IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 47, NO. 7, JULY 1999 1315

Optical Up-Conversion on Continuously VariableTrue-Time-Delay Lines Based on Chirped

Fiber Gratings for Millimeter-WaveOptical Beamforming Networks

Juan L. Corral, Javier Martı, Member, IEEE,and Jose M. Fuster,Member, IEEE

Abstract—An electrooptical up-converting modulation schemehas been considered on chirped-fiber-grating-based (CFG) true-time-delay lines for optical beamforming networks. A rigorousanalytical study about the effect of both up-conversion processand CFG dispersion on the amplitude and delay of detectedsignal is presented. Simulations and measurements show that thebandwidth limitation of the delay line due to the fiber gratingdispersion is clearly improved if compared to conventional inten-sity modulation scheme, while the time-delay performance on a2–30-GHz bandwidth is maintained (�max = �1 ps). This newconcept is used on an optical beamforming network architecturefor a millimeter-wave (15–25 GHz) phased array antenna showinga promising performance.

Index Terms—Chirped fiber grating, millimeter-wave photon-ics, optical beamforming, phased-array antennas, photonic delaylines.

I. INTRODUCTION

CHIRPED fiber gratings (CFG’s) have shown very inter-esting features as continuously variable true-time-delay

(TTD) elements in optical beamforming networks for mi-crowave and millimeter-wave phased-array antennas (PAA’s)[1]–[2]. In spite of the excellent performance as compact-size TTD lines, some limitations in the available bandwidthdue to the CFG dispersion have been previously reported [3],compromising its use on millimeter-wave systems. Single-sideband plus carrier (SSBC) generators based on dual-drive optical modulators have been proposed to overcomethe dispersive attenuation on the detected signal [3], butthe dispersion phase term still has to be compensated afterdetection. On the other hand, electrooptical harmonic up-converting schemes are expected to reduce the requirements onthe bandwidth of lasers or external modulators in millimeter-wave applications [4]–[5]. It has recently been proven thatthe optical up-conversion of an intermediate frequency (IF)modulated (either directly or externally) optical carrier bymeans of a Mach–Zehnder electrooptical modulator (MZM)driven by a strong local oscillator (LO) signal and biased atits minimum transmission bias (MTB) point, sharply reduce

Manuscript received October 7, 1998; revised March 10, 1999. This workwas supported by the Spanish Research and Technology Commission (CICYT)under Project TIC96-0611.

The authors are with the Departamento de Comunicaciones, UniversitatPolitecnica de Valencia, 46022 Valencia, Spain (e-mail: jlcorral@upv.es).

Publisher Item Identifier S 0018-9480(99)05201-1.

the chromatic dispersion effects in standard fiber-optic links[4]. This MTB up-converting scheme is applied in this paperto CFG-based TTD lines. The new concept presented inthis paper is proven on a optical beamforming architectureable to drive millimeter-wave PAA’s with total control onboth amplitude and phase distribution on array elements.In Section II, the principle about the proposed optical up-conversion scheme is introduced and compared with a conven-tional self-heterodyne modulation. Section III offers a rigoroustheoretical analysis of sensibility of the proposed CFG-basedTTD line with optical up-conversion to the CFG dispersionand up-conversion process. The results from the experimentscarried out are shown in Section IV.

II. PRINCIPLE

In Fig. 1, the proposed beamformer architecture is depicted;this beamformer is based on a previously presented opticalbeamforming network [2], where beamforming is achieved bytuning different narrow-band lasers, whose optical carriersare combined and modulated by the radio-frequency signal onone MZM. After passing all the modulated carriers through oneCFG, each carrier will suffer different delay according to itsrespective wavelength and CFG group delay response. Finally,wavelength demultiplexing will separate the carriers inorder to separately drive the antenna elements or subarrays.Before the detection, some extra length of standard fiber isadded on each beamformer branch in order to correct CFGdelay associated with each center wavelength position [2].Optical power and wavelength tuning on each laser source willprovide full control on PAA amplitude and phase distributions.Each branch of the beamformer is a continuously variableTTD line with optical up-conversion and with one CFG asa dispersive element, as is shown in Fig. 2. An IF electricalsignal is up-converted to a millimeter-wave frequency bymeans of two optical processes. First, the optical carrier isintensity modulated (either directly or externally) with theIF electrical signal. In the second stage, the IF signal is up-converted through an MZM driven by an LO signal (whoseangular frequency is ) and biased at the MTB point. Afterthe MZM, the optical signal is basically composed of twooptical carriers separated by and both intensitymodulated by the IF signal. As both carriers are IF modulated,

0018–9480/99$10.00 1999 IEEE

1316 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 47, NO. 7, JULY 1999

Fig. 1. Transmitting-mode block diagram of the TTD optical beamformer with one CFG and electrooptical up-conversion.

Fig. 2. CFG time-delay unit with a pseudo-self-heterodyne modulation scheme.

not true self-heterodyne (SH), but a pseudo-self-heterodynemodulation scheme (PSH) is achieved. The LO signal levelwill typically be chosen to maximize the level of both opticalcarriers. Finally, the optical signal passes through a linearlyCFG that introduces a time delay that may be varied by tuningthe optical wavelength of the laser. After detection, someelectrical filtering stage will separate the desired componentat .

In a true SH optical modulation technique, only one of theoptical carriers would be modulated. The generation of the SHspectrum would imply either two phase-locked optical sources[6] or one single source and some kind of optical carrierfiltering before IF modulation [7]. The first option is clearlyhandicapped by the need of two different optical sources foreach TTD line, while the second one can no longer be usedwhen the optical carrier is tuned, as in continuously variableCFG-based TTD lines [1]–[3].

Besides its easier generation, if compared to the SH tech-nique, the PSH modulation shows a different behavior whenthe optical system is dispersive. The dispersive medium willbe characterized by its parabolic phase response as

(1)

which is a typical phase response for a linearly chirped CFGwith group delay slope(ps/nm) or equivalent length,(km),of standard fiber with dispersion parameterps/(nmkmwhose group delay is

(2)

where is the group delay at due to thedispersive medium, and are the angular frequency andwavelength excursion from , and depends on orproduct according to

(3)

where is the center wavelength andis the speed of lightin vacuum. According to Fig. 2, if the CFG is defined by (1),the detected signal term at will ideallycome from the combination of two beat signals; one of themis the beat signal of the optical components atand and the second term will come as thebeat product of the optical components atand . Both beat signals independently correspondto the detected signal when SH modulation is used, which canbe expressed as

(4)

where the sign uncertainty depends on which beat signal ischosen.

From (2), it can be stated that the detected signal whenSH modulation is used will follow the time delay and dis-persion behavior of the dispersive medium with no dispersiveattenuation.

For PSH modulation, both beat products will be present,showing the same time delay, but inverse dispersion term.After adding both terms, the detected signal at

CORRAL et al.: OPTICAL UP-CONVERSION ON CONTINUOUSLY VARIABLE TTD LINES BASED ON CFG’S 1317

can be expressed as

(5)

If (4) and (5) are compared, PSH modulation technique isexpected to have no dispersion term in the phase of thedetected signal. On the other hand, a dispersive attenuationfactor is introduced.

III. A NALYSIS OF A CFG-BASED TTD LINE

A. General Expression

(6)

where is the optical modulation index at MZM. Thegeneral result in (6) will approximate the ideal PSH result in(5) only if is low enough to neglect all the terms in (6), butthe first one.

B. LO Low-Level Approximation

Assuming that the up-converting MZM is driven by a low-level LO signal, second-order terms in (6) are negligible andthe component of the detected signal atmay be expressed as

(7)

As expected, the PSH technique from Fig. 2 will lead to adispersion-free detected signal with a dispersion-dependentamplitude, as shown in (7). The time delay in (7)could be varied by tuning the laser wavelength according to(2).

When conventional CFG-based delay line (DSBC) mod-ulation is used, the dispersive attenuation factor in (7) willdepend on the factor instead of , reducing theavailable bandwidth of the TTD unit [3]. For instance, a CFG-based TTD unit with a group-delay slope of ps/nmand a data signal at GHz will show a 3-dBbandwidth of 6.4 GHz with DSBC modulation, but 39 GHzwith the PSH technique.

C. TTD Sensibility to LO Level

Equation (6) shows that the LO signal level will be lim-ited by the relative level of second-order terms, which aredependent on the optical modulation index at the up-converter

Fig. 3. Magnitude of detected power ripple at!RF = 2!LO + !IF

component against optical modulation index at optical up-converter. Solidline: second order approximation from (9). Circles: full expression resultsfrom (6).

. If the first two terms are considered, the detected signalcomponent at will be

(8)

thus, as the LO signal level grows, two main effects show.Firstly, even when no major dispersion is present, maximumdetected power is not reached when is maximum

, but just for , with 1.6-dBreduction in LO power input and 3 dB in detected power.Secondly, it appears as a ripple in amplitude of the detected RFcomponent, when the LO frequency is swept. This amplituderipple is due to the CFG dispersion and its magnitude can becalculated as

(9)

which is result that is plotted against the LO modulationindex in Fig. 3. From Fig. 3, one could notice that a 1-dBripple specification around full RF bandwidth would implya maximum modulation index of with 7.2-dBreduction in LO power input and 8.2 dB in detected powerfrom the maximum expected throughput from (7).

Concerning the time-delay performance, (6) and (8) showthat no dependence with the LO signal level could be expectedfrom the PSH technique. Furthermore, (6) implies that theRF detected current would follow the CFG delay responseaccording to (2) with no secondary effects expected from theLO signal level or IF and RF frequency values.

1318 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 47, NO. 7, JULY 1999

Fig. 4. RF power degradation simulations and measurements of DSB+C(m0 = 0:1, �0 = 1547:5 nm) and PSH (mIF = 0:1, mLO = 0:6,�0 = 1547:5 nm) modulations. (i) Solid line: DSB+C measurements.Dashed line: DSB+C simulation. (ii) Dotted points: PSH measurements[(�; fIF = 500 MHz), (�; fIF = 250 MHz), (�; fIF = 1 GHz)]. Dashedline: PSH simulation (fIF = 1 GHz).

IV. EXPERIMENT AND RESULTS

A. Amplitude Response

The CFG-based TTD line with PSH modulation, depictedin Fig. 2, has been measured for three different values ofIF frequency ( MHz, MHz, GHz). The CFGused is a 40-cm-long apodized linearly chirped CFG with anaverage group delay slope of ps/nm and almost flatreflectivity from 1547 nm to 1551 nm. The modulation indexesof both IF modulation and up-conversion processes were

and , thus, just a 0.1-dB ripple on thedetected power would be expected from (9). The LO frequencywas varied in 1-GHz steps from 1 to 9 GHz. Both PSHand DSB C modulation schemes have been measured andthe already calibrated results (measurements and simulations)are shown in Fig. 4. The dispersive attenuation shown in theDSB C case is fully overcome when using the PSH technique,as expected from (7). From Fig. 4, it may be observed that thePSH measurements show an almost constant slope growth overthe slight decay from the simulations. This behavior is due tothe laser chirp index, which was not considered in (7).

B. Time-Delay Response

In order to verify the time-delay behavior described in (2),i.e., a constant group delay for all frequencies and a lineardependency on the laser wavelength, the phase of the detectedsignal component at has been estimatedas a function of detected frequency from actual measurements(amplitude and delay) of the 40-cm CFG. Other elements inFig. 2 are assumed to show ideal performance, as expressed in(6). In Fig. 5, the relative detected phase for three different IFfrequencies and at five different optical wavelengths (

nm, nm, nm, nm, nm)inside the CFG bandwidth are plotted, showing the expectedlinear response against detected frequency, with differentslopes according to the relative CFG group delay at each wave-

Fig. 5. Phase term at!RF = 2!LO + !IF component of detected signalagainst detected frequency in a CFG-based TTD unit with PSH modulation(mIF = 0:1;mLO = 1) at five different optical wavelengths. Solid line:fIF = 1 GHz. Dashed line:fIF = 250 MHz. Dotted line:fIF = 500 MHz.

(a)

(b)

(c)

(d)

Fig. 6. Time-delay ripple at!RF = 2!LO+!IF against detected frequency(fRF) at four different optical wavelengths. (a)�1 = 1547:5 nm. (b)�2 = 1548:5 nm. (c)�3 = 1549:5 nm. (d)�4 = 1550:5 nm.

TABLE IRMS RIPPLE IN TIME-DELAY RESPONSE OF ACFG-BASED TIME-DELAY UNIT

WITH PSH MODULATION WHEN DETECTED FREQUENCY RANGES FROM

2 TO 30 GHz. RESULTS FORTHREE DIFFERENT IF VALUES AND

FOUR DIFFERENT WAVELENGTHS

length ( ps, ps, ps, ps, ps).In order to check the time-delay independence against detectedfrequency, the ripple around the ideally constant group delayhas been calculated at four different wavelengths inside CFGbandwidth. The results are plotted in Fig. 6 and the root meansquare (rms) values are summarized in Table I, showing aworst-case standard deviation of just1 ps from 2 to 30 GHz.

CORRAL et al.: OPTICAL UP-CONVERSION ON CONTINUOUSLY VARIABLE TTD LINES BASED ON CFG’S 1319

(a)

(b)

(c)

Fig. 7. Array factor for aN = 7 elements PAA with scanning angles� = 0�

(solid line), � = 20� (dashed line), and� = 40� (dashed-dotted line). (a)fLO = 7 GHz, fIF = 1 GHz, fRF = 15 GHz. (b) fLO = 9:5 GHz,fIF = 1 GHz, fRF = 20 GHz. (c) fLO = 12 GHz, fIF = 1 GHz,fRF = 25 GHz.

C. Beamforming

The performance of the beamformer from Fig. 1 has beenanalyzed by assuming a seven-isotropic-element PAA withinterelement spacing of mm (free-space wavelengthat GHz). Seven equally spaced optical carriersinside the CFG bandwidth ( nm, nm,

nm, nm, nm,nm, nm) have been assigned to every one of theseven array elements. The beamformer has been simulated forthree different detected frequencies ( GHz, GHz,

GHz) and an IF of GHz. In the simulations, the 40-cm-long CFG amplitude and phase measurements have beenincluded and the PSH modulation is fully characterized by (6).The array factor is shown in Fig. 7 for three different beampositions , , and for the three differentdetected frequencies. At each beam position, each laser wave-length is tuned to achieved the respective time delay at eachantenna element according to group-delay slope around eachcenter wavelength, . The beam-pointing accuracy in Fig. 7is really good for all three radio-frequency (RF) frequencies,while the sidelobes level keeps around10 dB.

V. CONCLUSION

A PSH modulation technique with optical up-conversion hasbeen proposed as a solution for millimeter-wave TTD unitsbased on CFG. The measurements results from Section IVprove the expected features of the PSH modulation tech-nique sketched in Section II and fully analyzed in Section III.Available bandwidth for CFG-based TTD lines is clearlyimproved if compared with conventional DSBC modulation.On the other hand, time-delay performance is maintained,with no phase dispersion at the detected signal if comparedto SH or SSB C modulation. The MZM drive level at theup-conversion stage has been shown to clearly impact the

ripple on detected power, but this effect is fully characterizedin Section III and it would imply a 5–10-dB reduction onmaximum detected power, depending on specified detectedpower ripple. Up-converter drive level influence on otherharmonics at detected signal outside the RF signal bandwidthhas not been considered, as they would be rejected by thefinal electrical filtering stage. Concerning the use of the CFG-based TTD line with PSH modulation on PAA’s, the resultsin Section IV for the optical beam-forming network (OBFN)architecture in Fig. 1 confirms the expected good performancefor millimeter-wave arrays if compared to other modulationtechniques (DSBC, SH or SSB C). The rise on sidelobeslevel in Fig. 7 (4 dB higher than expected) are due to CFGreflectivity ripple and group-delay ripple around ideally linearCFG group-delay response and its effect could be reduced ifthe CFG would be calibrated in advance.

ACKNOWLEDGMENT

The authors acknowledge the Optoelectronics ResearchCentre (ORC), Southampton, U.K., for supplying the wide-band CFG.

REFERENCES

[1] J. E. Roman, M. Y. Frankel, P. J. Matthews, and R. D. Esman, “Time-steered array with a chirped grating beamformer,”Electron. Lett., vol.33, no. 8, pp. 652–653, 1997.

[2] J. L. Corral, J. Marti, S. Regidor, J. M. Fuster, R. I. Laming, and M. J.Cole, “Continuously variable true time delay optical feeder for phasedarray antenna employing chirped fiber gratings,”IEEE Trans. MicrowaveTheory Tech., vol. 45, pp. 1531–1536, Aug. 1997.

[3] J. L. Corral, J. Marti, J. M. Fuster, and R. I. Laming, “Dispersion-induced bandwidth limitation of variable true time delay lines basedon linearly chirped fiber gratings,”Electron. Lett., vol. 34, no. 2, pp.209–211, 1998.

[4] J. M. Fuster, J. Marti, and J. L. Corral, “Chromatic dispersion effects inelectro-optical up-converted millimeter-wave fiber optic links,”Electron.Lett., vol. 33, no. 23, pp. 1969–1970, 1997.

[5] J. Park, M. S. Shakouri, and K. Y. Lau “Millimeter-wave electro-opticalup-converter for wireless digital communications,”Electron. Lett., vol.31, pp. 1085–1086, 1995.

[6] U. Gliese, T. N. Nielsen, N. Bruun, E. L. Christensen, K. E. Stubkjaer,S. Lindgren, and B. Broberg, “A wide-band heterodyne optical phase-locked loop for generation of 3–18-GHz microwave carriers,”IEEEPhoton. Technol. Lett., vol. 4, pp. 936–938, Aug. 1992.

[7] R. Hofstetter, H. Schmuck, and R. Heidemann, “Dispersion effectsin optical millimeter-wave systems using self-heterodyne method fortransport and generation,”IEEE Trans. Microwave Theory Tech., vol.43, pp. 2263–2269, Sept. 1995.

Juan L. Corral was born in Zaragoza, Spain, on April 20, 1969. He receivedthe Ingeniero de Telecomunicacion degree (first-class honors) and DoctorIngeniero de Telecomunicacion degree from the Universitat Politecnica deValencia, Valencia, Spain, in 1993 and 1998, respectively.

In 1993, he was an Assistant Lecturer in the Departamento de Comunica-ciones, Universitat Politecnica de Valencia. From 1993 to 1995, he was withthe Microwave Technology & Equipment Section (XRM), European SpaceResearch & Technology Centre (ESTEC), European Space Agency (ESA),where he was engaged in research on monolithic-microwave integrated-circuit-based (MMIC-based) technologies and photonics technologies foron-board PAA’s. Since 1995, he has been a Lecturer in the Departamentode Comunicaciones, Universitat Politecnica de Valencia. His research interestsinclude phased-array antennas, optical beamforming networks, and microwaveand millimeter-wave optical-fiber systems.

1320 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 47, NO. 7, JULY 1999

Javier Mart ı (S’89–M’92) received the Ingeniero Tecnico de Telecomuni-cacion and Ingeniero de Telecomunicacion degrees from the UniversidadPolitecnica de Catalunya, Catalunya, Spain, in 1988 and 1991, and the DoctorIngeniero de Telecomunicacion degree from the Universidad Politecnica deValencia, Valencia, Spain, in 1994.

During 1989 and 1990, he was an Assistant Lecturer at the Escuela Uni-versitaria de Vilanova, Barcelona, Spain. In 1991, he joined the Departamentode Comunicaciones, Universidad Politecnica de Valencia. From 1991 to 1994,he was a Lecturer of the Telecommunication Engineering Faculty and, since1995, he has been an Associate Professor. He is currently the head of the Radioover Fiber Group. He has published over 60 papers in referred internationaltechnical journals and 30 papers in international conferences in the fieldsof fiber-radio and microwave/millimeter-wave photonics, wavelength divisionmultiplexing (WDM) and subcarrier multiplexing (SCM) lightwave systems,optical processing of microwave signals, dispersion and fiber nonlinearitiescompensation employing fiber gratings and other techniques, and mobileand satellite communication systems. His current technical interest includemillimeter-wave fiber-radio systems, dense WDM (DWDM) broad-band op-tical communication networks, optical external modulators, fiber gratings,semiconductor optical amplifiers, and planar lightwave circuits. He has servedas a reviewer for several Institution of Electrical Engineering journals, and anAdvisory Board member for theJournal of Fiber and Integrated Optics.

Dr. Martı has served as reviewer of several IEEE journals and leadingconferences. He has served as a member of the Technical Program Committeeof the European Conference on Optical Communications, International TopicalMeeting on Microwave Photonics, and other international workshops andconferences.

Jose M. Fuster (S’84–M’86) was born in Tarragona, Spain, on March23, 1970. He received the M.Sc. and Ph.D. degree in telecommunicationsengineering from the Universidad Politecnica de Valencia, Valencia, Spain, in1993 and 1998, respectively.

Since 1993, he has been a Lecturer in the Departamento de Comunicaciones,Universidad Politecnica de Valencia, and is a member of the Fiber RadioGroup (FRG). His research interests include microwave and millimeter-waveoptical-fiber links, phased-array antennas, harmonic photonic mixing, andelectrooptical modulators.