STATE-OF-THE-ART AND TRENDS IN PHASED ARRAY RADAR

P. VAN GENDEREN
Delft University of Technology
Faculty of Information Technology and Systems
PO Box 5031, 2600 GA Delft, The Netherlands
E-mail: [email protected]
Also at: Hollandse Signaalapparaten BV
PO Box 42, 7550 GD Hengelo, The Netherlands
E-mail: [email protected]
The paper first introduces phased array configurations for radar and astronomy applications. In
radar various trends can be observed, on the one hand driven by cost issues and on the other hand
by increasing performance demands. The trends discussed, are the increasing bandwidth, the
improvement of the Power Added Efficiency of transmit/receive modules, the phase shifter
technology, alternative radiator designs, substitution of analog technology by digital technology
and the introduction of sparse arrays. The paper concludes by observing that in radar applications
phased arrays are now commercially viable, but that due to their complexity the cost will maintain
to be high.

Introduction

Principles of phased arrays have been applied in radar since World War II. Most of the advances in
theory and technology were achieved in the fifties and the sixties. However the phased arrays came
into operational use late in the sixties and the seventies. Two components were driving the
development at that time: fast phase shifters and computertechnology for phased array control. The
number of economies capable of bearing the cost of such radars was very low. While the technology
required for phased array radars was maturing in the eighties, in the current decade several companies
are capable of developing and manufacturing them. The paper discusses the trends in phased array
radar technology, combined with occasional side steps to the phased arrays for astronomy
applications.

Phased array configurations

Phased Array Radars appear in very different configurations.

Fig. 1 shows an example of one face of a phased array radar having four of these, installed in a large
anechoic chamber. It is the antenna part, including all transmit and receive electronics.

Figure 1: One face of the Active Phased Array Radar APAR

(Courtesy Hollandse Signaalapparaten)
A.B. Smolders and M.P. van Haarlem (eds.)
Perspectives on Radio Astronomy Technologies for Large Antenna Arrays
Netherlands Foundation for Research in Astronomy - 1999

The antenna consists of an array of active radiators. Active here means that the radiated power is
generated in the radiator module. The receive function is integrated with this design. Of course the full
radar has other system components for signal processing and management. These are not shown on the
figure. The radar, called APAR, operates in the X-band (so 10 GHz). The size of the antenna is
approximately 1 x 1 m2. [1]
Although many phased array radars have an appearance like this one, not all the phased array radars
look like the APAR. Other examples are the impressive phased array radars that served the Ballistic
Missile Early Warning system from the United States, with their equipment building as high as a 10
floor apartment building. Or the large Russian Over the Horizon mattress antennas, or the long
Jindalee Over The Horizon radar in Australia.
An extraordinary concept is the French RIAS (Radar a Impulsion et Antenne Synthetique), set up by
ONERA and built by Thomson-CSF/Airsys. Fig. 2 shows a picture of the antenna system. Note the
two circular arrays, sparsely populated. The array elements are mounted on the white vertical masts.
This radar operates in the HF/VHF band (50 MHz). The diameter of the outer annulus is
approximately 500 m. [2-3]

Figure 2: Radar a Impulsion et Antenne Synthetique (RIAS)

(Courtesy Thomson-CSF/Airsys)

From the antenna point f view, the arrays as used in radio astronomy may look like arrays in radar, as
illustrated by the artist impression of NFRAs Square Kilometer Array, shown in Fig. 3.

Figure 3: Artist impression of NFRAs Square Kilometer Array

2

Phased Arrays in Radar and Astronomy.

There are several reasons for selection of the phased array technique in radar.
Of course, to start with, radar is an active device (although we have seen some proposals for radars
using the solar noise as the emitted signal. So the radar itself then would be passive). Active device
here means that the radar transmits a signal and by matching the received signal to the emitted one,
analyses on the presence of echoes of objects and their properties. The associated signal processing
may either use or not use phase and amplitude of the received signal. Phase processing here serves
compression of the time domain signal to achieve improved resolution or some Doppler-related signal
property. Radar signal processing has evolved that far that identification routines are currently
required in al novel systems. These routines should support identification of objects, in order not to
start off inappropriate military action.
The emphasis in radar is on localization and identification in range, bearing and sometimes elevation
and speed. Angular resolution is of the order of 1 to 2 degrees. Range accuracy is of course depending
on the application, but typically for surveillance functions in the order of tens of meters.
High range and cross-range resolution profiles are required for the identification function. One should
think of less than 50 cm (bandwidth 300 MHz). Anti-stealth waveforms may go further in bandwidth.
Ground penetrating radar for anti-personnel landmine detection extends up to 3 GHz of bandwidth.
This is all in sharp contrast with the needs in radio astronomy, where the main quest is to achieve a
spatial resolution of 1 arc minute or even less than one milli arcsecond at a certain sensitivity.
Now the question may be posed: why a phased array solution.
The responses to this question in het communities of radio astronomy and radar are of course very
different, because the applications are wide apart.
The main reason in radio astronomy is that very large apertures can be created, whilst the angular
resolution is established by synthesizing the direction of arrival from this -maybe sparsely populatedaperture.
The main reason in radar is, that with the phased array technique, the beam of the antenna can be
steered in a certain new direction which is very much differing from its current position without delay
due to mechanical inertia. This might be called beam agility. This beam agility is important when
the radar has multiple functions to be performed simultaneously. The multifunction capability of
phased array radar is illustrated in Fig. 4. The grey level in a beam represents a particular waveform.
As the scenario evolves over time, the radar will spend its resources accordingly.

Figure 4: Phased Array Radars are preferred for their multifunction capability.
(Courtesy Hollandse Signaalapparaten)

The functions the radar must be capable of could be

limited volume search, with waveforms dedicated to classes of objects (e.g. missies, ships,
helicopters, )
accurate tracking of objects
identification waveforms, requiring relatively long dwell time in a certain angular sector
guidance of missiles
The order of magnitude of the antenna dwell time in a certain direction may vary in the same radarfrom less than 1 msec to more than 100 msec.
Phased array radars can be controlled to spend their time and power resources in a user defined
optimum way. This is not the case with the common surveillance or track radar, which are optimized
for one particular mission only.
Managing the radar so that it complies with its operational objectives, leaving no time and no power
unspent is a very complex task. The beam agility is just one aspect of the complexity of the design,
introduced by the multifunction capability. It is one of the cost drivers, making that phased array
radars cost multiple tens of millions of dollars per system.
Complexity is paid by the manufacturer at design stage.
There is one other cost driver: the antenna. Either it is the power distribution network and phase
shifting capability or -in other system approaches- the transmit/receive module which is pushing cost
sky high. Current cost of T/R modules at X-band is in the order of $1000 a piece, where the
expectation is that it will lower to $500 once mass production can be realized. Note that one phased
array is normally populated by 1000 to 3000 of these T/R modules per array.
Given this cost factor, it should be no surprise that the majority of technology actions concerns either
reduction of the cost by technology evolution or by technology substitution. The complexity however
remains.
This list of phased array systems presented in Table 1 does by no means pretend to be exhaustive. It
shows that many industries have the capability to design and manufacture phased array radars. In
effect, there are that many industries, that the tremendous investment required for the first of class per
system doesnt pay back soon. So technological advances will not be picked up readily. It might take
considerable time before budgets are available to make innovative-and costly- steps forward.

status

name

band

TX beamforming

Antenna structure

1974

AN/SPY-1

Constrained feed

fixed

Appr. 1980

PAVE PAWS

UHF

Active modules,
SS, thinned array

fixed

Under test

EMPAR

Under test

ARABEL

Space fed

SAMPSON

Active modules

concept

AMSAR

Active modules

Under test

APAR

RIAS

HFVHF

Based on
MESAR

Demonstrator

Direction finding Nr of faces

monopulse/
subarrays

4
2

rotating

monopulse

rotating

monopulse

rotating

fixed

monopulse/
subarrays
monopulse

Active modules

fixed

monopulse

Omni directional

fixed

digital
beamforming

2
circular

Constrained feed

Table 1: Some Phased Array Radars and their properties

Competition is fierce however, so that if a manufacturer could produce new systems at a considerably
reduced level of cost, he may have a decisive technological advantage and gain a larger market share.
National defence strategies of course interfere with this market mechanism.

4.

Trends in Technology for Phased Array Radar

Having thus introduced phased array radar, and having stated that certain aspects of the antenna are
cost drivers, now the question what technological trends are visible that could either lift the problems
of phased array radar designers or that may present solutions to important system performance
requirements will be addressed. Such a review can hardly be complete; so no conclusive statement can
be presented related to the ultimately achievable objectives in terms of cost or performance.
4.1 Increase of bandwidth
The first trend in all radar applications is, that they should provide a higher bandwidth , either by
synthesizing narrow band signals into a wide band or by instantaneous wide or ultra wide band
signals. The problem in synthesizing wide band signals based on a set of narrow band signals is that it
takes measurement time. Note that time is scarce in phased array radar.
The alternative to synthesized wide band would be instantaneous wide band, eventually even leading
to true time delay beam forming in stead of phase shifting. We saw an example in the RIAS radar.
The important fact is, that several of the technologies described in next sections can be optimized for
narrow band radar, but not for wide band radar, or only at a very high cost.
4.2 Transmit/receive module
One of the important functions in active phased array radar is the transmit/receive module. In active
phased array radar, there may be several hundreds to several thousands (in extreme cases even tens of
thousands) of these. Two aspects of this component are highlighted:

the power added efficiency, and

the cost per die.

First the Power Added Efficiency.
Fig. 5 shows the power added efficiency as a function of the RF frequency for the MESFET
technology. Power Added Efficiency (PAE) is defined as the power difference between the output and
the input of the device compared to the power supplied to it. One may see the PAE is acceptable at

Figure 5a PHEMT technology

Figure 5b MESFET technology

relatively low frequencies (1 GHz). And also that at Ka band (good for tracking accuracy) the
efficiency drops to below 20%. AT X-band the PAE is around 40%. This means that in order to obtain
10 kW of radiated power more than 25 kW should be supplied. And thus that 15 kW must be carried
off some way or another. For PHEMT technology (Fig. 5a), the efficiency at 10 GHz is higher than in
the MESFET technology, but at higher frequencies it is really getting very low. Now these numbers
are taken from literature as old as 1989 [4]. But they have been confirmed to be still valid today, at
industrial series production level.
The other important aspect is the cost of MMICs. Fig. 6a shows the cost per die, as a function of the
surface of the MMIC [5]. The far right of the horizontal axis is 100 mm2. The lower band of cost
numbers represent the optimum numbers when the yield of the production is very high. But the fact is
that the yield decreases significantly with the die size. In fact, in 1986 the yield for the largest size
(100 mm2) was as low as 10%.
These numbers have improved since then. State of the art industrial yield is now that the smaller dies,
say 2 mm2, can be manufactured at a yield level of 90 - 95 %. Larger dies, say 10 mm2, can be made
with, say, 65% of yield.

1986

$/Die

mm2
-32

Figure 6a: Cost per die

-28

-25

Figure 6b: dB side lobe level due to rms phase error

However, the cost of dies maintains a critical issue. The yield may be considerably improved when
tolerances of the performance could be relaxed. One particular example of the criticality of the
tolerances is illustrated by Fig. 6b [6]. This figure shows the rms error distribution of the phase of a
phase shifter over a number of wafers. So for each wafer, the rms phase error of all shifters on the
wafer was measured, resulting in one number per wafer. Then, the statistic was compiled over all
wafers, resulting in the indicated graph. Added to this figure is the impact on the root mean squared
beam sidelobe level (dB). Knowing that this example concerns a 6 bits phase shifter, where the phase
quantization alone would result in -32 dB side lobes below the gain of each element, a considerable
number of wafers should have been discarded and considered as unacceptable, because of unbalance
between the system side lobes and the contribution of these phase shift dispersion to these side lobes.
The message is: the performance specification is very critical; the size of the dies should be included
as system design parameter.

4.3 Phase Shifter Technology

4.3.1

The Micro Electrical Mechanical Switches

Looking for alternatives to Gallium Arsenide phase shifters, several technologies emerge. The first
one I will address here is the Micro Electrical Mechanical Switch [7].
It is a compact mechanical device, as illustrated in Fig 7, where a membrane can be switched to either:
contact closed or contact open. In the unactuated state, the membrane switch (in bold line) exhibits a
high impedance due to the air gap between the bottom and the lower metal plates. When a DC
potential is applied between the upper and the lower metal plates, the membrane will deflect
downwards due to electrostatic interaction. In this state, the top membrane rests directly on a
dielectric layer and it then is capacitively coupled to the bottom plate.

Figure 7: Micro Electrical Mechanical Switch, cross sectional view (left) and
phase shifter implementation (right)

By embedding the switches in a microwave circuit with various propagation path lengths in order to
create a switchable path length as illustrated in the other figure, one may control the propagation time
delay and thus the phase. A 10 GHz 4 bit phase shifter would have a size of 6 x 10 mm2. The main
advantage of such a device would be its cost, as can be seen from the table. It is one order of
magnitude cheaper than other techniques. Because it is a true time delay phase shifting its inherent
bandwidth is large. Compared to Gallium Arsenide it has the disadvantage of a higher insertion loss
and a switching time of 4 microseconds.
4.3.2

Voltage Variable Dielectric phase shifters

Another alternative could in some applications be the use of Voltage variable dielectric phase shifters.
The principle of operation is based on the change of the dielectric properties of some ceramic material
upon application of an electric field [8].

RFin
V

Metallized areas

Figure 8: Layout of a Voltage Variable Dielectric Device

In Fig.8 it means the application of a DC voltage over a layer of the material. The RF wave is injected
at the left side edge of the layer and propagates through it. By application of the DC voltage the
permittivity and thus the speed of propagation is changed. Therefore the phase of the signals between
the coupling stubs can be controlled. This technology seems very interesting: it results in very simple
and compact two dimensional scanning capability. However it is not yet matured for application:
The value of the DC voltage to be applied in order to achieve a change in velocity such that in a
half wavelength of spacing between the stubs sufficient phase shift is obtained is rather high. It
may be as much as 10 kiloVolt.
The insertion losses are high. They are approx. 1.5 dB per 360 degree of phase shift.
The permittivity is changing with temperature. Because of the considerable insertion loss and
propagation loss, an antenna system based on this technology would require permanent
monitoring and calibration.
The homogeneity of the dielectric material in production is poor.

4.3.3

Photonics

A technology which is not yet matured either, is the use of optical means. In research projects various
approaches to the use of electro-optical devices are underway. Three of them will be indicated here.
First: The control of which element of an array is activated. Such an array would then be
reconfigurable. Research has demonstrated the application of activation by laser of photoconductive
silicon bowtie radiating elements.[9]
Second: In phased arrays all elements must be controlled individually. Various concepts ranging from
centralized to fully decentralized computation of the phase setting per antenna element can be
considered, each requiring considerable cabling or active computation in the antenna. An alternative to
this would be the use of fiber optic data transport, with a very wide bandwidth and little vulnerability
to EMC.
Thirdly: a very interesting one is the phase shifting by modification of the refractive index due to an
electric field, as illustrated in Fig. 9 [10]. Here the RF signal is modulated on the light emitted by a
laser. By changing the refractive index of an optical wave guide, the phase of the optical signal at the
output of the propagation path is changed. After the application of optical delay, the signal is mixed
down to obtain the RF signal, however now changed in phase.

Experiment:
1550 nm (optical) wavelength; 10 GHz radar
16 element array on 8.5x8 mm2
phaseshift 720 over 4.9 mm, applying 20V
on chip losses very high (25 dB) depending on
phaseshift

Figure 9: Photonic phase shifter

4.4 Alternative radiator designs

Other advances in phased array are in the radiating elements and the structures carrying the elements.
Of course as referred to before, the trend is towards greater bandwidth, resulting for typical application
into a requirement of 30% fractional bandwidth. Various solutions have been studied, with another
common property, next to the bandwidth issue, namely the weight. An extreme example is the
spaceborn inflatable L-band (25 cm wavelength) microstrip array [11]. It has a relatively small
bandwidth, but it is dual polarized. The weight of a 10x3 m2 aperture is 100 kilograms. Microstrip
arrays are investigated at many places, sometimes combined with proper selection of dielectric
materials in order to reduce their size.
Under development still is also the application of conformal surfaces, especially in the airborne phased
array radars. In this context it is worth mentioning that a trend can be observed to combine different
devices emitting in RF bands in the same aperture. This trend arises from both the reduction of the
effective radar cross section of antennas and of the increasing number of RF emitters on each
platform.
Polarization in phased array radars mostly is single linear. But in the remote sensing applications the
use of dual polarization is common nowadays, whereas it is penetrating in surveillance radar.
However, the improvement of detection of moving objects using polarimetry is not yet fully assessed.
4.5 Substitution of analog components by digital ones
Important for the phased array system designers is the increasing ability of digital processing
techniques. This permits the beamforming to be digital eventually, substituting analog technology by a
digital one. Especially in this area some very interesting demonstrator radars have been built. Refer
again to the RIAS radar. This RIAS is a sparse array. The mutual distance between the array elements
is several wavelengths and beamforming is done in digital processing. In this processing phase shifting
is implemented by true time delay, which is very simple to achieve once the signals are digital. RIAS
emits omni directionally. The signal consists of a set of orthogonal signals. Upon reception in two
concentric annuli of antennas mutually displaced both horizontally and vertically RIAS is able to
perform 3D measurement of direction of arrival. The sequence of various linear operations is also
uncommon: in each array element, the signal is converted to digital, then Doppler filtered in each
member of the set of orthogonal signals, then the direction of arrival is assessed and finally automatic
detection of objects is performed. The digital technology of today already allows for this complicated
and massive computational process.
So: digital techniques enabled the substitution of analog circuitry. This trend is manifest over a
number of years in the past.
4.6 Sparse arrays
Stimulated by the cost issue, the radar community has evaluated the feasibility of alternatives to the
densely populated gridded array, arriving at the concept of the sparse array. Demonstrator radars exist.
However their application in radar is in niches, like ground based HF radar (refer to RIAS).
The disadvantages of this technique relevant to application in radar are mainly related to the gain and
the side lobes of the antenna. Especially the sidelobes of sparse arrays are difficult to control at design
stage. The mean side lobe level is inversely proportional to the number of radiating elements. Most
difficult is the peak side lobe level. Putting strong constraints on this, converts to requiring large
numbers of elements. Thus the net effect on cost savings is dubious.

5.

Conclusion

Phased Array Radars now enter the market. Many products exist / are under customer funded
development Technology of phased array radar is still developing and will enable significant cost
reduction on the long run. In radar, the complexity of phased array radars will stay high, and thus one
of the cost components will maintain to be high. Radar may profit from radio-telecommunications
technology improvements.

References
[1] A.K.Golshayan, S.v.d.Schoot, P.van Genderen, Active Phased Array Radar (APAR), Proc.of the
5th International Conference on Radar Systems, Brest (F), 1999.
[2] J.Dorey, G.Garnier, RIAS, Radar a Impulsion et Antenne Synthetique, LOnde Electrique,
November-December 1989, Vol.69, nr 6, pp 36-44.
[3] A.-S.Luce, H.Molina, D.Muller, V.Thirard, Experimental results on RIAS digital beamforming
radar, Proc. of the International Conference Radar 92, Brighton (UK), 1992
[4] Y.C.Shih, H.J.Kuno, Solid state sources from 1 to 100 GHz. State of the art reference,
Microwave Journal 1989, p.145
[5] A.Podell, D.Lockie, S.Moghe, Practical GaAs ICs designed for microwave subsystems. The
Microwave System Designers handbook, 4th ed., Vol.16,No.7, p.327. EW Communications Inc,
Palo Alto (USA), 1986.
[6] F.L.M.van den Boogaart, J.G.bij de Vaate, Production results of a transmit/receive-MMIC chip
set for a wide band active phased array radar at X-band, Conf.Proc. MM92, pp 138-143, Brighton
(UK), 1992.
[7] B.R.Norvell, R.J.Hancock, M.L.Pugh, S.W.Theis, J.Kviatkofsky, Micro Electro Mechanical
Switch (MEMS) Technology Applied to Electrically Scanned Arrays (ESA), Proc. Of the
International RadarSymposium IRS98, pp 867-875, Munich (D), 1998
[8] T.W.Bradley, W.W.Chen, J.R.Chow, S.B.Coppedge, T.K. Dougherty, P.J.Fitzgerald,
S.A.Gabelich, N.H.Harris, N.T.Hrobak, R.T.Lewis, J.F.Linder, W.W.Milroy, B.M.Pierce,
R.J.Wittmond, Development of a Voltage-Variable Dielectric (VVD), Electronic Scan Antenna,
Proc. of the International Conference Radar 97, pp 383-385, Edinburgh (UK), 1997
[9] D.Liu, D.Charette, M.Bergeron, H.Karwacki, S.Adams, F.Kustas, K.Farley, R.Munson, A 1-18
GHz photonically reconfigurable phased array antenna, 1998 IEEE Aerospace Conf.Proc., pp
483-490, Snowmass at Aspen (USA), 1998
[10] J.Stulemeijer, D.H.P.Maat, I.Moerman, F.E.van Vliet, M.K.Smit, Photonic Integrated
Beamformer for a Phased Array Radar, 1997.
[11] J.Huang, M.Lou, A.Feria, Y.Kim, An inflatable L-band microstrip array, IEEE Antennas and
Propagation Society International Symposium 1998 Digest, pp 2100-2103, Vol 4, Atlanta (USA),
1998

10