a phase-shifting surface consisting of an array of unit cells. These focus the wavefronts from a feeding antenna into a narrower beamwidth. The transmitarray is fed by the 8-element linear phased array antenna designed in Chapter 3. Beam focusing is achieved for two different directions, 0° and 53.1° in simulation, using combinations of crossed-slot unit cells. By aligning the focusing direction with the beam direction from the feed array, the scan loss can be reduced compared to conventional planar transmitarray designs.

Geometry and Radiation Pattern

This chapter is structured as follows. Section \ref{ch6_2} describes the transmitarray concept, and Section \ref{ch6_3} explains details of the novel transmitarray, including the unit cell design and unit cell placement rule. Section \ref{ch6_4} presents simulation and measurement results for the fabricated prototypes. Section \ref{ch6_5} compares the performance with the fixed lens design from Chapter 4, and concluding remarks are provided in Section \ref{ch6_6}.

A conventional transmitarray consists of a planar arrangement of unit cells, illuminated by a feed source (Fig. \ref{ch6_fig_coord_system}). For this structure, the required phase distribution is : $\phi _{m}(x_{m},y_{m},z_{m})=-k_{0}(\sin {\theta _{0}}\cos {\phi _{0}}x_{m}+\sin {\theta _{0}}\sin {\phi _{0}}y_{m}+\cos {\theta _{0}}z_{m})$

where ($\theta _{0}$, $\phi _{0}$) are the elevation and azimuth steering directions, and $(x_{m},y_{m},z_{m})$ are the coordinates of unit cell $m$. In Fig. \ref{ch6_fig_coord_system}, $x_{m}=(m+{\frac {M-1}{2}})d$, $y_{n}=(n+{\frac {N-1}{2}})d$, and $z_{m}=0$. $M$ and $N$ are the total numbers of unit cells in the $x$- and $y$-directions respectively.

When steering in azimuth only, this simplifies to : $\phi _{m}(x_{m},y_{m})=k_{0}(d_{m}-\sin {\theta _{0}}(x_{m}\cos {\phi _{0}}+y_{m}\sin {\phi _{0}}))$

where $d_{m}={\sqrt {(x_{m}-x_{f})^{2}+(y_{m}-y_{f})^{2}+z_{f})^{2}}}$

and ($x_{f}$,$y_{f}$,$z_{f}$) are the coordinates of the feed, in this case (0,0,-$F$).

The overall radiation pattern can be calculated, using . Here, terms are combined to express the formula in full: Failed to parse (syntax error): {\displaystyle E(\theta, \phi) = \sum_{m=1}^{M} \sum_{n=1}^{N} \cos^{q_{e}}\left({\theta - \theta_{0}}\right) \frac{\cos^{q_{f}}\left({\theta_{f mn}}\right)}{\sqrt{(md)^2 + (nd)^2 + F^2}}|T_{mn}| e^{j \Psi_{mn}}\\ \times e^{-jk\left(\sqrt{(md)^2 + (nd)^2 + F^2} - d(m\sin{\theta}\cos{\phi} + n\sin{\theta}\sin{\phi})\right)} } where the radiation pattern of the steered array source is modelled as $\cos ^{q_{e}}\left({\theta -\theta _{0}}\right)$. The term $e^{j\Psi _{mn}}$ corresponds to the phases applied to the transmitarray unit cells, to undo the phase variation due to the geometry of the cells from the feed (represented by the right-hand part of (\ref{ch6_eq_txarray})), i.e. $e^{j\Psi _{mn}}e^{-jk\angle G_{mn}}=1$.

$G_{\textrm {edge}}=10\log _{10}\left(\cos ^{n}{\theta _{sub}}\right)$ An edge taper of around -10 dB is desired, so that the illumination efficiency is maximised.

For a planar (conventional) transmitarray, fed by an antenna with radiation pattern $G_{f}(\theta ,\phi )=\cos ^{n}{\theta }$, and subtended angle $\theta _{sub}=\tan ^{-1}\left({\frac {D}{2F}}\right)$, the taper efficiency is calculated by : $\eta _{s}=1-\cos ^{n+1}{\theta _{sub}}$ $\eta _{t}={\frac {2n}{\tan ^{2}{\theta _{sub}}}}{\frac {(1-\cos ^{(n/2)-1}{\theta _{sub}})^{2}}{({\frac {n}{2}}-1)^{2}(1-\cos ^{n}{\theta _{sub}})}}$

$\theta _{sub}$ is a function of $F/D$. Note that $\cos(\tan ^{-1}{x})={\frac {1}{\sqrt {1+x^{2}}}}$, so using $x={\frac {D}{2F}}$, this formula can be expressed in terms of $F/D$, rather than the subtended angle. The illumination efficiency is the product of these: $\eta _{i}=\eta _{s}\eta _{t}$. The overall aperture efficiency $\eta _{ap}$ is obtained by multiplying by material losses and any directivity reduction terms, as evaluated later in Table \ref{ch6_table2} in Section \ref{ch6_4}.

Unit Cell Design

A variety of unit cell shapes have been proposed in the literature, including double square loops , microstrip patches , and slots. The double square loop has the best transmission performance at wide angles of incidence, whereas a large bandwidth can be achieved if Jerusalem cross slots are used. A switchable FSS using MEMS capacitors was demonstrated in . The four-legged loaded element was used to obtain full control of the bandwidth and incidence angle properties. For space applications, in which thermal expansion must be considered, air gaps between layers can be used instead of dielectric, to minimise the insertion loss (metal-only transmitarray) . However, this increases the thickness, and requires a large number of screws for mechanical support.

Fig. \ref{ch6_fig2} shows the structure of the proposed 1-bit unit cell, which operates at 28 GHz. It is based on the design presented in . It consists of two metal layers, printed on a Rogers\textsuperscript{\textregistered} RT5880 substrate material having a thickness of 0.254 mm, a dielectric constant of 2.2, and a loss tangent of 0.0009. Each metal layer consists of a pair of crossed slots, and the incident fields are vertically polarised ($E_{y}$). By selecting a symmetrical unit cell shape, they can be adapted for dual linear or circular polarisation . As shown in Fig. \ref{ch6_fig2c}, the two metal layers are separated by a 3 mm thick layer of ePTFE material (of dielectric constant $\epsilon _{r}$ = 1.4), which creates a 100° phase shift between these layers. The unit cell has reduced thickness and insertion loss compared with multilayer designs such as in .

The unit cell can be reconfigured between two phase states, OFF (0°) and ON (180°), as shown in Figs. \ref{ch6_fig2a} and \ref{ch6_fig2b}. For the OFF state, it has a Jerusalem cross slot structure. In the ON state, the slots are not loaded with Jerusalem cross (JC) shaped caps, producing a large phase change, as displayed in Fig. \ref{ch6_fig3b}. Due to the use of single-pole resonators (a two-layer structure), the transmission performance in Fig. \ref{ch6_fig3a} was challenging to achieve, requiring fine-tuning of the unit cell physical dimensions.

Figs. \ref{ch6_fig_mag_sweep_l} and \ref{ch6_fig_phase_sweep_l} show the change in transmission magnitude and phase as the slot length is swept from 3.464 mm to 4.833 mm. Due to presence of the dielectric layers, which shorten the guided wavelength in the effective medium, the slot length is scaled by a factor of $2/(1+{\sqrt {\epsilon _{r}}})=0.8055$. As shown in Fig. \ref{ch6_fig2b}, setting $l$ = 4.4 gives a length of 3.584 mm across the structure. A continuous phase variation can be observed at 28 GHz. Sweeping the length of the slot achieved a large phase change of almost 270° (from -143.1° to +125.8°), but the transmission magnitude drops significantly, to -2.157 dB, if the slot is too long ($l$ = 4.6). This is because it only resonates when its length is half a guided wavelength. Additionally, if the slot length is increased, it can overlap with adjacent unit cells. Hence, JC cells had to be used to create a large phase shift, and the cap length had to be optimised.

Both unit cell states were simulated in CST Microwave Studio\textsuperscript{\textregistered} using Floquet ports and the frequency domain solver. Simulation details are available in Appendix \ref{AppdxB_sim_methods</ref>. Fig. \ref{ch6_fig3} shows the magnitude and phase of the $E_{y}$ transmission coefficient through the unit cell in ON and OFF states. A phase change of 189° was observed, which is close to 180°, and the transmission magnitude is at least -1.76 dB at 28 GHz for both states. As observed in Figs. \ref{ch6_fig_CS_fields} and \ref{ch6_fig_JC_fields}, in the ON state, the surface current density, which is proportional to the H field magnitude, is concentrated around the ends of the slots. ON state, the $H_{x}$ phase varies in a uniform way. For both the ON and OFF states, the $E_{y}$ magnitude is large within the slots, and varies in the $z$-direction. Both the $E_{y}$ and $H_{x}$ phase vary rapidly at the air-copper boundaries. In the OFF state, the resonance of the caps creates a large phase shift, because the currents travel through a longer path. As shown in Fig. \ref{ch6_fig_surface_current}, for the JC cells, the surface currents are in opposite directions (anti-phase) on each conductor layers, whereas for the CS cells, the surface currents are in the same direction (in-phase).

The phase difference between states is given by: $\Delta \phi =\angle {S_{21}}_{\textrm {OFF}}-\angle {S_{21}}_{\textrm {ON}}$.

Electronic reconfiguration can be achieved by several possible methods. Liquid crystal was initially considered for the design, as the material can be biased by applying a voltage between two parallel conducting plates. However, it was not selected, due to several issues. The liquid must be hermetically sealed in a cavity, and the crystal orientations aligned with the cavity walls in an unbiased state. The liquid can flow between cells, causing a variation in the RF properties of the transmitarray, and dynamic instabilities .

Alternatively, PIN diodes could be placed across the ends of the Jerusalem cross caps, applying a different bias voltage for each state. DC blocking in the form of interdigital capacitors would be needed to isolate the bias voltages , and RF choke inductors would be needed at the ends of the bias lines. To demonstrate the transmitarray concept, unit cells with fixed phase shifts were used in the fabricated prototypes. For electronic reconfiguration, PIN diodes would need to be placed on both the top and bottom layers. When the diodes are forward biased (ON), incident radiation is transmitted through the slots with a 180° phase change, but when the diodes are reverse biased (OFF), the current path is lengthened so that there is minimal phase change (around 0°).

The MACOM MA4GP907 diode has an ON resistance $R_{\textrm {ON}}$ = 4.2 $\Omega$, an OFF resistance $R_{\textrm {OFF}}$ = 300 k$\Omega$, and small parasitic inductance and capacitance values ($L_{\textrm {ON}}$ = 0.05 nH, $C_{\textrm {OFF}}$ = 42 fF in the 28 GHz band) . Given that it has a high OFF resistance value, and that the switching time is very fast (2 ns), this component is suitable for the design.

The position and orientation of the bias lines must be chosen to minimise their effect on the transmission of the incident waves through the structure. If the lines are sufficiently narrow (width up to 0.1 mm), they will present a high impedance, so will have less effect on the wavefronts . As they act as a polarising grid, the bias lines should be perpendicular to the incident field direction . This design has no ground plane, so each group of active unit cells must have both a $V_{\textrm {bias}}$ and a ground connection. As groups of cells share the same bias voltages, these lines can be routed between adjacent cells. The required number of external control lines is equal to the number of beam directions supported, so is inversely proportional to the steering resolution.

The bias lines could be implemented as large blocks of copper around the unit cells, separated by thin gaps (through which the RF wave propagation is heavily attenuated). The gaps may need to be meandered to form DC block capacitors. Radial stubs or high-impedance lines of length ${\frac {\lambda _{g}}{4}}$ could be used as chokes (inductors) on the external control lines, to prevent the RF signal from affecting the DC control circuitry .

External links

• www.example.com