Professional Documents
Culture Documents
CHAPTER NINE
9.1 INTRODUCTION
333
334 MONOLITHIC MICROWAVE CIRCUIT TRANSMITRECEIVE (T-R) MODULES
The antenna needs to have a diameter of several wavelengths in order to direct the
radiated energy. Features of these functions are discussed in the following paragraphs.
The RF transmitter must provide power to reach the target, be scattered, and then
be received back at the system site. The received signal must be detectable by a
suitable low noise receiver. The required power is determined by the radar range
equation [1]. The antenna aperture size is important also as the antenna directs the
radiated RF power into a concentrated beam rather than radiating energy in all
directions equally. Readers are referred to [1] for a complete discussion of radar
system operation. As the RF energy is transmitted and received by the same antenna,
RF pulses of energy must be transmitted and received in time sequence. That is, an
RF pulse of high energy is transmitted; the transmitter is turned off; a return
scattered pulse is received; and the received signal is passed through a duplexer, is
amplied by a low noise receiver, and is processed by the signal processor.
9.1 INTRODUCTION 335
conguration, a network is provided that divides the RF energy so that it may pass
through each phase shifter. Each phase shifter in turn feeds an individual radiating
element. Congurations are shown in Fig. 9.2.
Phase-shifted elements are made up of ferrite or diode phase shifters. Although
diode phase shifters were used in a few lower frequency search radar applications,
the major effort and applications have used ferrite phase shifters. Diode phase
shifters use a variety of circuits including switched length of transmission line or
3-dB hybrid couplers terminated in varactor diodes. For passive array applications,
diode phase shifters present more insertion loss than ferrite types. In ferrite phase
shifters, phase shift is obtained in increments by partially switching the magnetiza-
tion of the ferrite material [2]. Ferrite phase shifters are often referred to as ``latching
ferrite phase shifters.'' A 5-bit phase shifter would provide phase shift in increments
of 0 , 22:5 , 45 , 90 , 180 , and 360 .
Most present elded passive arrays use ferrite phase shifters. During the late
1960s and early 1970s nonreciprocal phase shifters received considerable develop-
FIGURE 9.1 l=2 l=2 waveguide egg crate or honeycomb antenna structure.
9.1 INTRODUCTION 337
mental efforts [35]. Slightly later, emphasis changed to reciprocal phase shifters as
these need not be switched between the transmit and receive operations. Today, both
types are utilized in elded systems. Nonreciprocal and reciprocal phase shifters are
depicted in Fig. 9.3. A partial list of radar types that have been elded using ferrite
phase shifters are listed in Table 9.3
FIGURE 9.3 (a) Nonreciprocal toroidal phase shifter and (b) dual-mode reciprocal phase
shifter.
by a transmitreceive (TR) module that includes a good portion of the radar system.
The TR module includes the RF front end fabricated in solid-state format. The
module architecture may vary slightly from system to system. However, the TR
module usually includes the following:
Solid-state transmitter
Phase shifter
Attenuator
TR switch or circulator
Receiver protector
Low noise amplier
Other required circuitry
Energy storage
Control and Switching logic
Bias circuitry
A simplied block diagram of a TR module is shown in Fig. 9.4. In an active
array, each module's solid-state amplier must provide 1=n of the total radiated
power. Here n is the total number of l=2 l=2 elements in the array.
9.1 INTRODUCTION 339
Work on active arrays has gone on since the late 1960s with major emphasis in
the late 1980s through the mid 1990s. Thus far, the driving force has been the
complex system needs of X- to Ku-band airborne radars. Here the large number of
TR modules per system and the large number of aircraft platforms per system class
provide the potential quantities to justify large development programs.
The U.S. Air Force through contractural program at Texas Instruments has
provided pioneering early efforts [68]. These rst programs were named MERA
(Molecular Electronics for Radar Applications) and RASSR (Reliable Advanced
Solid State Radar). A MERA module is depicted in Fig. 9.5 and a RASSR module is
shown in Fig. 9.6. These modules were fabricated before X-band GaAs FETs were
available. The RF transmitters in these modules utilize S-band silicon bipolar
transistors and frequency multipliers to get to X-band.Such a conguration produces
only a fraction of a watt with very low dc RF efciency. Additionally, these modules
contain hundreds of parts, resulting in high cost and poor reliability. However, much
(a)
(b)
FIGURE 9.7b RASSR array with radiators. (Photographs courtesy of J. Wassel, Texas
Instruments.)
important groundwork was performed in these programs. Figure 9.7 shows MERA
and RASSR arrays.
With continuing support from DARPA (ARPA), the three military services, and
private industry, much progress has been made. Programs such as MIMIC (Micro-
wave and Millimeter Wave Monolithic Integrated Circuits), HDMP (High Density
Microwave Packaging), and MAFET (Microwave Front End Technology) have
provided substantial funding. The MIMIC program has been a four-year program
to support GaAs TR module chip set optimization along with developing design
and manufacturing tools to economically produce such chip sets. MAFET is a
follow-on program that places emphasis on computed-aided design (CAD) and
module packaging activities. These ARPA triservice programs supported major
industry teams. Figure 9.8 depicts a state-of-the-art module containing nine GaAs
chips [9]. Key parts of the TR module include the following:
342 MONOLITHIC MICROWAVE CIRCUIT TRANSMITRECEIVE (T-R) MODULES
At this time a state-of-the-art TR module may still contain 50100 parts. This in
turn leads to small quantity costs of a few thousand dollars per unit with cost goals
for quantities of 100,000 being at about the $5002000 range.
GaAs FETs
PHEMTs
HFETs
HBTs
GaAs FETS are used in most present designs although the devices with more
complex doping proles (PHEMTs, HFETs, and HBTs) may offer slightly increased
efciencies and possibly more power per unit periphery. The least mature of these
device types is the power HBT (heterojunction bipolar transistor). The life of the
HBT device (MTBF) has been somewhat less than the other device types.
Performance of all these device types is continuing to improve.
The overall dc RF efciency for a well-designed X-band power amplier chain is
on the order of 2535%. Thus about three times the amplier's output power must be
removed as heat. This is accomplished by thinning the GaAs substrate to 34 mils
thickness. Vias are etched through the substrate material to make thermal contact
with the active device region. Gold is back-lled into the vias and the resulting
substrate is soldered or otherwise bonded to a thermal carrier or in some cases
directly to the module housing. Strips or rows of modules are placed against a cold
plate, which in turn has channels for liquid cooling. This approach is depicted in Fig.
9.10. Flip chip mounting of GaAs ampliers has been used as an alternate approach
at a few industrial organizations. This approach is depicted in Fig. 9.11. The ip chip
mounting technique eliminates the need for active device vias and thinning of the
GaAs substrate. However, the heat must still be removed and visual inspection of
devices=circuits is difcult. For an array in which 1500 6-W modules are used,
considerable cooling energy is required:
For arrays in which higher power is required, the available power from the air
platform for cooling may present a major problem. Dissipated power will be
discussed further under thermal considerations.
TR protector
Circulator
Phase shifter
Attenuator
TR switch
FIGURE 9.13 Candidate chip sets for TR modules: (a) 5-chip design, (b) 3-chip design,
and (c) 2-chip design.
batch approach is favored. Two such approaches are being considered as replace-
ments for the conventional chip and wire approach. These are low temperature
cored ceramics (LTCCs) and microwave high density interconnects (MHDIs).
For the LTCC approach substrates within a module are replaced by a structure
made up of ceramic layers that contain layers of metal interconnects. These layers
are made from tapes, which are red at somewhat lower temperatures than other
ceramics. Some components such as 3-dB directional couplers, which are used for
combining amplier stages, may be fabricated directly into the LTCC layers. With
this approach, some wire bonds are still required for nal component assembly. The
LTCC layering approach presents a three-dimensional CAD problem for
circuit=device designers. At this time, good progress is being made with LTCC
substrates. Tolerance on layer thickness, accuracy of metal pattern layouts, and
complete three-dimensional CAD of LTCC parts are all being addressed.
The MHDI approach uses conventional ceramic substrates and makes all
interconnects at one level. In this approach a layered exible lm made up of
dielectric and thin metal interconnect patterns is applied as a top layer above active
and passive circuits within the module. A batch processing technique is used to open
windows in the dielectric lm and to make connections to the circuits below. At
frequencies above X-band, the electrical effects (loading) of the layer must be
factored into the design. The MHDI approach is well suited for large scale
production of TR modules. As with the LTCC approach, further work is required
before the full value of MHDIs can be determined.
RF output
RF input
FIGURE 9.14 Module has one port as RF input, one port for dc bias and control signals, and
one port as RF output, which goes to antenna radiator.
attached directly to the module wall. Several modules are typically connected in a
row to a coupling manifold. For broadband modules that might operate over the 6
18 GHz band, the output connectors need to be spaced no more than l=2 at the
highest operating frequency. Thus placing two or four modules in a single package is
desirable. Figure 9.15 shows four 4.518 GHz 0.5-W modules in a tray. Here the
driver electronics are packaged at the input end of the tray.
To date, the major deterrent to active phased array systems has been the cost of
fabricating and arraying large numbers of individual TR modules into an antenna
face. As indicated earlier, much hard work has gone into cost reduction studies for
individual modules. The reduction of the number of circuits or other parts within a
TR cell has been important, as has automated assembly and testing. The volume
production of modules for a few systems has contributed to cost reductions.
ARPA=triservice programs such as MIMIC and MAFET have helped reduce the
number of design cycles in module development.
Other efforts for cost savings have utilized two or four TR modules in a common
metal package. The use of LTCC substrates has been valuable in such efforts.
In present designs the cost of active arrays is further increased by the mechanical
structure that holds the individual TR modules, radiating antenna elements, RF feed
structure, and cooling manifold.
An ARPA-sponsored program with Westinghouse was established in the early
1990s to explore the possibility of fabricating subarrays of modules compete with
cooling and radiating elements at the same time as other structural elements. The
program was named RF Wafer Scale Integration (RFWSI). The program was
completed in 1994. The RFWSI program was followed up by three HDMP
contracts to industry teams. The HDMP program was somewhat more conservative
than the earlier RFWSI program. Some key features of these programs are
described below.
In the RFWSI approach, individual TR modules are eliminated and are
replaced by a multielement coplanar subarray. A very thin RFWSI subarray
structure results, which may be ush mounted or incorporated into a vehicle's
surface. The goal of the RFWSI program was to demonstrate the realization=
fabrication and testing of ve 16-element subarrays. The specic performance goals
for an individual TR cell were:
Transmit: Bandwidth 6 12 GHz
Radiated power 0:5 W
Phase bits 6
Amplitide bits 5
Receive: Bandwidth 4 12 GHz
Noise figure < 7 dB
Gain > 20 dB
9.4 PRESENT TRENDS AND FUTURE DIRECTIONS 351
GaAs layer
Cooling layer or manifold
dc Bias circuits
RF feed structure
Control circuits
Antenna radiating elements
These are depicted in Fig. 9.16. The entire subarray was less than 2 inches thick,
with the circulator layer and the wideband radiating element consuming most of the
thickness [13]. The block diagram of a TR cell is given in Fig. 9.17. The completed
subarray is shown in Fig. 9.18. In the RFWSI approach, individual TR cells are
fabricated in l=2 l=2 areas that are at right angles to the signal ow. Thus layers in
this structure have been called ``tiles'' as opposed to ``bricks'' in the more usual
architecture. In the following HDMP program these terms were widely used. The
two approaches are depicted in Fig. 9.19.
In the RFWSI conguration, thermal, dc, and RF feed-throughs between the
various layers are required. The individual layers must be thermally matched to
prevent deformation or cracking as thermal cycling occurs. The realization of large
area GaAs chips with high yield was accomplished by using circuit redundancy.
Automatic testing and closure of mechanical switches allowed the better parts to be
utilized.
The TR cell that was developed is shown in Fig. 9.20. The reduncancy in circuit
elements is as follows:
With this degree of redundancy, about a 75% TR cell yield was obtained.
The RFWSI demonstration chose broadband applications with modest power
requirements. The HDMP program focused on higher power radar TR cells. This
program utilized existing GaAs module chip sets. The program has been carried out
by three industrial teams. The teams retained the ``tile'' concept and used a variety of
interconnect techniques. Two teams used ip chip mounting of GaAs chips. The
352 MONOLITHIC MICROWAVE CIRCUIT TRANSMITRECEIVE (T-R) MODULES
FIGURE 9.16 RFWSI subarray approach: (a) assembled view and (b) exploded view.
HDMP program helped dene CAD problems, which are still being addressed in the
MAFET program. A good discussion of the Hughes team approach is found in the
literature [14].
Advances gained in the RFWSI, HDMP, and MAFET programs should lead to
the next generation of active arrays. Airborne applications will continue to focus on
X-band and Ku-band frequencies. New applications will be found at higher
354 MONOLITHIC MICROWAVE CIRCUIT TRANSMITRECEIVE (T-R) MODULES
FIGURE 9.19 The tile and brick approach to TR module packaging: (a) brick approach
and (b) tile approach.
Acknowledgments
The author is grateful to Mr. John Wassel and Dr. David McQuiddy, Jr. of Texas
Instruments for providing photographs of early module work. The author also is
appreciative of past interactions with Mr. John Zingaro and Dr. James E. Degenford
of Westinghouse (now NorthropGrumman).
REFERENCES 357
REFERENCES