See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/3010684

Beamformer architectures for active phased-array radar antennas

Article in IEEE Transactions on Antennas and Propagation · April 1999

DOI: 10.1109/8.768777 · Source: IEEE Xplore

CITATIONS READS
62 4,046

2 authors, including:

Eric Louis Holzman

39 PUBLICATIONS 380 CITATIONS

SEE PROFILE

All content following this page was uploaded by Eric Louis Holzman on 25 March 2016.

The user has requested enhancement of the downloaded file.

432 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 47, NO. 3, MARCH 1999

Beamformer Architectures for Active

Phased-Array Radar Antennas
Ashok K. Agrawal, Senior Member, IEEE, and Eric L. Holzman, Senior Member, IEEE

Abstract— In active phased-array antennas, the transmit and stripline. Manifold insertion loss may range from 2 to 3 dB
receive functions are distributed at the antenna aperture using for an all waveguide to as high as 8 dB for an all-stripline
transmit and receive (T/R) modules. The use of T/R modules beamformer. For wide-band operation, stripline medium is
provides a significant improvement in antenna performance and
flexibility in the choice of array architectures. In this paper, used. A general discussion of beamforming feeds for various
we present a review of various beamformer architectures for type of antennas is provided in [2] and an overview of array
active phased-array antennas. This review is limited to corporate- architectures is provided in [3], [4]. In this paper, we focus on
fed active phased-array antennas for radar applications, as the the beamformer architectures for active phased-array antennas.
general discussion to include all phased-array types is too broad High-performance radars use pulse-doppler detection modes
for any one presentation. Beamformer architectures for narrow
and wide bandwidth arrays, including the choice of applying to separate small signals from large levels of clutter received
an amplitude taper in the T/R module or beamformer network [1]. The system dynamic range from the T/R module through
are discussed. Beamformer architectures that increase antenna the A/D converter has to be large enough to handle the clutter
reliability are also presented. return and small target signal. Therefore, the receive channel
Index Terms—Active arrays, phased arrays. architecture needs to consider the entire receiver chain. The
gains of various components in the receive chain must be
I. INTRODUCTION set to meet the system requirements. The transmit channel
architecture is also affected by system desires to radiate a

T HE requirements for target detection radars operating

in complex threat environments have driven system de-
signers to use active phased-array antennas. Transmit/receive
transmit beam that has a sidelobe structure different from
the receive beam sidelobes. Since very low sidelobes are
not critical for the transmit beam, a uniform distribution is
(T/R) modules are key elements in any active aperture system.
typically used that results in higher gain in transmit than
Active phased-array antennas use a T/R module for each
receive and increased radar range. This may necessitate the
radiating element. The transmit power is distributed at the
use of separate transmit and receive manifolds in order to
aperture, eliminating the central high-power transmitter needed
preserve low receive sidelobes and high transmit gain.
in passive phased-array antennas.
For active phased-array antennas, the receive amplitude
Since the RF power and low-noise amplification in an active
taper can be generated either by the T/R module or receive
phased array are distributed at the aperture, the effect of losses
beamformer. The decision to provide the amplitude taper in
in the transmit and receive beamforming networks is greatly
the T/R module or receive beamformer depends on the receive
reduced. Sufficient transmit gain in the T/R module allows a
noise figure, input third-order intercept (TOI) and dynamic
very low RF input power to the antenna. In receive, a high-
range requirements. The portion of the receiver chain behind
gain low-noise amplifier (LNA) reduces the noise contribution
the array manifold, which processes the combined receive
from secondary stages of the T/R modules and beamformers.
output power of all the T/R modules needs to have high
By reducing the impact of transmit and receive beamformer
dynamic range. Its noise figure impacts the overall system
losses, large active phased-array antennas provide increased
noise figure. System architecture tradeoffs of input third-order
radar sensitivity in the range of 8–12 dB compared to passive
intercept and noise figure may or may not allow arbitrarily
arrays. In addition, T/R modules can provide wider bandwidth
high gain in the T/R module to swamp poor noise figure at
and higher average transmit power as compared to passive
the receiver [5].
phased arrays with central tube transmitters.
We first review beamformer architectures for passive
Radar systems often use monopulse techniques to derive
phased-array antennas used in radar applications in Section II.
angle tracking information from a single-echo pulse, requir-
In passive phased arrays, the transmit power from a single
ing a separate receive beamformer for each of the three
transmitter is distributed to the array elements through a
monopulse beams, viz., sum , elevation difference (EL
passive beamformer network. Separate transmit and receive
DELTA), and azimuth difference (AZ DELTA) [1]. Corporate
beamformers are used for the transmit and receive beams. The
beamformers are generally preferred in phased arrays that need
amplitude taper to provide low receive sidelobes is provided in
wide instantaneous bandwidth and good impedance match.
the receive beamforming network. The beamforming networks
The beamformers can be constructed either in waveguide or
for passive phased arrays form the basis for the discussion of
Manuscript received May 29, 1997; revised June 15, 1998. beamformer architectures for active phased-array antennas.
The authors are with the Lockheed Martin Government Electronic Systems,
Moorestown, NJ 08057 USA. If the receive amplitude taper is applied in the T/R module
Publisher Item Identifier S 0018-926X(99)04423-3. with either a variable gain amplifier (VGA) or attenuator, the
0018–926X/99$10.00  1999 IEEE
AGRAWAL AND HOLZMAN: BEAMFORMER ARCHITECTURES FOR ACTIVE PHASED-ARRAY RADAR ANTENNAS 433

Fig. 1. Receive beamformer architecture of a passive phased-array antenna.

receive beamformer architecture can be greatly simplified [6], II. BEAMFORMER NETWORKS FOR
[7]. This beamformer architecture is discussed in Section III. PASSIVE PHASED-ARRAY ANTENNAS
In Section III, a receive beamformer architecture with separate Radar systems often use monopulse techniques to derive
amplitude and phase control for each of the monopulse beams angle-tracking information from a single echo pulse. This
is presented. This architecture results in a very simple, low- is accomplished by generating two or more, usually three,
cost beamformer network where amplitude and phase errors
antenna beams [1], so that the simultaneously received echoes
in the beamformer network are not critical, as these errors can
from the multiple beams can be compared. A phased-array
be calibrated in the T/R module for each of the monopulse
antenna that generates three simultaneous beams in order to
beams independently. Therefore, the sidelobe performance in
support monopulse operation requires a separate beamformer
each of the three monopulse beams can be optimized.
for each of the three beams.
In Section IV, a tradeoff between two receive beamformer
High-performance phased-array monopulse antennas in cur-
architectures for antenna performance in terms of noise figure,
rent use have several thousand radiating elements. A beam-
input TOI, and dynamic range is presented. In the first beam-
former for use with such an antenna array must, therefore,
former architecture, the receive amplitude taper is applied in
include several thousand inputs and may require different
the T/R module, and in the second beamformer architecture,
the receive amplitude taper is applied in the beamformer. Since amplitude weighting for signals applied to each input. The am-
the transmit and receive amplitude tapers are different, the plitude weighting is provided in order to control the sidelobe
second architecture requires separate beamformers for the sum performance of the beams.
receive and transmit beams. On the other hand, a common A simplified block diagram of a typical receive beamformer
transmit and sum receive beamformer can be used for the first for a monopulse phased-array radar antenna is shown in Fig. 1.
architecture. Signals from elements in each column equidistant from the
Since active phased-array antennas contain large numbers of array centerline are fed to the column beamformers of sum
active components, the antenna performance degrades grace- and difference hybrids where the two vector signals are added
fully as components fail over time. In addition to T/R modules, and subtracted. The sum and difference outputs are separately
active phased arrays may contain large numbers of power weighted and combined to form one elevation sum and one el-
supplies or dc to dc converters; each dc to dc converter feeds evation difference signal per column of radiators. The column
a group of T/R modules. The array architectures that provide sum and difference signals are combined symmetrically about
higher reliability are discussed in Section V. By employing the array centerline in the horizontal beamformers to form the
fault-tolerance in the design of an antenna architecture, the three monopulse signals. To obtain low sidelobes in the receive
antenna reliability can be increased so that replacement of mode, a Taylor weighting is applied in the sum beamformers.
failed components can be avoided for an extended period To obtain Taylor weighting in two dimensions, the weighting
of time. in the column and horizontal beamformers is such that their
In Section VI, the beamforming networks for wide-band multiplication results in the desired two-dimensional weighting
active phased-array antennas are presented. To increase the function. Similarly, a Bayliss weighting is applied in the
instantaneous bandwidth of an array, the aperture is divided beamformers to provide low sidelobes in the difference beams.
into subarrays, and a time delay unit is used behind each A typical beamformer requires a large number of coupler
subarray to provide true-time delay for phase shift. designs that are selected to provide the necessary amplitude
434 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 47, NO. 3, MARCH 1999

tapers with the designs modified after test to compensate for a transmit/receive module for each radiating element. Each
coupler interactions and manufacturing errors. The amplitude T/R module contains a transmit chain consisting of power
and phase errors across the aperture contribute to the sidelobe amplifiers to provide power to each radiating element and
level for each of the simultaneous beams. The random error a phase shifter and a receive chain consisting of a low
contribution to sidelobe level is largely set by design, manu- noise amplifier, phase shifter, and amplitude control. Since
facturing, and testing technology. To obtain low peak and rms the transmit power is distributed at the aperture, there is no
sidelobes, phase and amplitude errors must be reduced and central high-power transmitter. Placement of power amplifiers
correlation of errors between elements must be minimized. and low-noise amplifiers at the aperture eliminates the transmit
The correlated errors can be minimized by measuring the and receive beamformer losses, resulting in increased radar
phase errors in the feed network to each element of the array sensitivity of 8–12 dB for large phased arrays. In addition, the
and using the phase shifters to compensate for that error. solid state technology used in the T/R modules can provide
This approach reduces the effective element phase error to wider bandwidth and higher average power by operating at a
that of phase shifter quantization and measurement errors. higher duty than tube technology.
The residual error after correction is a function of bandwidth, For active phased-array antennas, the receive amplitude
random errors in the phase shifters, and tracking and alignment taper can be provided either in the T/R module or the beam-
errors between beamformer channels. forming network. When the amplitude taper is generated in
The effects of the resulting amplitude and phase errors the modules, the resulting beamformers are simple uniform
depend on where they occur in the combining network. The weight beamformers. However, the amplitude taper can also
elements that share a given error act together as a subarray. be provided in the beamformers as in passive arrays. In
The effect of the error has gain due to the directivity of the this section, we will discuss beamformer architectures with
concerted action of the elements within the correlation interval. amplitude taper provided in the T/R modules. In the next
Therefore, correlated errors are multiplied by the number of section, the two beamformer architectures will be compared
elements affected by the error. for their dynamic range and noise figure performance.
While the phase errors in the beamforming network for the Since there is no central high-power transmitter in active
sum beam can be corrected by the phase shifters, such errors arrays, the transmit beamformer can be constructed using low-
cannot be simultaneously corrected for the difference beams power transmission lines fabricated from printed circuit boards
with only a single-phase shifter per element. Since the phase and can also be shared with the receive sum beamformer.
errors for the difference beamformers cannot be corrected Fig. 1 illustrated the use of column sum-and-difference
simultaneously for monopulse operation, the phase errors for hybrids and combiners to form column-sum and column-
the difference beamformers in passive phased arrays are larger difference signals and further used sum-and-difference hybrids
than the sum beamformer and the achievable sidelobe level for
and combiners to form the sum , AZ DELTA, and EL
the difference beam is generally higher than for the sum beam.
DELTA beams. Fig. 2 shows a simplified block diagram of a
The interconnections within the beamformer, which lie
monopulse active phased-array antenna architecture in which
between the radiating element and the common output port of
the beamformer, must be carefully phase controlled so as not to each T/R module associated with the antenna element has three
introduce errors. When three such beamformers are used—one receive outputs. The architecture in Fig. 2 consists of three
each to produce a sum , azimuth difference (AZ DELTA), beamformers—one each for sum, elevation delta, and azimuth-
and elevation difference (EL DELTA) beam—the insertion delta receive beams. The transmit beamformer and sum receive
phase of the sum and delta beamformers must track each beamformers are common and these two functions are shared
other closely so that the desired performance for monopulse by this beamformer using a switch.
operation is achieved under all operating conditions. For Fig. 3 is a simplified block diagram of a T/R module with
example, phase errors between sum and delta beamformers multiple outputs [7]–[9]. The transmit chain consists of a phase
can result in beam pointing angle error. shifter, a low noise amplifier, a variable gain amplifier (VGA),
The complexity of the beamforming arrangement of Fig. 1 a driver amplifier, and a power amplifier. The power amplifier
is apparent. Additional complexity arises because of the am- may consist of several power amplifiers, typically two or four,
plitude weighting of the signals relative to each other in where the output power of these amplifiers is combined to
each column and from column to column in order to achieve obtain required output power. Broadcasting of harmonics of
the appropriate sidelobe level for both elevation and azimuth the transmit signal is reduced by a harmonic filter. The receive
beams. Even if phase shifters are set correctly, assuming equal chain consists of an LNA proceeded by an amplitude limiter.
phase signals arrive at the phase shifters, cumulative phase A circulator is used to provide the duplexing function and
errors through the combiners and hybrid arrays may adversely isolation between transmit and receive signals. The output of
affect performance. In this regard, it should be noted that the the LNA is divided equally into three signals using a three-
actual physical lengths of interconnecting cables must nearly way power divider. These three signals go through a phase
be equal for wide bandwidth signals. shifter, low-noise amplifiers, and a VGA, and the outputs are
connected to the three monopulse beamformers: sum, elevation
III. BEAMFORMER NETWORKS FOR delta, and azimuth delta. In addition, the sum receive channel
ACTIVE PHASED-ARRAY ANTENNAS is also shared with the transmit beamformer using a double-
In active arrays, both transmit and receive functions are pole double-throw (DPDT) transfer switch as shown in Fig. 3.
placed at the aperture. This function is achieved by using This module architecture provides optimum performance with
AGRAWAL AND HOLZMAN: BEAMFORMER ARCHITECTURES FOR ACTIVE PHASED-ARRAY RADAR ANTENNAS 435

Fig. 2. Beamformer architecture of an active phased-array antenna with three independent receive beams.

Fig. 4. Block diagram of a T/R module with two receive outputs.

Fig. 3. Block diagram of a T/R module with three receive outputs. El delta beams. For the difference beams, an extra 180
respect to module noise figure, TOI, and dynamic range [10]. phase shift can be added to the elemental outputs from
There are several advantages to this architecture compared to the selected half arrays. The VGA’s can be adjusted not
a more conventional common phase-shifter architecture: 1) it only to correct amplitude errors in the beamformer, but can
minimizes the number of components since gain stages are provide the amplitude taper required to reduce the sidelobe
shared between transmit and receive; 2) the receive dynamic level. The beamformers can be designed using standard-
range is maximized since gain stages are distributed before ized coupling values rather than coupling values that are
and after the lossy phase shifter; 3) the VGA can be used for customized to provide the desired amplitude weighting. At
programmable transmit aperture as well as receive; and 4) the each operating frequency of the array, the phase shifters and
shared components can be designed as an integrated chip using variable gain amplifiers can be programmed with the amplitude
a versatile high-yield monolithic microwave integrated circuit and phase required to correct the errors that occur at that
(MMIC) process without degrading module performance. frequency.
The beamformer architecture of Fig. 2 has the salient ad- Since, for the beamformer architecture in Fig. 2, the
vantage that the beamformer designs are not critical and residual phase and amplitude errors in all three beamformers
need not have connections made in matched pairs equidistant can be corrected using VGA’s and phase shifters in the T/R
from the center lines of the array. Thus, there may be a module, the achievable sidelobe performance for all three
great saving in cable length and weight and a reduction monopulse beams is the same, a significant improvement
in criticality of the phase through various paths. The phase over the passive architecture of Fig. 1 where the achievable
shifts of the phase shifters of each T/R module can be sidelobe level for the difference beam is higher than for the
adjusted to optimize the phase shift through a particular path sum beam. Accordingly, the limiting factor in the sidelobe
connected to the output port of that module. The perfor- performance of such an array is only the amplitude and phase
mance in either transmission or reception can, therefore, be performance of T/R modules together with the long-term
optimized separately for each of the sum, AZ delta, and stability of the components.
436 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 47, NO. 3, MARCH 1999

Fig. 5. Beamformer architecture of an active phased-array antenna with two independent receive beams.

A disadvantage of the arrangement described in Figs. 2 and power since the final power amplification takes place in the
3 lies in the number of phase shifters and VGA’s and com- T/R modules at the antenna elements.
plex controls that may therefore be required. A compromise
IV. IMPACT OF BEAMFORMER ARCHITECTURE
between the beamformer architecture of Fig. 1 with a single-
ON SYSTEM-NOISE TEMPERATURE
phase shifter for multiple beamformers and the beamformer
architecture of Fig. 2 with a single-phase shifter for each In this section, two common receive beamformer architec-
beamformer of the array, may be the set of two phase tures, used in active phased arrays to generate low sidelobe
shifters for three beamformers. Fig. 4 illustrates a T/R module sum patterns, are compared for architectural complexity and
architecture with two phase shifters resulting in two inde- their performance in terms of noise figure, input TOI, and
pendent receive channels. Fig. 4 differs from Fig. 3 in that dynamic range. In the first beamformer architecture, the re-
the low-noise amplifier has its output coupled to a two- ceive amplitude taper is applied in the T/R module, and in
way power divider. Fig. 5 is a simplified block diagram of the second beamformer architecture, the receive amplitude
an array antenna using the T/R module of Fig. 4. One of taper is provided in the beamformer [11]. Since the transmit
the outputs of the T/R module is coupled to an elevation- and receive amplitude tapers are different in the second
architecture, separate beamformers are required for the sum
difference beamformer illustrated in Fig. 4. This beamformer
receive and transmit beams. On the other hand, for the first
is similar to the elevation-difference beamformer of Fig. 2,
architecture, a common transmit and sum receive beamformer
which combines the elevation signals from T/R modules
can be used.
in a column to provide a column elevation difference sig-
Figs. 6 and 7 show two popular corporate beamforming
nal.
architectures; they differ in how they form the sum channel
The uppermost or sum outputs from each T/R module receive weighting. For comparison of these two architectures,
in a column are connected to the inputs of a sum-column only sum beamformers are considered. In Fig. 6, the weighting
beamformer. From the single output of the column beamformer is implemented actively with the T/R modules, while in Fig. 7,
both sum and azimuth beam must be generated. The desired nonuniform (unequal) combiners are employed within the
pair of beams are generated from the output of the column beamformer to passively create the low sidelobe weighting. In
beamformers by coupling these outputs to a further array of general, the passive weighted architecture provides the lowest
T/R modules. Each T/R module receives a signal from the cor- system noise figure and highest output intercept point, while
responding column in the receive mode and divides the signal the active weighted architecture is the easiest to construct
into two signals and applies them to a sum beamformer and an and produce. It is beneficent to gain more insight into the
azimuth beamformer. This arrangement has the advantage of differences between the two architectures. Toward this end,
relative simplicity compared with the arrangement of Fig. 2, a detailed comparison of these two beamformers in terms of
while maintaining the advantage of substantial control over their architectural complexity and receive performance (noise
the three antenna beams. The column T/R module in Fig. 4 is figure and output intercept point) is presented.
similar to the T/R module in Fig. 3, but has only two receive The active-weighted architecture of Fig. 6 has the advantage
channels and the transmit channels have a lower transmit out of using the same beamformer for transmit and receive. In the
AGRAWAL AND HOLZMAN: BEAMFORMER ARCHITECTURES FOR ACTIVE PHASED-ARRAY RADAR ANTENNAS 437

Fig. 6. Active-weighted beamformer architecture.

Fig. 7. Passive-weighted beamformer architecture.

Fig. 6, an -element array of T/R modules is backed by a by using nonuniform beamformers. For a rectangular array
combiner for each column of the array and the columns are with a separable amplitude distribution function, the vertical
summed by a single horizontal combiner to give the sum- combiners in each column are identical; however, for a circular
channel receive output. On transmit, the T/R modules are set array, each column will require a different combiner.
to maximum gain and output power and, on receive, the low Although the passive-weighted array is more complex, it
sidelobe taper is applied by varying the gain of the receive has a lower receive noise figure and higher intercept point
VGA’s in the T/R modules. than the active-weighted array. A low antenna noise figure is
The passive-weighted architecture of Fig. 7 is more com- critical if a radar is to detect small returns, while a high antenna
plex. It requires separate beamformers for transmit and receive TOI is essential when there are big scatterers such as clutter
operation. The transmit combiners are uniform. However, on present at the same time. Lee [12] has analyzed the noise
receive, all T/R module VGA’s are set to maximum gain and performance of active and passive-weighted architectures and
the low sidelobe taper is provided in the receiver combiners Holzman [13] has derived equations for the output TOI for
438 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 47, NO. 3, MARCH 1999

TABLE I TABLE II
RECEIVE PERFORMANCE PARAMETERS FOR ACTIVE- AND PASSIVE-WEIGHTED RECEIVE PERFORMANCE PARAMETERS FOR ACTIVE- AND PASSIVE-WEIGHTED
PHASED ARRAYS NORMALIZED TO THOSE OF AN ARRAY WITH UNIFORM PHASED ARRAYS NORMALIZED TO THOSE OF AN ARRAY WITH UNIFORM
ILLUMINATION. M ELEMENTS; wi IS THE VOLTAGE AMPLITUDE OF THE ith 0
ILLUMIANTION. 16 384 ELEMENTS, 40 dB ILLUMINATION [11]
ELEMENT; Fi IS THE NOISE FIGURE OF THE ith T/R MODULE (INCLUDING
RADIATING ELEMENT); Fmin IS THE NOISE FIGURE OF THE T/R MODULE
OPERATING WITH MAXIMUM GAIN; I0 IS THE T/R MODULE OUTPUT
THIRD-ORDER INTERCEPT (TOI) FOR MAXIMUM GAIN; Ii;0 IS THE OUTPUT TOI
OF THE ith T/R MODULE; " TAPER IS THE APERTURE EFFICIENCY [11]

As a vehicle to compare the two architectures, we use a

rectangular array of 16 384 elements arranged on a triangular
lattice with 256 rows and 256 columns [11]. If we apply a
linear–linear 40-dB Taylor taper to the array, we obtain an
aperture efficiency of 2.2 dB. Table II compares the resulting
active and passive-weighted performance.
We see that the passive-weighted array has 7.5 dB more
signal and noise gain than the active architecture. If the passive
array has 15-dB noise gain and a 5-dB input noise figure and
both array architectures. Table I summarizes their equations is followed by a receiver with a 5-dB input noise figure, the
for beamformer signal gain, noise gain, input noise figure, and system noise figure is 5.09 dB. An active antenna with the
output TOI for both phased-array architectures [11]. Signal same T/R module would have only 7.5-dB total noise gain
gain includes all passive combining gain and active T/R and a system noise figure of 5.50 dB.
module gain plus any losses in the beamformer. The noise The choice of providing amplitude taper in the beamformers
gain excludes the gain of the combiners. The input noise or the T/R modules will depend on the required system
figure is the noise figure referenced to the input of the T/R noise performance, beamformer complexity, and cost. The
module and the output TOI is referenced to the output of the additional beamformer required for the passive-weighted array
sum channel. The equations are normalized to the performance will impact array packaging as it requires additional volume
of a uniformly illuminated array, i.e., an array with uniform and will increase array cost ( 10%).
combiners and all T/R modules with maximum receive gain.
If we assume that reactive combiners are used in the passive- V. BEAMFORMER ARCHITECTURES FOR HIGH RELIABILITY
weighted array, the noise gain and input noise figure will be Since active phased-array antennas contain large number of
the same as those of the uniform array. The TOI and signal active components, the antenna performance degrades grace-
gain are reduced by the taper efficiency, typically 1–2 dB. The fully as components fail over time. In addition to the T/R
T/R modules are operated at maximum gain, so no reduction modules, the active phased arrays may contain large number
in noise gain occurs. On the other hand, for active-weighted of power supplies or dc to dc converters; each dc to dc
array, the T/R module gain varies with position in the array. converter feeds a group of T/R modules. A block diagram of
For tapered aperture illuminations, the signal and noise gains the antenna with these active components is shown in Fig. 8.
of the active-weighted architecture can easily be 6–8 dB less In Fig. 8, a group of T/R modules share a common control
than those of the passive-weighted array. The noise figures of module that houses voltage regulators, drain switches, digital
the two arrays are not much different, since the T/R module’s controls, and memory chips. A distributed power system is
weighting is set by a VGA or attenuator at the back of the generally chosen to increase system clutter improvement factor
module’s receive chain. For a module at the array edge where (CIF) [14] and reliability [15]. The choice of beamformer
the attenuation is severe, the module noise figure is higher; architecture and packaging of components will affect the
however, that module’s contribution to the system noise figure system reliability. Consequently, the design of active phased
is small because its noise figure is multiplied by the square of arrays often involves a tradeoff between production cost, life-
its voltage weight. The effect of the active-weighted array’s cycle cost, and performance.
lower gain can be significant in a radar system when we Strategies for designing active phased-array antenna archi-
consider the stages that follow the phased array. For instance, tecture with high reliability are discussed in this section. It
if a receiver follows the antenna, the contribution of its noise is shown how to consider fault tolerance in the design of the
figure is reduced by the noise gain of the antenna, viz., antenna architecture, so that replacement of failed components
can be avoided for an extended period of time. The reliability
of an antenna is described by its mean-time-between-failures
(MTBF)—failures that cause the antenna performance to fall
where all quantities are referenced at the input. This equation below the acceptable level. We discuss the design of active
applies to both array architectures. The passive weighted an- phased-array architectures for maximizing antenna MTBF. The
tenna will have the lower overall system noise figure because antenna MTBF is generally defined in terms of a specified
its noise gain is not reduced by the amplitude taper. degradation in receive peak and/or rms sidelobes from the
AGRAWAL AND HOLZMAN: BEAMFORMER ARCHITECTURES FOR ACTIVE PHASED-ARRAY RADAR ANTENNAS 439

TABLE III
ALLOWED FRACTIONS OF ELEMENT FAILURES FOR 3- AND 6-dB INCREASE IN
THE PEAK SIDELOBE LEVEL AS A FUNCTION OF CLUSTER SIZE FOR AN
8000-ELEMENT CIRCULAR ARRAY WITH TAYLOR 40-dB AMPLITUDE TAPER
AND TRIANGULAR LATTICE, HALF-WAVELENGTH ELEMENT SPACING.
ELEMENTS ARE ADJACENT AND WITHIN THE SAME COLUMN

TABLE IV
CALCULATION OF ANTENNA MTBF FOR A 3-dB RISE
IN PEAK SIDELOBES FOR AN 8000-ELEMENT ARRAY

Fig. 8. Reliability block diagram of the antenna. We have curve fit the data in Table III to develop a single
equation describing sidelobe level increase as a function
nominal level [15]. For a phased array that is deployed in of array size. We assume that the number of failures is
the field for a length of time, it is desirable to have the proportional to array size for a given cluster size. As the
antenna MTBF exceed the deployment time, so that the failed array size decreases, a particular cluster failure causes more
components can be replaced at the depot and the need for spare sidelobe level increase in proportion to its area. The sidelobe
parts in the field is minimized. level (SLL) increase in dB as a function of cluster size, array
As antenna components fail randomly, the sidelobe perfor- radius, and frequency is given by the following equation:
mance degrades. If the components fail randomly in single
clusters, a fairly large number of components can be allowed SLL Increase
to fail before the antenna sidelobe performance falls below where is the element cluster size, is the frequency in Ghz,
a specified value. However, if the components fail in large and is the radius of the array in feet.
clusters, the number of allowed failures for a given sidelobe For example, consider the 8000-element array. Twelve two-
degradation is much smaller. Therefore, the antenna architec- element cluster failures causes 0.46-dB sidelobe level increase.
tures which minimize failure of big clusters of elements will If we lose three four-element clusters of elements, we also get
increase the antenna MTBF. approximately 0.46 dB increase in sidelobe level. Now, halve
Based on this definition, we can calculate the MTBF of a the array radius so the array size reduces to 2000 elements.
large active phased-array antenna given the types of compo- For this smaller array, three two-element clusters failure also
nents, their quantities, and their MTBF’s. In addition, we must causes approximately 0.46-dB sidelobe increase. As the SLL
know the antenna performance degradation per failure for each increases, we would expect failures to have less and less
component-type and the antenna’s minimum acceptable level impact.
of performance. Consider peak sidelobe level. Table III shows In this section, we provide guidelines for selecting an active
the number of element failures allowed for clusters of one, array architecture that maximizes the antenna MTBF. We
two, four, and eight for the peak sidelobe level increase of begin with the phased-array antenna architecture in Fig. 8 and
3 and 6 dB for an 8000-element circular array with a 40-dB assume that each control module and each power supply drives
Taylor sidelobe taper. The element lattice is triangular with eight T/R modules. From the maximum allowable number of
about half-wavelength spacing. failures given in Table III for an 8000 element array, we can
The data in Table III was obtained by calculating the calculate the antenna MTBF as shown in Table IV for a 3-dB
antenna patterns for an aperture with different size clusters rise in peak sidelobes.
of failed elements removed in a random fashion. The peak The antenna MTBF calculation in Table IV shows antenna
sidelobe degradation was determined for many different trials MTBF due to failures of single component types and cu-
for each cluster size and the numbers shown in Table III mulative antenna MTBF. The antenna MTBF due to single
represent the average. These simulations were done for a large component is obtained by multiplying the failure rate by the
array size (about 8000 elements). The results in Table III could number of element failures allowed. From the third row, we
be used for other arrays of comparable size or larger. As the see that the MTBF of the antenna is only 162 h, i.e., the
size of the array is increased, the fractions in Table III will antenna must be shut down approximately once a week for
increase and as the size of the array is decreased, the fractions repair of failed components. Clearly, if we only considered
in Table III will decrease toward zero. Note that the data in T/R modules, the antenna MTBF would be 6400 h, so the T/R
Table III is independent of frequency. module cluster size and MTBF are acceptable. However, the
440 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 47, NO. 3, MARCH 1999

Fig. 9. Active-weighted beamformer architecture with subarrays for time-delay steering.

TABLE V of repairing or replacing failed elements in the field. If the

NO CLUSTER DESIGN. CALCULATION OF antenna MTBF is greater than the deployment period, the
ANTENNA MTBF FOR A 3-dB RISE IN PEAK
requirement for maintenance personnel can be minimized.

VI. BEAMFORMER NETWORKS FOR WIDE-BAND

ACTIVE PHASED-ARRAY ANTENNAS
An array that is steered by phase shift, rather than time
MTBF due to the power supplies alone is only 250 h, and for delay, will scan as frequency is changed. The instantaneous
the control modules it is only 500 h; these two components bandwidth of a phased-array antenna can be increased by
are determining the antenna MTBF. dividing the aperture into subarrays and placing a time-
We can raise the antenna MTBF if: 1) the MTBF of the delay network behind each subarray. The size of the subarray
individual components is increased; 2) the elements are driven determines the instantaneous bandwidth of the antenna. Band-
in smaller clusters; 3) we use redundancy in the design; or 4) width criteria for phased-array antennas in terms of loss of
we raise the maximum allowable level of acceptable sidelobe antenna gain and relative amplitude of grating lobes have been
increase. The best MTBF we can achieve using smaller developed for both narrow-band signal with frequency agility
clusters occurs when each element is driven by a single T/R and for instantaneous wide-band signals [16]. In this section,
module, control module, and power supply. Table V shows we present beamformer architectures for instantaneous wide
these calculations. We have increased the antenna MTBF to bandwidth active phased arrays.
1120 h. However, this improvement has come at a high cost Fig. 9 shows a typical wide-band beamformer architecture
since we have had to increase the number of components in the where the array is divided into subarrays with a time-delay-
antenna by a factor of two and a half. Therefore, the selection unit (TDU) steering each subarray. The TDU provides true
of antenna architecture and cluster size will depend on the time delay for each subarray and the T/R modules provide
tradeoff between antenna cost and MTBF. phase compensation for the quantized time steering. The size
Another way to increase the MTBF is to provide redun- of the subarray determines the instantaneous bandwidth of the
dancy in some components. From Table V, we note that the array. A rule of thumb for the bandwidth of the array was
antenna MTBF is significantly reduced by the power supplies. established in [16] as
Therefore, the antenna MTBF can be significantly improved Bandwidth Beamwidth (deg.) (for CW signals)
by providing redundancy in the power supplies. This can be Bandwidth Beamwidths (deg.)
achieved by combining power supplies and feeding larger
(for Pulsed signals).
clusters of T/R modules. The redundancy in power supplies
can be provided by an additional power supply in the paralleled The beamwidth in the above equations represents the
group. beamwidth of the subarray in degrees.
From the above discussion we note that the antenna man- The amplitude taper in the architecture of Fig. 9 is provided
ufacturing cost increases the antenna MTBF is increased. To in the T/R modules, so all the combiners are uniform and, thus,
increase the antenna MTBF, we had to increase the number of all the subarrays are identical. In Fig. 9, for a subarray of
components to minimize the number of large element cluster elements, elements in vertical columns are combined
failures due a single component failure. The life cycle cost using vertical combiners and the outputs of the -
of the antenna includes initial manufacturing cost and cost vertical combiners is combined with an horizontal
AGRAWAL AND HOLZMAN: BEAMFORMER ARCHITECTURES FOR ACTIVE PHASED-ARRAY RADAR ANTENNAS 441

Fig. 10. Passive-weighted beamformer architecture with subarrays for time-delay steering.

Fig. 11. Hybrid-weighted beamformer architecture with active-weighted subarray columns.

combiner to form the subarray input/output. The TDU provides nonuniform combiners so that each subarray has unique set
the common time delay to all the elements in the subarray. The of components. For an antenna with quadrantal symmetry, the
outputs of the TDU’s are further combined using vertical and number of unique subarrays can be reduced by a factor of four.
horizontal combiners to provide transmit input and receive out- Figs. 9 and 10 show beamformer architectures for the
put of the antenna. All the beamformers in Fig. 9 are uniform sum beam only. This architecture can be extended to three
and all TDU’s are identical. For large phased arrays, which monopulse beams by providing a TDU in each of the
may have hundreds of subarrays, one common set of parts monopulse beams.
is very desirable from the producibility and maintainability As discussed in Section IV, the active weighted architecture
prospective. has lower noise performance than the passive-weighted archi-
Fig. 10 shows a wide-band passive-weighted array archi- tecture. As a compromise between the more complex passive-
tecture. We need two beamformers, one for transmit and weighted architecture and lower performance active-weighted
one for receive. Further, we must use two sets of TDU’s architecture, we can combine the characteristics of both in
or diplex the transmit and receive signals through a single a hybrid architecture shown in Fig. 11. For a beamformer
set; either option is more complex than the active-weighted with multiple levels of combining, some combiners can be
architecture. In addition, the receive beamformer requires uniform, while others can be nonuniform. In that way, we
442 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 47, NO. 3, MARCH 1999

can reduce the number of unique nonuniform combiners in [11] E. Holzman and A. Agrawal, “A comparison of active phased array,
the antenna and have higher performance [11]. In Fig. 11, corporate beamforming networks,” in IEEE Int. Symp. Phased-Array
Technol., Boston, MA, Oct. 1996, pp. 429–434.
the vertical combiners and the horizontal combiner in the [12] J. J. Lee, “G/T and noise figure of active array antennas,” IEEE Trans.
subarray are uniform weighted and common to both transmit Antennas Propagat., vol. 41, pp. 241–244, Feb. 1993.
[13] E L. Holzman, “Intercept points of active phased array antennas,”
and receive paths. Partial amplitude weighting can be applied in IEEE Microwave Theory Tech. Soc. Int. Microwave Symp., San
in the modules in the receive mode. For a large phased Francisco, CA, June 1996, pp. 999–1002.
array, by making the first-level vertical combiners in the [14] E. Holzman and A. Agrawal, “Active phased array design for high
clutter improvement factor,” in IEEE Int. Symp. Phased-Array Technol.,
subarray uniform, we have reduced the number of nonuniform Boston, MA, Oct. 1996, pp. 44–47.
combiners with performance still significantly better than the [15] A. Agrawal and E. Holzman, “Active phased array design for high
active-weighted architecture. reliability,” IEEE Int. Symp. Phased-Array Technol., Boston, MA, Oct.
1996, pp. 159–162.
[16] J. Frank, “Bandwidth criteria for phased array antennas,” Phased Array
Antennas. Norwood, MA: Artech House, 1972.
VII. CONCLUSION
Active phased-array antennas provide significant improve-
ment in performance over their predecessor passive phased- Ashok K. Agrawal (S’76–M’79–SM’82) received
array antennas. In addition to the reduction in transmit and the M.S. and Ph.D. degrees in electrical engineering
from the University of New Mexico, Albuquerque,
receive losses, active arrays can provide wide instantaneous in 1976 and 1979, respectively.
bandwidth and high average radiated power. Active phased From 1976 to 1982, he worked as a Research
arrays offer many choices for beamformer architectures. This Scientist at Mission Research Corporation, and from
1982 to 1983 he was a Senior Research Engineer at
paper has examined the beamformer architectures for active Dikewood Corporation, both in Albuquerque, NM.
phased-array radar antennas for narrow and wide bandwidth From 1982 to 1983 he was also an adjunct faculty
and for high reliability. The passive-weighted and active- member at the University of New Mexico, where
he taught graduate courses on antennas. In 1983
weighted architectures were compared for beamformer com- he joined the Government Electronic Systems Division of Lockheed Martin
plexity and antenna noise performance. The material presented Company (previously, RCA, GE, Martin Marietta) as a Principal Member
in this paper should provide sufficient information for an of Engineering Staff. Since then he has been involved in research and
development work on phased-array antennas. He has led the development
antenna design team to choose an optimum beamformer ar- of several passive and active phased-array antenna and T/R module projects.
chitecture for a given set of requirements. He holds five U.S. patents on phased-array antennas.
Dr. Agrawal was Vice Chairman of the APS/MTTS/EMC Albuquerque
Chapter from 1982 to 1983 and Chairman of the APS/MTTS Philadelphia
REFERENCES Chapter from 1986 to 1987. He was Chairman of the Benjamin Franklin
Symposium held in Cherry Hill, NJ, in 1987. He is a member of Tau Beta Pi.
[1] M. Skolnik, Introduction to Radar Systems. New York: McGraw-Hill,
1980.
[2] J. Ajioka and J. L. McFarland, “Beam-forming feeds,” in Antenna
Handbook. New York: Van Nostrand Reinhold, 1993, vol. III, ch. 19.
Eric L. Holzman (S’86–M’89–SM’95) was born
[3] R. J. Mailloux, “Antenna array architecture,” Proc. IEEE, vol. 80, Jan.
January 17, 1962, in Los Angeles, CA. He received
1992.
[4] , Phased Array Antenna Handbook. Boston, MA: Artech House, the B.S., M.S., and Ph.D. degrees from the Univer-
1994. sity of California at Los Angeles, all in electrical
[5] T. R. Turlington, F. E. Sacks, and J. W. Gipprich, “T/R module engineering, in 1984, 1987, and 1989, respectively.
architectural consideration for electronically steerable arrays,” in IEEE He worked at Hughes Missile System Company,
Microwave Theory Tech. Soc. Symp. Dig., 1992, pp. 1523–1526. Canoga Park, CA, from 1984 to 1993, where he was
[6] A. Agrawal and N. Landry, “Independent control of sum and difference involved in all stages of the complete conception-
patterns in active phased array antennas,” in IEEE Antennas Propagat. to-prototype design of electromagnetics portions of
Int. Symp., San Jose, CA, June 1989, pp. 392–395. microwave and millimeter-wave seekers. He is cur-
[7] , “Monopulse phased array antenna with plural transmit-receive rently a Principal Member of the Engineering Staff
monopulse phase shifters,” U.S. Patent 5 017 927, May 21, 1991. at Lockheed Martin Government Electronic Systems, Moorestown, NJ, where
[8] J. Komiak and A. Agrawal, “Design and performance of octave S/C- he is engaged in the design of advanced solid-state phased arrays. He is
band T/R modules for multifunction phased arrays,” IEEE Trans. the lead author of Solid-State Microwave Power Oscillator Design (Norwood,
Microwave Theory Tech., vol. 39, pp. 1955–1963, Dec. 1991. MA: Artech House, 1992) and author of over 15 papers on phased arrays,
[9] , “Octave S/C-band MMIC T/R modules for multifunction phased antennas, and microwave components. He holds four patents and has two
arrays,” in IEEE Microwave Theory Tech. Soc. Int. Microwave Symp., pending in the microwave field.
Boston, MA, June 1991, pp. 35–38. Dr. Holzman is Chairman of the Philadelphia chapter of the IEEE Anten-
[10] A. Agrawal, R. Clark, and J. Komiak, “T/R module architecture tradeoffs nas/Microwave Societies. He was a member of the Organizing Committee for
for active phased array antennas,” in IEEE Microwave Theory Tech. Soc. the Benjamin Franklin Symposium in 1995, 1996, and 1997. He is a member
Int. Microwave Symp., San Francisco, CA, June 1996, pp. 995–998. of Tau Beta Pi and Eta Kappa Nu.

View publication stats