Patent Application
20230074075
The present disclosure generally relates to reconfigurable wireless networks and, more particularly, to steerable antenna devices for implementing node-to-node and backhaul communications in wireless mesh networks.
Wireless mesh networks can bring flexible Internet connectivity to outdoor environments. A mesh network includes multiple wireless nodes, at least some of which are connected to each other, along with nodes that are “wired” into the Internet for backhaul communication. One advantage of the mesh networks is their resilience. When one node malfunctions, the wireless traffic can be automatically rerouted through other nodes.
Network scalability of mesh networks, however, remains a significant challenge. Particularly, throughput loss per hop can lead to significant performance degradation as the coverage area and number of nodes increases. Because a communication path between an access node of a mesh network and a node connected to the Internet may include multiple hops between adjacent nodes, losses from each hop multiply, leading to exponential signal loss from multiple hops. At least in part, the losses for each hop stem from radio interference (e.g., from neighboring nodes). Better radios, and, in particular, antenna devices can ameliorate interference problems to improve performance.
The antenna devices and techniques described in this disclosure can improve wireless mesh network performance at least in part by reducing radio interference among distinct node links. In particular, an antenna array may be configured with a discrete set of phase options for each antenna element and directionally-disordered antenna placement to steer direction and/or directivity while substantially minimizing radiation pattern side lobes.
In one implementation, a steerable antenna device for a reconfigurable wireless mesh network comprises a substrate including a network of antenna feed traces connected to a primary feed port. The steerable antenna device further comprises a directionally-disordered quasi-uniform two-dimensional array including a plurality of antenna elements attached to the substrate, the array configured to operate at an operating wavelength. Still further, the steerable antenna device comprises a plurality of switches for each one of the plurality of antenna elements, the switches configured to select, for each one of the plurality of antenna elements, a respective phase delay from a respective set of possible phase delays by selecting a respective path from a set of possible respective paths in the network of antenna feed traces. Additionally, the steerable antenna device comprises a controller configured to: i) obtain a pointing direction of the steerable antenna array, and ii) control the switches to select, for each one of the plurality of antenna elements, the respective phase delay based on the obtained pointing direction of the steerable antenna device.
The methods and devices described in this disclosure can improve operation of radio devices for wireless mesh networks. Mesh network radio devices can include steerable antenna arrays which can have radiation patterns with a main lobe in a certain pointing direction, configured by selecting a carrier phase from a set of possible phases for each antenna element. Radiation pattern side lobes, however, can cause interference which, in turn, can increase signal loss or decrease throughput (e.g., cause increased bit error rates, dropped packets, etc.).
To ameliorate throughput degradation, steerable antenna devices can be configured to substantially minimize radiation pattern side lobes. One approach, described in the present disclosure, includes introducing directional disorder in an antenna array. A directionally disordered array includes elements that are arranged in no particular direction, i.e. statistical difference between any two directions is substantially minimized. For example, the antenna elements in the array are not arranged in lines, rectilinear grids, nor with any other Cartesian regularity. One implementation includes arranging antenna elements along a Fermat spiral at incremental azimuthal intervals determined by the golden ratio (i.e., the golden angle), as described below.
The steerable antenna device 100 includes a substrate 110 at which a primary feed port 112 and a network of antenna feed traces, such as traces 120a-e of
A controller 160 may be configured to control each of the multiplexers 140a-d to select, for each antenna, a respective phase delay by selecting among alternative paths between the primary feed port 112 and the antenna element. The controller may select a path in view of an intended radiation direction, other radiation pattern constraints, and one or more operating wavelength.
The legend in
The substrate 110 in
Traces (e.g., traces 120a-j) may, for example, be printed, machined (e.g., by removing part of a metallic layer), or lithographically defined on the substrate 110. The traces may implement transmission lines (e.g., coplanar, microstrip, etc.) with suitable characteristic impedances (e.g. 25, 50, 75, 100Ω, etc.). In some implementations, as illustrated in
Traces may be configured to meander along the substrate 110 to have equal cumulative lengths between the primary feed port 112 and each of the antenna elements (i.e., antenna element feeds). Alternatively, total paths lengths to antenna feeds may vary by integral number of wavelengths (in the transmission lines). Still alternatively, the total path lengths may vary by fractions of wavelengths and may be compensate by the phase-selecting MUXs, as discussed in more detail below.
Antenna elements (e.g., 130a-e), splitters (122a-d), MUXs (140a-d) are discussed in more detail with reference to, respectively,
Although the device 100 is illustrated in
Generally, antenna elements need not be monopoles. For example, antenna elements may be dipoles. In some implementations, the two halves of a dipole may be on opposite side of the substrate 110 (e.g., the plane of the substrate). In other dipole implementations, both halves of a dipole may be on the same side of the substrate, and a portion of the feed for the dipole may run along the length of the dipole, departing from the substrate 110 and electrically connecting to the trace feeding the antenna.
Still more generally, the antenna 130e, may have any suitable shape and need not be a monopole nor a dipole antenna. Furthermore, the antennas 130a-e (or, for that matter, any of the antennas in the device 100 need not be identical to one another. In the case of monopole implementations, monopole lengths or capacitive loading may vary. Still in some implementations, the antennas 130a-e may be of different types.
The substrate 110 may include a ground plane 170. The ground plane 170 and the monopole antenna 130e may together terminate a microstrip or a coplanar transmission line implementing the trace 120f. The ground plane 170 may be implemented on either or both sides of the substrate 110. In the implementations where the ground plane 170 is disposed at both sides of the substrate (or within the substrate), portions of the ground plane 170 may be electrically connected, for example, using vias. The substrate 110 may include an electrically insulating region 172, isolating the pole of the antenna 130e from the ground plane 110.
The splitters 122a-f need not be equal power splitters. For example, a 1:2 ratio splitter followed by 1:1 ratio splitter may equally partition power to three antenna elements. Furthermore, in some implementations, powers fed to distinct antenna elements (e.g., antennas 130a-e) may not be equal.
The phase delays 142a-d may be implemented with different length transmission line segments. Additionally or alternatively, the phase delays 142a-d may be implemented with filters. In either case, the amount of phase in each of the phase delays 142a-d may depend on the frequency of a radio signal. For narrowband signals, the variability of phase delays across the band can be negligible. On the other hand, phase delay variability with respect to wavelength may be designed for broadband operation. For example, a redundant number of phase delays, non-uniform distributions of phase delays, and engineered dispersion of the phase delays may help with broadband operation. Furthermore, rather than broadband operation across a range of wavelengths, the delays may be designed for a select group of two or more wavelengths.
The MUX 140e may include two digital selector inputs 144a, b corresponding to two selection bits B0 and B1. The selection bits can determine which of the phase delays 142a-d add to the total propagation phase delay of the trace 120j. Analogously, MUXs (e.g., MUXs 140a-d) for other antenna elements (e.g., antenna elements 130a-d) may have respective selector inputs for bits determining corresponding phase delays. The controller 160 may determine and send a two-bit selection to each MUX (e.g., MUXs 140a-e) in the device 100 to set one of four possible phases at each antenna element (e.g., antenna elements 130a-e) to implement a phased antenna array.
In general, MUXs may provide any suitable number of alternative paths. The number of possible paths to each antenna element may be a power of two. For example, a MUX selecting among eight paths may be implemented with three selector bits. Generally, the number of possible delays and selector bits may trade off phase resolution (which, as discussed below, may somewhat affect side lobe suppression ratio) and propagation loss in between a central feed and an antenna element. The propagation loss may be affected by the increased number of switches in any given feed path between the central feed and an antenna element, as described below.
The digital selector inputs 144a,b may be logical inputs using, for example, transistor-transistor logic (TTL) or diode-transistor logic (DTL), or complimentary metal-oxide (CMOS) integrated circuits. The digital selector inputs 144a,b may accept digital signals in parallel. In some implementations, on the other hand, two bits to determine a MUX phase may be sent to the MUX in series. Generally, the MUX may include electronics to select the phase based on a sequence of bits.
A radiation pattern of the device 100 set by the phases sent to the MUXs (e.g. MUXs 140a-e) by the controller 160 may depend on the spatial arrangement of the antenna elements (e.g., antenna elements 130a-e), the geometry of the antenna elements themselves, and the configuration of the substrate 110. In particular, regular structures (e.g., statistically anisotropic patterns), in the arrangement of antenna elements (e.g., antenna elements 130a-e) may lead to spurious maxima (i.e., lobes) in the radiation pattern of the device 100. Thus, reducing such regularities in structure may enable radiation patterns with large side-lobe suppression.
The elements of the array 200 (e.g., the elements 230a-d), represented by small open circles, are arranged to minimize directionality (i.e., directional order). For the purpose of illustration, four cardinal direction lines 252a-d and three concentric circles 254a-c partition the plane of the array 200 into eight slices and three annular regions. A center point 256 of the partition may be the first geometric moment of the array 200 or another suitable center point. With respect to the center point 256, the four cardinal lines 252a-d are uniformly distributed along the angular coordinate of a polar coordinate system centered at the center point 256. The concentric circles 254a-c are at uniformly increasing radii of the polar coordinates with respect to the center point 256.
The elements of the array 200 do not tend toward any one of the cardinal lines 252a-d, nor any intermediate direction. The angular distribution of the elements can be described as directionally disordered. A metric of directional disorder in the array 200 may be defined and used for constructing the array 200. For example, an optimization function may be constructed with the metric of directional disorder, possibly along with other optimization parameters. Such an optimization function, for example, may be a weighted sum or a weighted sum of squares of the various optimization parameters. In some implementations, the optimization function may be maximized using a search among various candidate array patterns. In other implementations, the optimization function may be maximized iteratively, using, for example, a gradient descent algorithm. Additionally or alternatively, an array (e.g., the array 200) may be selected based on achieving a metric of directional disorder that is above a predetermined threshold of the metric.
In some implementations, a metric of directional disorder may be an inverse of amplitude of correlation between radial and azimuthal coordinates (between 0 and 2π radians) of elements (e.g., the elements 230a-d). For example, a correlation coefficient of 0.1 would yield a higher directional disorder than a correlation coefficient of −0.5. A threshold correlation magnitude for sufficient directional disorder may be 0.1, 0.2, 0.3, 0.4, 0.5 or another suitable threshold.
In other implementations, the metric of directional disorder may be the measure of isotropy of the array (e.g., array 200). In other words, a directional disorder metric may be a metric reflective of the isotropy. One such metric may be variability in a histogram of elements with respect to azimuthal directions. For example, an eight-bin histogram may be constructed for the array 200 based on the sectors (i.e., wedges) between cardinal direction lines 252a-d. The number of elements in each such edge varies between four and five. In other implementations, a histogram may be constructed with overlapping bins. In any case, a metric of isotropy may be defined as relative variability among bin counts. A threshold isotropy metric for a directionally-disordered array (e.g., array 200) may be 10%, 20%, 30% or any other fraction of an average bin count.
Other metrics of directional disorder and/or isotropy may include a measure of entropy with respect to azimuthal position of array elements, variability of moments of array coordinates projected on cardinal direction lines (e.g., lines 252a-d), etc.
Besides directional disorder, the array 200 may be configured for quasi-uniformity. Generally, in a quasi-uniform array, the elements may be substantially evenly distributed over a region, albeit not on a regular grid. An array optimization function may include a metric of uniformity along with a metric for directional disorder.
One metric of uniformity or quasi-uniformity may be based on an inverse of relative variance among nearest-neighbor distances of array elements. An additional or alternative metric of uniformity may be based on modeling elements as having identical electrical charges. Then, for each element (e.g., elements 230a-d), the sum of virtual forces from all of the other elements, and, possibly, a boundary represented by a circularly-distributed charge may be calculated. A variance in the magnitudes of the virtual forces on each element, relative to the mean force, may be used as a measure of uniformity.
In yet another implementation, a local density at each element location may be calculated as a sum of values, at the location of the element, of isotropic kernels centered at the locations of the other elements. The effect of a circular boundary may be represented by an isotropic boundary function decreasing radially inward. The measure of uniformity may be derived from the statistical distribution of local densities.
Still another measure of uniformity may be based on a statistical distribution of Voronoi cells defined by array element locations. This measure is described in more detail with reference to
The mean density of elements within the array (e.g., array 200) may be set based on a number of considerations. For example, the density may be a trade-off between reducing coupling between neighboring antennas and device compactness. The density may be configured, for example, to ensure a minimum spacing between antenna elements with respect to a nominal wavelength. The minimum spacing may be 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1 or any other suitable multiplier of a nominal wavelength.
As discussed above, a suitable optimization algorithm may yield, based on the metrics above, a suitably directionally disordered and quasi-uniform array. In some implementations, however, locations of elements in a directionally-disordered quasi-uniform array may be determined directly using a closed-form equation, as discussed with reference to
Mathematically, the Fermat spiral is given by the polar equation, r=a√{square root over (θ)}, with radius r varying as the square root of angle θ, and proportionally to a scaling constant a. One property of the Fermat spiral is that it encloses approximately equal areas with each subsequent loop (i.e., 2π increment in θ). The scaling constant, a, may be chosen to achieve a minimum spacing constraint as discussed above. Furthermore, the scaling constant may be selected in view of the intended operating wavelength or a set of operating wavelengths of the device.
The array 260 may be generated by placing elements at azimuthal position given by the equation, θn=θ0+nθG, where the n-th azimuthal position θn is the sum of the initial azimuthal position θ0 and n times the golden angle of π(3−√{square root over (5)}) radians. As the golden angle is maximally irrational, the array 260, placed along the Fermat spiral 265 is directionally disordered.
In some implementations, the initial angle θ0 is zero. In other implementations, the initial angle may be chosen to optimize sideband rejection ratios in one or more radiation directions.
The array 260 has quasi-uniformity owing to the property of the Fermat spiral of enclosing substantially equal areas with every turn. Thus, each element (e.g., 264a-i) has approximately the same area apportioned to it as described, for example, in more detail with reference to
An area of Voronoi cell may define the local density of the array at the location of the element corresponding to the Voronoi cell. For example, the local density may be defined as the inverse of the area of the Voronoi cell. In other implementations, the local density may be based on the Voronoi cell using another suitable algorithm. In summary, the array 200 may be designed using a quasi-uniformity measure based on Voronoi cell areas.
An array of antennas (e.g., the array 200) may be designed so that a mean (or median) of the distribution of Voronoi cell areas is within a certain range of values encompassing a target Voronoi cell area. The target Voronoi cell area may be given by AV=Cλ2, where λ is an operating wavelength and C is a constant (e.g., 0.01, 0.02, 0.05, 0.1, 0.2, 0.5, 1, 2, etc.) selected to achieve desired spacing between antenna elements, as described above.
In one operating mode, the input line 341a may be electrically connected to the output line 341b via the switch 342a, the delay line 343a, and the switch 342b. In another operating mode, the input line 341 a may be electrically connected to the output line 341a via the switch 342a, the delay line 343b, and the switch 342b. A binary digital logic signal B0 applied to the digital selector line 344 may select between the two operating regimes by controlling the switches 342a,b. That is, for example, when the signal B0 at the digital selector line 344 is high (e.g., binary 1), the input line 341a may be electrically connected to the output line 341b via the phase delay line 343a. Conversely, when the signal B0 at the digital selector line 344 is low (e.g., binary 0), the input line 341a may be electrically connected to the output line 341b via the phase delay line 343b.
The multiplexer 345 includes an input line 346a, an output line 346b, a connecting line 346c, four single-pole double-throw switches 347a-d, four phase delay lines or, simply, delay lines 348a-d, and digital selector lines 349a,b. The phase delay lines 348a-d may be implemented as traces on a suitable substrate. In one implementation, the trace 348a and the trace 348c may be of equal lengths, while the trace 348b may have extra length to implement a π/2 phase delay at an operating frequency, and the trace 348d may have extra length to implement a π phase delay at the operating frequency.
In operation, binary digital logic signals B0,1 applied to the digital selector lines 349a,b may select among four possible phase delays between the input line 346a and the output line 346b. More specifically, the binary digital logic signal B0 controls the switches 347a, b to select between the delay lines 348a,b to make an electrical connection between the input line 346a and the connecting line 346c. On the other hand, the binary digital logic signal Bi controls the switches 347c,d to select between the delay lines 348c,d to make an electrical connection between the connecting line 346c and the output line 346b.
In one operating mode, a (0, 0) two-bit combination (of B0, B1) applied to the digital selector lines 349a,b may connect the input line 346a to the output line 346b via the delay lines 348a and c having, in combination, a nominally zero phase delay. In another operating mode, a (1, 0) two-bit combination (of B0, B1) applied to the digital selector lines 349a,b may connect the input line 346a to the output line 346b via the delay lines 348b and c having, in combination, a π/2 additional phase delay. In yet another operating mode, a (0, 1) two-bit combination (of B0, B1) applied to the digital selector lines 349a,b may connect the input line 346a to the output line 346b via the delay lines 348a and d having, in combination, a π additional phase delay. Finally, a (1, 1) two-bit combination (of B0, B1) applied to the digital selector lines 349a,b may connect the input line 346a to the output line 346b via the delay lines 348b,d having, in combination, a 3π/2 additional phase delay.
In the manner described with reference to
In a coordinate system 400 of
ϕi=−2πdi/λ,
where di is a delay distance between the element 406 and the origin 401 along the direction of propagation (e.g., given by ray 408a) and λ is the wavelength. The delay distance, in turn, may be computed as
d
i
=r
i cos(Θr−Θi).
Thus, the radiation phase delay may be written as
ϕi(Θr)=−2πri cos(Θr−Θi)/λ.
In a coordinate system 410 of
The steerable antenna device, at which the elements 416a-c are disposed, may add an adjustable phase delay, αi, to each of the elements 416a-c, to generate, through interference, an array factor corresponding to any given direction. The array factor, AF, for a given (by Θr) radiation direction may be determined as:
AF(Θr)=Σi=1Nej(ϕ
where j=√{square root over (−1)}. The magnitude of the array factor, |AF(Θr)|, determines gain as a function of direction, i.e., a radiation pattern, of an array of isotropically radiating antenna elements (e.g., elements 416a-c implemented as monopole or dipole antennas).
As described above, the steerable antenna device may select added delays αi from an array of predetermined delays (e.g., using MUXs and delay lines). In some implementations, the delays may come from a set of M=2N possible delays, where N is the number of bits required to select a delay using a MUX and may be referred to as a resolution of delay selection. The value of N may be 1, 2, 3, 4, 5, 6 or any other suitable integer. In some implementations the number of different predetermined phases for each of antenna element may be an integer not represented by a power of 2, the number of possible phases may be 3, 5, 6, 7, 9, or any other suitable integer. The device may include appropriate switches, such as single-pole triple throw in selecting among possible phases. Still in other implementations the added phase may be continuously tunable.
An antenna device (e.g., the device 100) may use a controller (e.g., the controller 160) to compute a suitable additive phase for each of the elements in the array. In some implementations, the controller may choose the phases to maximize the array factor in a particular direction. Additionally or alternatively, the controller may compute the phases to minimize the array factor in a particular direction. The controller may compute optimal phases and then round each phase to the nearest available phase from the predetermined set. The controller may change the phases of the antenna elements at a suitable rate to steer or reconfigure the radiation pattern of the antenna device. The rate may be determined by switching delays of switches implementing the MUXs. The delays may be 0.1, 0.2, 0.5, 1, 2, 5, 10, 20, 50, 100, 200, 500, 1000, 2000, 5000 ns or any other suitable switching delays. It should be noted that the antenna radiation pattern need not switch at the maximum rate of the switches and may be held constant for any suitable length of time (e.g., from fractions of a nanosecond to hours or even days).
In some implementations, configuring an array for radiation in multiple directions may include partitioning an array by assigning one direction to some antenna elements and another direction to other antenna elements. The partitioning may be according to regions (e.g., sectors of a circle), random assignments of antenna elements to different directions, or following any other suitable algorithm. In other implementations, the phases of all antenna elements may be simultaneously optimized to achieve a desired radiation pattern. The optimization may follow a gradient descent or any other suitable algorithm. Furthermore, an initial point of the optimization may be based on the phases obtained from a partitioned array implementation of a multi-directional radiation pattern.
In some implementations the substrates 810a,b may both be conductive. In other implementations, one or more of the substrates may be constructed from dielectric materials. Generally, a conductive surface may act as a ground plane, i.e., have a nominally zero voltage difference with respect to other ground points. At least a portion of the surface of one of the substrates 810a,b may thus act as a primary ground plane with a flat circular shape. In some implementations, at least a portion of the surface of the other substrate may act as a secondary ground plane. Alternatively, a conductive surface may be floating, i.e., the voltage may be allowed to vary in response to, for example, induced currents. Still alternatively, at least some points on a conductive surface may be connected to a fixed voltage, different from the ground. The two conductive surfaces disposed at the substrates 810a,b may, in a sense, form a finite parallel plate waveguide. The conductive surfaces, and, therefore, the parallel plate waveguide may have rotational symmetry around a center point.
More generally, two conductive surfaces (e.g., of substrates) sandwiching (disposed at the ends of) antenna elements, need not be flat. For example, they may be in a shape of a dome, i.e., a surface of revolution produced by rotating a continuous function around an axis of rotation. Such surface-generating functions are illustrated as cross-sections of surfaces in
The electromagnetic fields radiated by the antenna elements may be configured to couple to modes (e.g., with field distributions illustrated with dashed lines) of the curved finite parallel plate waveguides formed by the surface pairs 1010a,b and 1040a,b. The guided modes may then diffract outside of the parallel plate waveguides, with the peak of the radiation pattern along parallel to the waveguide direction at the edge of the domes, as illustrated by arrows 160a-d.
The devices 1100a-c may be configured to transmit and or receive at the same frequency or, in some implementations, at different frequencies. Furthermore each of the devices 1100a-c may be configured to transmit in a different direction.
Each of the nodes in
It should be noted that the configurations in
At block 1302, the controller 160 directs a signal at a certain operating wavelength from a primary feed (e.g., element 112) to antenna elements (e.g., elements 130) disposed on a substrate (e.g., element 110), along a network of antenna feed traces (e.g., elements 120). The antenna elements can form directionally-disordered, quasi-uniform two-dimensional array.
Next, at block 1304, the controller 160 obtains a pointing direction of the steerable antenna array. In some implementations, the device 100 operates in a wireless mesh network, and the controller 160 dynamically determines the pointing direction and/or beam width (i.e., angular extent of a lobe of a radiation pattern) in view of a routing decision for a data packet.
At block 1306, the controller 160 computes respective phase delays for the antenna elements, so as to generate an appropriate radiation pattern (e.g., a pattern with a main lobe aligned with the pointing direction of the antenna device). In some implementations, each phase delay is selected from a limited set (e.g., four values, eight values, 16 values) of possible phase delays, in view of the implementation of the antenna array.
At block 1308, the controller 160 can use multiplexers (e.g., elements 140) to select a phase delay from the corresponding set and apply, to each of the antenna elements, the corresponding phase delay. In the case of multi-wavelength operation of the antenna device, the controller 160 may select a corresponding time delay from the time delays corresponding to the alternative paths selected by the multiplexer.
