36 February 2009

Near-Field DirectAntenna Modulation

Modern silicon-based technology processes haveopened a plethora of opportunities fordesigning highly integrated millimeter-wave systems by providing transis-tors with cutoff frequency, fT, of

more than 200 GHz [1]–[5]. At millimeter-wavefrequencies, the wavelengths are comparable tothe die size, and this inspires the integration ofthe radiating elements and active circuitcomponents on a single silicon die.Although integration of millimeter-wavesystems on a silicon substrate lowers thecost and improves reliability, there are sev-eral challenges that must be addressedappropriately [1], [2]. Because of con-straints imposed by the fabrication ofactive components, the substrate resistivityof the silicon substrate has to be very small(∼1–10 � cm). This low resistivity causesenergy loss into the substrate and lowers thequality factor Q of unshielded on-chip passivecomponents such as inductors, transmission lines,and antennas and hence results in the degradation ofpower efficiency and noise performance. The finite con-ductivity of metal structures causes further energy loss in inte-grated systems. Since the skin depth becomes very small at mil-limeter-wave frequencies (e.g., the skin depth of copper at 60 GHz is approximately 300 nm), the ohmic loss in metalstructures significantly increases, degrading the performance of passive devices.

As transistors become smaller and faster, their breakdown voltages become very low, introducing a serious chal-lenge to power generation in integrated systems. Rather than implement a single amplifying block that requires ahigh voltage swing, it is more promising to use a parallel structure of multiple amplifying elements with a smallerswing and combine the output power.

Aydin Babakhani, David B. Rutledge, and Ali Hajimiri are with California Institute of Technology, Pasadena, California 91125 USA.

© EYEWIRE

Aydin Babakhani, David B. Rutledge, and Ali Hajimiri

The combination of many high-performance transis-tors and on-chip metal structures comparable to thewavelength, which can be used as on-chip radiators, isrevolutionizing the design of millimeter-wave and sub-millimeter-wave transceivers and antennas.

In this article, the design of the millimeter-waveintegrated antennas is discussed, the concept of near-field direct antenna modulation (NFDAM) is intro-duced, and the implementation of a 60-GHz proof-of-concept chip based on NFDAM is presented.

Millimeter-Wave Integrated AntennasOne of the primary challenges in designing on-chipantennas on a silicon substrate is the high dielectricconstant of silicon (εr = 11.7). To understand theimportance of the dielectric constant, a simple dipole isplaced on the boundary of semi-infinite regions of airand dielectric, as shown in Figure 1. The percentage ofthe power coupled into both the air and the dielectric isplotted in Figure 2. The figures indicate that for the caseof a silicon substrate with εr = 11.7, more than 95% ofthe power is radiated into the silicon rather than the air.

To avoid coupling the signal into the silicon sub-strate, a ground plane may appear to resolve the prob-lem. There are two ways to implement this approach.One way is to make an on-chip ground plane by using

the lowest metal layer (e.g., M1) while placing an on-chip antenna on the topmost metal layer. Unfortunately,the distance between the topmost and the lowest metallayers is only a few micrometers in today’s processtechnologies. This small distance almost shorts theantenna and reduces its efficiency to below 4–5% [1].The other way to implement a ground shield is to placean off-chip metal plane onto the backside of the silicon.In this case, because of the rectangular shape of the dieand its thickness of several hundred micrometers, mostof the dipole power gets coupled into substrate modes.

February 2009 37

Figure 1. Radiating from top side without any groundshield [1].

Air, εr = 1

Dielectric, εr = εd

Topside

Figure 2. Ratio of the radiated power into air and into thedielectric substrate as a function of the dielectric constant [1].

Nor

mal

ized

Pow

er (

%)

Dielectric Constant

100

80

60

40

20

01 4 7 10 13 16

Pdielectric

Pair

Figure 3. Radiating from the back side of the chip bymeans of a silicon lens [1].

Air, εr = 1

Air, εr = 1

Sio2, εr = 4

Silicon,εr = 11.7

MatchingLayer Backside

Topside

Substrate-Wave

Figure 4. Measured E-plane patterns of two adjacentantennas on a silicon chip with a back side silicon lens.

Angle (°)

Gai

n (d

B)

Antenna 2

Antenna Pattern in E-Plane

Antenna 3

105

0−40 −30 −20 −10 0 10

φ = 25°

20 30 40−5

−10

−15

−20

−25

−30

To understand the importance of thedielectric constant, a simple dipole isplaced on the boundary of semi-infinite regions of air and dielectric.

38 February 2009

A significant part of the power coupled into the sub-strate modes disappears as heat, and the rest is guidedto the chip edge, which often results in an undesirableradiation pattern [1]. To avoid these problems andminimize the power coupled into substrate modes, ahemispherical dielectric lens, with a dielectric constantsimilar to that of silicon, can be placed on the back side

of the silicon substrate [6]–[14]. The lens combines thesubstrate modes and converts them into useful radiatedpower, as illustrated in Figure 3.

To measure the antenna pattern, a W-band hornantenna is used to irradiate the on-chip dipoles on thereceiver side. The E-plane patterns measured by thespherical method are shown in Figure 4. A maximumpeak gain of +8 dBi is achieved in the measurement.Because of the off-axis properties of the silicon lens, thepeak points of two patterns have an angular separationof 25◦, which is close to the theoretical value of 23◦ [14]corresponding to the lens dimensions shown in Figure 5.

Near-Field Direct Antenna ModulationThe availability of a large number of transistors with lit-tle incremental cost and high reproducibility in siliconenables a designer to take into account multiple levels

of abstraction in a single chip—from sys-tem-, architecture-, circuit-, and device-levelto electromagnetic structures. This interdis-ciplinary integration on multiple levelsallows for implementing novel conceptsand architectures that would not be feasiblefor conventional module-based approaches.

Near-field direct antenna modulationis a novel concept that is established bytaking full advantage of such interdisci-plinary integration capability [15], [16].By combining antenna electromagneticboundary conditions with data-transmit-ting architectures and circuits in a singleintegrated system, an efficient andsecure communication scheme is possi-ble, which is fundamentally differentfrom conventional modulation schemes.The NFDAM-based system allows fordirection-dependent data transmissionwith higher security than the conven-tional systems. In this section, the con-cept of the NFDAM technique isdescribed in detail along with the associ-ated advantages. Subsequently, theimplementation of a NFDAM-based 60-GHz integrated transmitter is describedand accompanied by measurement

results, which serve as a proof of concept.To introduce the idea of NFDAM, two modulation

schemes are illustrated in Figures 6 and 7. A conven-tional scheme is shown in Figure 6, where the informa-tion is created at baseband and the modulated signal isup-converted by a mixer. In this configuration, becausethe antenna pattern remains unchanged at each sym-bol transition, any change in the baseband signalappears at every direction in space. For example, if theamplitude of the baseband signal is increased by a fac-tor of two, the amplitude of the transmitted signal willbe doubled at every direction in space. In this case, an

Figure 5. Lens dimensions.

1

2φ

3

4

1.6 mm

12.7 mm

0.88 mm

Figure 6. Modulation at baseband [16].

Signal Is Modulated before the Antenna

Q

Base-BandData

Mixer

Antenna

Oscillator

DesiredDirection

UndesiredDirection

Q1

1

2

22

1

I

I

I

fc

Figure 7. Modulation at antenna level [16].

Signal Is Modulated at the Antenna

UndesiredDirection

DesiredDirection

fc

2

I

1

2Q

Oscillator

1I

Q

In an NFDAM system, only a lockedunmodulated carrier signal drives theantenna. In this case, the informationis created by changing the phase andamplitude of the antenna pattern atdifferent directions.

eavesdropper with a sensitivereceiver located at an undesireddirection can receive all thechanges in the baseband data andhence properly demodulate thesignal. In an NFDAM system,only a locked unmodulated carri-er signal drives the antenna. Inthis case, the information is creat-ed by changing the phase andamplitude of the antenna patternat different directions. In otherwords, because the antenna pat-tern changes at each symboltransmission, the change inamplitude and phase of thereceived signal becomes a func-tion of the direction. In this case, itis possible to double the ampli-tude of the received signal at onedirection but decrease it (or leaveit unchanged) in another direc-tion. Figure 7 illustrates the ideaof NFDAM, where the informa-tion is sent to the desired direc-tion and the misinformation issent in the undesired directions.In an NFDAM system, in order tomodulate the signal, the antenna

February 2009 39

Figure 8. Signal modulation using switches on the reflectors (a) open switch and (b) closed switch [16].

Bit = 0

A0cos (ωt + ϕ0)Main Signal

A1cos (ωt + ϕ1)Reflected Signal

A0cos(ωt + ϕ0) + A1cos (ωt + ϕ1) = A’cos (ωt + ϕ’)

Antenna Reflector

Bit = 1

Antenna Reflector

Switching

A0cos (ωt + ϕ0)Main Signal

A2cos(ωt + ϕ2)Reflected Signal

A0cos (ωt + ϕ0) + A2cos (ωt + ϕ2) = A”cos (ωt + ϕ”)

Signal Constellation

I

Q

X O

Z-Direction

(a) (b)

Figure 9. Arbitrary signal modulation [16].

Reflector Reflector Reflector ReflectorAntennab3

b1

b2

b4

b5

b6 b9

b8

b7

b11

b10

b12

Q

I

40 February 2009

characteristics (its near- and far-field) have to vary atthe symbol rate.

An oversimplified example of the NFDAM tech-nique is shown in Figure 8, where a reflector is placedin close proximity to a dipole antenna. An ideal switchis placed in the middle of this reflector to change its

effective electrical length. Let us assume the maindipole antenna radiates a signal A0 cos(ωt + ϕ0) in thez-direction, normal to the plane (boresight). Some partof the main signal couples to the adjacent reflector inthe near field of the antenna, causing the reflector toscatter a signal A1 cos(ωt + ϕ1) at its open state and asignal A2 cos(ωt + ϕ2) at its closed state. On the basis ofthis assumption, the far-field signal in the z-directioncan be calculated by

Open switch: A0 cos(ωt + ϕ0) + A1 cos(ωt + ϕ1)

= A′ cos(ωt + ϕ′). (1)

Closed switch: A0 cos(ωt + ϕ0) + A2 cos(ωt + ϕ2)

= A′′ cos(ωt + ϕ′′). (2)

Figure 8 shows the real andimaginary parts of the combinedsignal in the far field. Although thisoversimplified example showshow a crude signal modulation canbe done by a simple switch, it doesnot capture the complete near-fieldelectromagnetic behavior of thesystem. In our design, where reflec-tors are placed in the near field ofthe antenna, changes in the antennaparasitic components modulate thefar-field radiated signal withoutnecessarily varying the path delay.

Figure 9 shows a more practicalconfiguration where a total num-ber of N switches produce 2N con-stellation points. By having a suf-ficiently large number of switch-es, it is possible to cover most ofFigure 10. Simulation results of the switch-based NFDAM transmitter (10,000 points).

2

1.5

1

0.5

0

−0.5

−1

−1.5

−2

I-Part (10−2 V)

Q-P

art (

10−2

V)

−2 −1.5 −1 −0.5 0 0.5 1 1.5 2

Figure 11. Communication security: (a) desired direction and (b) undesired direction [16].

(a) (b)

Antenna Reflector Reflector

Q

II

Q

Antenna Reflector Reflector

Change of Angle

At the undesired direction,constellation points are completelyscrambled demonstrating thefunctionality of the NFDAM system.

the points on the signal constellation diagram andselect any arbitrary modulation (Figure 9). Figure 10shows a simulation result where 10,000 random switch-ing combinations generate 10,000 distinct points on thesignal constellation diagram. Each point corresponds toan induced voltage on a receiving dipole antenna locat-ed at the far field. In this simulation, five reflectors areplaced at each side of the antenna and nine idealswitches are placed on each reflector, resulting in a totalnumber of 90 switches.

By changing the switching configuration, the nearfield of the antenna and hence its impedance arechanged, but this can be alleviated by exploiting theredundancy. For a maximum of 1.5 dB variation inpower gain, 2 dBm variation in delivered power, and3% variation in power added efficiency of the poweramplifier (PA), it is possible to cover one quadrant withresonant switches that present a maximum impedance(off impedance) of 70 �.

Properties of Systems Utilizing Near-FieldDirect Antenna Modulation

Using a Narrowband (PA) to TransmitWideband InformationIn NFDAM systems, because the modulated signaldoes not pass the PA, specifications/requirements onthe PA and the modulated signal can be totally differ-ent. In fact, a highly efficient narrowband switching PAcan be used to amplify the carrier signal while antennaparasitic modulation can be utilized to transmit a non-constant-envelope wideband signal.

Secure Communication Link and Information Beam WidthAssume that a set of switching combinations can befound that generates 16 quadrature amplitude modu-lation (QAM) points at boresight [Figure 11(a)]. If the

receiver moves to a different direction, it will seescrambled points on the signal constellation diagram[Figure 11(b)]. This is because the scattering propertiesof these reflectors—and hence the phase and ampli-tude of the reflected signal—vary with angle. This isone of the unique properties of the NFDAM systemsthat prevent undesired receivers from properlydemodulating the signal. Figure 12 shows that the con-stellation points start to move to seemingly randomlocations, and this movement is increased at largerangles. As Figure 12 depicts, some of these points fallinto adjacent compartments and introduce error. Thesimulated error rate versus angle in the E-plane (paral-lel to the dipole) and the H-plane (normal to thedipole) of an on-chip dipole antenna is also shown in

February 2009 41

Figure 12. Information beamwidth for a modulation with 210 equally spaced constellation points [16].

Desired Direction

Err

or R

ate

(%)

100Error Rate Versus Angle

80

60

40

20

0

A Different Angle

−30 −20 −10 0Angle (°)

10 20

ER (%)(H-Plane)ER (%)(E-Plane)

30

Figure 13. Enhancing security by leveraging redundancy [16].

DesiredDirection Undesired

Direction

Q

QI

I

A highly efficient narrowbandswitching PA can be used to amplifythe carrier signal while antennaparasitic modulation can be utilizedto transmit a nonconstant-envelopewideband signal.

42 February 2009

Figure 12, where 210 equally spaced constellationpoints at boresight are selected and viewed at differentangles. In the H-plane, the error rate rises rapidly and

reaches approximately 50% at offset angles of 2 to 3degrees. In the E-plane, the error rate reaches approxi-mately 50% at offset angles of 6 to 7 degrees off bore-sight. Receivers located at angles ±1◦ can completelyrecover the modulated signal in the absence of noiseand other channel nonidealities. From this simulation,the information beamwidth of the antenna, which is ametric of information directionality of the beam, can bedefined as the angle in which a noiseless receiver canrecover the signal without introducing any error. ForNFDAM systems, the information beamwidth

represents the informationdirectionality of the beam whilethe power beamwidth repre-sents the power directionality ofthe antenna pattern.

Redundancy and Added SecurityWith enough redundancy, it ispossible to find different switchcombinations that producealmost the same point in thedesired direction while generat-ing widely scattered points inother directions (Figure 13). Thisunique property of NFDAMsystems makes it possible totransmit a set of predefinedmodulation points in a desireddirection while simultaneouslyrandomly changing the patternof the constellation points in theundesired directions, prevent-ing unintended receivers fromfinding a one-to-one mappingbetween the received signal atthe desired direction and theundesired directions.

Phased-ArrayConfigurationBecause direct antenna modula-tion is performed in the nearfield of the antenna, eachNFDAM system can behave as asingle element in a phased-array configuration. In this case,NFDAM system can control theinformation beamwidth of theantenna while the phased-arrayfunctionality can be used to nar-row the power beamwidth ofthe array pattern and henceincrease the power efficiency ofthe whole system, as illustratedin Figure 14.

Figure 14. NFDAM transmitter in a phased-array configuration: (a) transmitting toboresight and (b) transmitting to a different angle [16].

PatternBeam Width

PatternBeam Width

PatternBeam Width

Single NFDAM Transmitter

Single NFDAM Transmitter

NFDAM in a Phased-Array Configuration

NFDAM in a Phased-Array Configuration

PatternBeam Width

InformationBeam Width

InformationBeamWidth

InformationBeamWidth

InformationBeamWidth

λ/2 λ/2

λ/2 λ/2

(a)

(b)

NFDAM systems provide a uniquesolution for transmitting highlysecured direction-dependentcommunication data.

Implementing a NFDAM 60-GHz TransmitterFigure 15 shows the block diagram of the 60-GHztransmitter [15], [16], which is implemented in the IBM130-nm BiCMOS Si-Ge (8HP) process. In this design,five reflectors are placed on each side of the antennaand nine switches are placed on each reflector, result-ing in a total of 90 switches. Reflectors are implement-ed on metal layers M1 to M3, each having a length of1,300 µm. A dipole antenna with a length of 835 μm isalso implemented on metal layers M1 to M3. A differ-ential carrier signal generated by an on-chip lockedvoltage-controlled oscillator (VCO) is amplified withan on-chip PA. This signal travels inside a shielded dif-ferential transmission line implemented on top metallayers M6 and M7. As shown inFigure 15, the baseband datacontrol the state of switchesthrough a digital control unit.The electromagnetic structure,including the dipole antennaand reflectors, occupies an areaof 1.3 × 1.5 mm2.

The coverage of the signalconstellation diagram is a func-tion of size, shape, and locationof the reflectors as well as theirdistance from the main anten-na. If the reflectors are placedvery close to the antenna, theywill not be able to generate asufficient number of radiatingmodes necessary to transmitindependent signals to multipledirections simultaneously. Onthe other hand, if the reflectorsare placed too far from the mainantenna, the reflected signalwill be very weak and hencewill not be able to significantlyaffect the phase and amplitudeof the main signal. In thisdesign, in order to maximizethe coverage of the signal con-stellation diagram, the distanceof the reflectors from the mainantenna is optimized.

Figure 16 shows the blockdiagram of an optional coarsecontrol unit that can behave as aquadrant selector with controlsignals A and B taking values 1and −1. This unit can be turnedon in situations where the anten-na parasitic modulation fails togenerate widely scattered pointsto cover the four quadrants.

One of the challenging tasks of the design is imple-menting switches with a relatively high ratio of off andon impedances. The switch size cannot be very small asit has to provide a small impedance (about 1 to 5 �) atits closed state. On the hand, a large switch limits themaximum impedance achievable at the open state. To

February 2009 43

Figure 15. Block diagram of the NFDAM transmitter architecture [16].

Reflectors: M1+M2+M3

Dipole: M1+M2+M3

Transmission Line: Ground M6, Signal M7

Switch

PA

VCO

Digital Control UnitBase Band

Dipole

0°

180°

Reflectors Reflectors

1,500 μm

560 μm

375 μm

835

μm

1,30

0 μm

Figure 16. Optional coarse control unit (quadrant selector) [16].

VCO

A,B: Coarse Control

A

90° 180°

0°

B

PA

Base Band Digital Control Unit

DipoleReflectors

Optional FeatureReflectors

The coverage of the signalconstellation diagram is a function ofsize, shape, and location of thereflectors as well as their distancefrom the main antenna.

44 February 2009

alleviate this problem, a circular shielded transmissionline is used to connect the drain and source of the N-type metal-oxide-semiconductor (NMOS) switch res-

onating out the parasitic capacitance of the switch in itsopen state and hence increasing the off impedance ofthe switch (Figure 17). Utilizing this technique provides

a maximum off impedance of 70 � at 60GHz. The NMOS switch used in thisdesign has a length of 120 nm and a widthof 150 μm. The circular transmission lineis implemented on metal layers M2 to M4and has a diameter of 60 μm.

A micrograph of the chip is shown inFigure 18, which includes the reflectors,dipole antenna, PA, up-converter mixers,local oscillator (LO) buffers, VCO, injec-tion locked divider, digital divider chain,and baseband buffers. In addition to thetransmitter, a complete receiver includinglow-noise amplifier (LNA), down-con-verter mixers, baseband buffers, andreceiving on-chip antennas has beendesigned and implemented.

Measurement ResultsA hemispherical silicon lens with a diam-eter of 25.4 mm and an extension lengthof 1 mm is placed on the back side of thesilicon chip. This lens minimizes the sub-strate modes by converting them intouseful radiation modes, as describedabove. Figure 19 shows the block dia-gram of the measurement setup. ALabView program communicates with adata acquisition card through generalpurpose interface bus (GPIB) ports. Thedata acquisition card transmits the base-band data to an on-chip digital controlunit and controls the state of switches.The LabView program also communi-cates with an Agilent N5250A PNA net-work analyzer using a GPIB card. Thenetwork analyzer measures the real andimaginary parts of S21 for every switch-ing combination. In this measurement,the transmitter is disconnected (withlaser trimming) from the on-chip dipoleantenna. In fact, a V-band signal generat-ed by a network analyzer directly drivesthe on-chip dipole antenna through awire-bond connection. The other port ofthe network analyzer measures thereceived signal at the horn antenna.

Figure 20 shows the measured real andimaginary parts of the S21 in two differentdirections with an angular separation ofapproximately 90◦. In this measurement,the same set of switching combinations isused for both directions. Out of 2,000switching combinations, which are selected

Figure 18. NFDAM chip micrograph [16].

QuadrantSelectors

Buffer

Buffer Down-Mixers

LNA

Digital Control

Reflectors

Dip

ole

Reflectors

Bas

eban

d

PAUP-Mixers

90°T-Line

LO Buffer

VCO

Divider Chain

1.5 mm

1.3 mm

5 mm

2.5 mm

Figure 19. Measurement setup [16].

Lab View

GPIB

Port 1 Port 2

GPIB

V-BandRigid Cable

V-BandRigid Cable

Network AnalyzerAgilent N5250A PNA

Data AcquisitionCard (DAC)

Power Supply

1.5V

5V

Silicon Chip

Silicon Lens

Figure 17. A 60-GHz resonant NMOS switch [16].

T-Line NMOSL = 120 nm

W = 10 × 15 μm

Reflector

60 μm

20 μm

S

D

G

randomly, 20 equally spaced constellation points in thedesired direction are chosen. For the same set of switch-ing combinations, real and imaginary parts of S21 aremeasured at an undesired direction as shown in Figure20. At the undesired direction, constellation points arecompletely scrambled demonstrating the functionality ofthe NFDAM system. Vertical color coding is used toemphasize the scrambling effect.

In another measurement, the optional quadrant-selector unit in conjunction with reflector switching is

used to cover the four quadrants on the signalconstellation diagram (Figure 21). The picture ofthe measurement setup is shown in Figure 22.

To verify the repeatability of S-parametermeasurement, several methods are used. In onecase, real and imaginary parts of S21 are mea-sured for a fixed set of switching combinations1000 times. This measurement was intentionallyperformed over the course of 10 hours to capturethe supply voltage noise, room temperaturechange, and other variations. According to themeasured results, the error in repeatability ismuch smaller than the distance of two adjacentpoints in Figures 20 and 21.

A separate measurement is made to charac-terize the performance of the transmitter. In thismeasurement, the antenna is disconnected fromthe PA (with laser trimming) and a V-band waferprobe is used to capture the output power of the

February 2009 45

Figure 20. Measured constellation points of the switch-based NFDAM chip. In this measurement only the switches areused [16].

Desired DirectionUndesired Direction

Constellation Points Constellation Points

0

−1

156789

10

111213141516

17181920

234

−2

−28.2

1718

11

201410

1316

4

62

19

15

198

73 12

5

−28.6

−29.0

−29.4

−29.8

−1.5 −16.2 −15.6 −15.4 −15.2−15.8−16.0−1 −0.5I(10−4)

Q(1

0−4 )

Q(1

0−4 )

I(10−4)

0 0.5 1

−3

−4

For NFDAM systems, the information beamwidthrepresents the informationdirectionality of the beam while thepower beamwidth represents thepower directionality of the antenna pattern.

Figure 21. Measurement results of the four-quadrant coverage ofthe signal constellation space using the optional quadrant-selectorand switches (switch-based NFDAM chip).

Q

Constellation Points

−3e-4 −2e-4 −1e-4 0 1e-4 2e-4 3e-4

0

−1e-4

−2e-4

−3e-4

1e-4

2e-4

3e-4

I

All Switches Offx

x

transmitter. An Agilent signal source (E8257D) sendsthe input signal to the transmitter input. An Agilent V-band power sensor (V8684A) and power meter(E4418B) are used to measure the output power of thetransmitter. The measured results indicate an outputpower of +7 dBm, a small-signal gain of 33 dB, and asaturated gain of 25 dB.

In a separate measurement, VCO performance ischaracterized by directly connecting the on-chipprobes to the output of an LO buffer. Phase noise andtuning range of the VCO are measured by an Agilentspectrum analyzer (E4448A) and a SpacekLabs off-chipdown-converter (GE-590). This measurement shows atuning range of more than 2.5 GHz and a phase noiseof −100 dBc at 10 MHz offset.

ConclusionsNFDAM systems provide a unique solution for trans-mitting highly secured direction-dependent data andhence preventing eavesdroppers from properlydemodulating the signal. A 60-GHz proof-of-conceptchip was designed and measured.

AcknowledgmentsThe authors appreciate the help of Yu-Jiu Wang ofthe California Institute of Technology and supportof the Lee Center for Advanced Networking at

California Institute of Technology. The technicalsupport from Cadence Design Systems for CADtools and from Agilent Technologies, ZelandSoftware Inc., and Ansoft Corporation is alsoappreciated.

References[1] A. Babakhani, X. Guan, A. Komijani, A. Natarajan, and A. Hajimiri,

“A 77 GHz phased array transceiver with on chip dipole antennas:Receiver and on-chip antennas,” IEEE J. Solid-State Circuits, vol. 41,no. 12, pp. 2795–2806, Dec. 2006.

[2] A. Natarajan, A. Komijani, X. Guan, A. Babakhani, and A. Hajimiri,“A 77 GHz phased array transmitter with local LO-path phase-shifting in silicon,” IEEE J. Solid-State Circuits, vol. 41, no. 12, pp. 2807–2819, Dec. 2006.

[3] B.A. Floyd, S.K. Reynolds, U.R. Pfeiffer, T. Zwick, T. Beukema, andB. Gaucher, “SiGe bipolar transceiver circuits operating at 60 GHz,”IEEE J. Solid-State Circuits, vol. 40, no. 1, pp. 156–167, Jan. 2005.

[4] B. Razavi, “A 60-GHz CMOS receiver front-end,” IEEE J. Solid-State Circuits, vol. 41, no. 1, pp. 17–22, Jan. 2005.

[5] C.H. Doan, S. Emami, A.M. Niknejad, and R.W. Brodersen,“Millimeter-wave CMOS design,” IEEE J. Solid-State Circuits, vol. 40,no. 1, pp. 144–155, Jan. 2005.

[6] D.B. Rutledge, D.P. Neikirk, and D. Kasilingam, “Integrated-circuitantennas,” in Infrared and Millimeter-Waves. New York: Academic,1983, pp. 1–90.

[7] N. Engheta and C.H. Papas, “Radiation patterns of interfacialdipole antennas,” Radio Sci., vol. 17, no. 6, pp. 1557–1566, 1982.

[8] H. Kogelnik, “Theory of dielectric waveguides,” in Integrated Optics,T. Tamir, Ed. New York: Springer-Verlag, 1979, ch. 2, pp. 66–79.

[9] N.G. Alexopoulos, P.B. Katehi, and D.B. Rutledge, “Substrate opti-mization for integrated circuit antennas,” IEEE Trans. MicrowaveTheory Tech., vol. 83, no. 7, pp. 550–557, July 1983.

[10] B. Chantraine-Barès, R. Sauleau, L. Le Coq, and K. Mahdjoubi,“A new accurate design method for millimeter-wave homoge-neous dielectric substrate lens antennas of arbitrary shape,” IEEETrans. Antennas Propagat., vol. 53, no. 3, pp. 1069–1082, Mar. 2005.

[11] T. Nagatsuma, A. Hirata, M. Harada, H. Ishii, K. Machida, T.Minotani, H. Ito, T. Kosugi, and T. Shibata, “Millimeter-wave pho-tonic integrated circuit technologies for high-speed wireless com-munications applications,” in IEEE ISSCC Dig. Tech. Papers, Feb.2004, pp. 448–449.

[12] M.J.M. van derVorst, P.J.I. de Maagt, and M.H.A.J. Herben,“Effect of internal reflections on the radiation properties and inputadmittance of integrated lens antennas,” IEEE Trans. MicrowaveTheory Tech., vol. 47, no. 9, pp. 1696–1704, Sept. 1999.

[13] P. Focardi, W.R. McGrath, and A. Neto, “Design guidelines forterahertz mixers and detectors,” IEEE Trans. Microwave TheoryTech., vol. 53, no. 5, pp. 1653-1661, May 2005.

[14] D.F. Filipovic, G.P. Gauthier, S. Raman, and G.M. Rebeiz, “Off axisproperties of silicon and quartz dielectric lens antennas,” IEEETrans. Antennas Propagat., vol. 45, no. 5, pp. 760–766, May 1997.

[15] A. Babakhani, D.B. Rutledge, and A. Hajimiri, “A near-field mod-ulation technique using antenna reflector switching,” in IEEEISSCC Dig. Tech. Papers, Feb. 2008, pp. 188–189.

[16] A. Babakhani, D.B. Rutledge, and A. Hajimiri, “Transmitterarchitectures based on near-field direct antenna modulation,”IEEE J. Solid-State Circuits, vol. 43, no. 12, Dec. 2008.

46 February 2009

Figure 22. Measurement setup.

NFDAM systems provide a unique solution for transmittinghighly secured direction-dependentdata and hence preventingeavesdroppers from properlydemodulating the signal.