4-2 optical heterodyne detection for 60- ghz-band radio-on … · 2013. 11. 21. · 4-2 optical...

45

1 Introduction

External modulation will provide the bestsolution to configure simpler base stations(BSs) in future microwave or millimeter-wave(mm-wave) radio-on-fober systems, leading tocost-effective system configuration[1]. Wehave reported on radio-on-fiber systems thatuse a 60-GHz-band electroabsorption modula-tor (EAM) for downlinking and uplink-ing[2][3]. To receive optical power largeenough to get high transmission quality, usingsome optical amplifiers is necessary in theoptical link. However, accumulated ampli-fied-spontaneous-emission (ASE) noise fromthe optical amplifiers cannot be removed com-pletely, in analog optical communication sys-tems, such as radio-on-fiber applications. Thenoise fatally affects the photodetected signal,resulting in a degradation in system perform-ance. The theory that optical coherent detec-tion using a remote local light source hashigher-sensitivity to a received optical signalthan direct detection under a shot-noise-limit-ed condition is well known[4]. Thus, we

expect that if we use the coherent detectiontechnique, we can avoid using optical ampli-fiers. Laser phase noise and polarization mis-match cause this degradation if we considerusing a remote local light source for opticalheterodyne detection. Because the polariza-tion mismatch is a common problem for opti-cal fiber systems, we concentrate on usingoptical heterodyne detection that is insensitiveto laser phase noise.

In this paper, we describe a novel methodfor optical heterodyne detection in radio-on-fiber systems at a 60-GHz band. A free-run-ning dual-mode local light is used to detect a60-GHz-band radio-on-fiber signal. We showthat the system, in principle, is not affected bythe phase noises not only for the transmittinglaser but also for the remote local one. Thisscheme is also theoretically immune from thefiber-dispersion effect which causes signalfading, because only two components of theoptical signal are selected out by the locallight to demodulate themselves. This immuni-ty persists even if the transmitted optical sig-nal is in a double-sideband (DSB) format.

Toshiaki KURI and Ken-ichi KITAYAMA

4-2 Optical Heterodyne Detection for 60-GHz-Band Radio-on-Fiber Systems


A novel method for optical heterodyne detection in millimeter-wave-band radio-on-fibersystems is described. The key to detection is to use a remote dual-mode local light.Although the light is free-running, our method is in principle free from laser phase noise.This method is also theoretically immune from the fiber-dispersion effect, because only twocomponents of the optical signal are selected out by the local light to demodulate them-selves. This immunity persists even if the transmitted optical signal is in the doubleside-band format. We derive the theoretical limit of the system performance and experimentallydemonstrate a 25-km-long fiber-optic transmission and the optical heterodyne detection ofa 59.6-GHz radio-on-fiber signal using 155.52-Mb/s DPSK-formatted data.

Keywords Optical heterodyne detection, Dual-mode local light, Radio-on-fiber, Optical fiber dis-persion, Laser phase noise

46

Although some dispersion-free systems[5][6]have been proposed, they have never essen-tially solved the ASE problem. We present aprinciple together with a mathematicaldescription of signals in our method andderive the theoretical limit of the system per-formance. We experimentally verify the prin-ciple of our method by demonstrating a 25-km-long fiber-optic transmission and the opti-cal heterodyne detection of a 59.6-GHz radio-on-fiber signal using 155.52-Mb/s differential-phase-shift-keying (DPSK).

2 Principle

2.1 Mathematical DescriptionFig.1 shows the fundamental block dia-

gram of the optical heterodyne detection usingdual-mode local light source. The transmitterconsists of a single-mode light source andoptical external and electrical modulators.Also, the optical heterodyne receiver consistsof a dual-mode local light source, opticalpower combiner, photodetector (PD), phasenoise canceling (PNC) circuit, and electricaldemodulator.

The complex electric field of the opticalcarrier, ec1(t), which is oscillated from the sin-gle-mode light source in the transmitter, iswritten as

where Ec1, fc1, and φc1(t) are the amplitude, thecenter frequency, and the phase noise of theoptical carrier, respectively. The radio-fre-quency (RF) signal, eRF(t), is written as

where VRF , fRF , and θ(t) represent the ampli-tude, the carrier frequency, and the phase-modulated data of the RF signal, respectively.The optical carrier, ec1(t), is modulated by theRF signal, eRF(t), and transmitted over the opti-cal fiber. Taking into account the fiber-disper-sion effect, after the propagation in the opticalfiber of the length, L, the complex electricfield of the modulated optical signal is gener-ally written as

Here, β(f) is the propagation constant and isapproximated as[7]

Journal of the Communications Research Laboratory Vol.49 No.1 2002

Block diagram of proposed optical heterodyne detection schmeFig.1

47

where β1L corresponds to the group delaytime and β2 is related with the dispersion, D,as

λand c are the wavelength in the fiber and thevelocity in the vacuum, respectively. Notethat the last term in Eq.(6) is constant andindependent of time and the last second termrepresents the chromatic fiber-dispersioneffect. Then, MOD[•] represents the responsefunction of the optical modulator, dependingon the scheme used in the modulator, such asintensity, phase, or frequency. Therefore, an isgiven as the Fourier coeffcient of MOD[•]. Ifthe intensity modulation is performed and themodulation index, mIF, is small, for example,then the Fourier coeffcients are writtenapproximately as[8]

where (2n＋1) !! ≡ 1・3 … (2n＋1), (2n) !! ≡2・4 … (2n), and 0 !! ≡ (－1) !! ≡ 1. Thedetails are described in Appendix A[8]. If theinput amplitude of the modulating signal issmall (VRF≪1), then es(t) is approximated as

The free-running dual-mode local lightthat has a frequency separation of fLO is used todetect the radio-on-fiber signal, es(t, L), in theoptical heterodyne receiver. Here, the dual-mode light source is considered to have a fre-quency separation that is either highly stabi-lized or jitter-free. The complex electric fieldof the dual-mode local light, el(t), is written as

where Ec2, fc2, and φc2(t) are the amplitude, thecenter frequency, and the phase noise of thedual-mode local light, respectively. The dual-mode local light, el(t), is combined with thereceived optical signal, es(t, L), as shown inFig.2. The optical heterodyne detection is per-formed using a square-law response of the PD.Let us assume that the polarizations betweenthe received optical signal and the local lightare matched. Then, the photocurrent becomes

In the above equation, Pc1 and Pc2 are the pho-todetected power of the signal and local light,respectively and are given as[9]

Here, A and Zw represent the photodetectingeffective area and the wave impedance,respectively. Note that is the responsivityof the PD. We focus on the two phase compo-


Optical spectra of signal es(t) andlocal el(t)

Fig.2

48

nents:

where Δf and Δφc (t, L) are defined as

There are two possible configurations forthe PNC, (a) using an electrical square-lawdetector (PNC1) and (b) using an electricalmultiplier (PNC2)[10], as shown in Fig.3. InFig.3 (a), BPF11 in PNC1 filters out two com-ponents of Δf＋fLO/2 and Δf＋fRF－fLO/2simultaneously. They are squared by PNC1.Then, only the down-converted signal at fIF [≡fRF－fLO], i2(t), is filtered out by BPF12.

In PNC2, two components of Δf＋fLO/2 and Δf＋fRF－fLO/2 are filtered out each by BPF21 andBPF22, respectively, as shown in Fig.3 (b).They multiply each other, and only the down-converted signal at fIF , i2(t), is filtered out byBPF23, which is identical to BPF12. Also, inthis case, i2(t) becomes the same signalexpressed in Eq. (23), except that the ampli-tude becomes half. The difference of PNC1

and PNC2 is the amount of the noise put intothe square-law detector or the multiplierbecause of the difference in the bandwidths of

BPF11, BPF21, and BPF22, as shown in Fig.3 (c).This will cause an absolute amount of noisebecause of a noise×noise term included inPNC1 output that is larger than PNC2 output.However, the noise×noise term is usuallynegligible for practical use, in which thesignalto-noise power ratio (SNR) is relativelyhigh so that data can be transmitted with highquality. Therefore, the system performancesfor the systems using PNC1 and PNC2 will bealmost the same.

Based on the above theoretical description,we can describe the features. First, theabsence of laser phase-noise remains in Eqs.(23) and (24). This means that our method isin principle free from the laser phase noise.Second, the last term in Eq. (24) represents thephase delay for the fiber length of L and isconstant. If the RF signal is modulated usingDPSK-encoded data, the fiber-dispersioneffect, which causes the fading of the photode-tected RF signal, does not seriously affect thetransmission quality because only two opticalcomponents, i.e., one single-sideband (SSB)component and carrier, are detected. Note that


PNCs (a) with square-law detector(PNC1) and (b) with multiplier (PNC2).(c) Photodetected signals and pass-bands of BPF11, BPF21, and BPF22

Fig.3

49

this method has the same effect as in the opti-cal SSB-formatted transmission[2]. Third, thefiltering is not performed in the opticaldomain but in the electrical domain to selectthe desired signal components. Our methodwill be able to provide fine and tunable filter-ing for the desired optical components bymaking good use of both the local light andthe electrical BPF. This is because the fre-quency stability and controllability of thelasers have recently progressed in conjunctionwith the development of dense wavelengthdivision multiplexing (DWDM) technologies.

2.2 Theoretical System PerformanceWe will now describe the SNR of the

down-converted signal, i2(t), in the systemusing our detection method. As shown in theprevious section, there was no fiber-dispersioneffect in the demodulation, and therefore, wewill omit the propagation constant, β(f).

Let us rewrite the photodetected signalwith no data (θ(t)≡0) and the local compo-nents in the real part

When the shot noise is dominant, the noisespectral density,η, is given as

If the input put into the multiplier is optimallybandwidth-limited, then the SNR becomes(see Appendix B)

Here,α[= /e] is the sensitivity of the PD.Under the local shot-noise limited environ-

ment (Pc2≫Pc1), the SNR is given as the theo-retical limit.

Let us consider the intensity modulation /direct detection (IM/DD) system, as shown inFig.4[2]. The optical signal is assumed to bemodulated using the same transmitter shownin Fig.1 (a). Then, the two transmitting opti-cal signals that have no data (θ(t)≡0) areexpressed as

where G is the gain of an optical amplifierwith a spontaneous emission coeffcient ofnsp[7]. Using the square-law detection, thephotodetected signal, i3(t), appears at fRF . Thephotocurrent is expressed as

By downconverting with a mm-wave mixerand mm-wave local oscillator of fLO, thedesired signal, i2(t), which is the same signalas that given in Eq. (23), appears. If thedownconverter is assumed to be ideal, theSNR of i2(t) for the conventional systembecomes (see Appendix C)

If the gain in the optical amplifier is largeenough to suppress the other noise compo-nents (G≫1), then the theoretical limit ofSNR is given as

Comparing Eqs. (31) and (36), we can seethat the SNR for our method is improved bythe amount of nsp. Note that nsp represents theASE noise, which is accumulated in the linkand fatally affects system performance.


50

3 Experiment

3.1 Experimental SetupTo verify the principle of our method, we

conducted an experiment using the followingsetup. As shown in Fig.5, the optical systemin the setup consisted of a distributed feedbacklaser diode (LD1), 60-GHz-band electroab-sorption modulator (EAM)[2], 3-dB opticalcoupler, tunable laser diode (LD2), LiNO3

modulator (EOM), erbium-doped fiber ampli-fuer (EDFA), optical isolator, PD, and twopolarization controllers (PCs). The PNCbased on the electrical square-law detectorconsists of two electrical amplifiers and anelectric mixer (the bandwidth of the RF, LO,and IF ports were 5–18 GHz, 5–18 GHz, andDC–3 GHz, respectively).

An optical carrier (fc1) was intensity modu-lated using a 59.6-GHz signal (fRF) by the

EAM. The mm-wave signal was DPSK-encoded at the data rate of 155.52 Mb/s(PRBS = 223－1). An optical local tone (fc2)and the optical signal were combined via the3-dB coupler and optically heterodyne-detect-ed by the PD. The tone was modulated usinga 28.5-GHz [=fLO/2] sinusoidal wave at the DCbias of Vπ, which enables the suppression ofthe optical carrier. The EDFA was used onlyto amplify the optical local tone. The optical-ly heterodyne-detected signal was amplified inthe PNC by the first electric amplifier with thebandwidth of DC–26.5 GHz, was power-divided, put in the electric mixer, and ampli-fied again by the following amplifier (2–4GHz). The 2.6-GHz [= fIF ] IF signal obtainedafter mixing was demodulated by a DPSKdemodulator to extract the data and the clock.

3.2 Experimental ResultsThe measured optical spectrum is shown

in Fig.6. The thick and thin lines represent theoptical signal and the dual-mode local tone infront of the PD, respectively. The wave-lengths of LD1 and LD2 were 1550.27 nm [=c/fc1] and 1550.17 nm [= c/fc2], respectively.The optical-insertion losses of the EAM at abias of －1.5 V and the EOM at a bias of 1.5V were 11 and 25 dB, respectively. The


Conventional IM/DD systemFig.4

Experimental setupFig.5

51

power of fRF applied to the EAM was 6 dBm atthe same bias, corresponding to a modulationdepth of about 44 %. In the optical local tone,the carrier still remained but could be removedusing the present allocation of fc1 and fc2, asshown in Fig.3.

The photodetected signals before and aftermixing were also measured. In Fig.7, the pho-todetected signals appear at around 12.4 GHz[=Δf +fLO/2], 15.0 GHz [=Δf＋fRF－fLO/2],and 16.2 GHz [=－Δf]. Note that fc1＜fc2.Large phase noises caused by the individuallydriven LD1 and LD2 were observed from themeasured linewidths because the interval fre-quency between LD1 and LD2 were not con-trolled. As shown in Fig.7(b), however, a 2.6-GHz IF signal was stably generated. Theobserved linewidth was less than 30 Hz, andthe measured SSB phase noise was less than－73 dBc/Hz at 10 kHz apart from the carrier,despite the linewidths of LD1 and LD2 being 5MHz and 100 kHz, respectively. Hence, wesuccessfully demonstrated the phase-noisecancellation.

We also measured the BERs as a functionof the optical signal power put into the PD.Fig.8 (a) shows the BERs after the 25-km-long SMF transmission and the BERs for theback-to-back, which are plotted as a functionof the optical signal power received by thePD. Optical local power put into the PD was－10.0 dBm. RF input power put into EAM

of fRF was 6.3 dBm. No BER floor wasobserved. The minimum optical signal powerthat was received to get the BER of 10–9 was－16.0 dBm. Compared with the back-to-back case, the small power penalty was pre-sumably due to the polarization matchingerror. When the RF power was below 5.5dBm, the BER floor appeared. This was dueto the ASE noise from the EDFA that boostedthe optical local tone. However the appear-ance of the BER floor can never pose a seriousproblem because it can be circumvented byusing ether a 2-mode DBR-MLLD[11] or a 2-mode injection-locked FP-LD[12] as the dual-mode local light source. Fig.8 (b) shows theBERs as a function of the RF power into theEAM after the 25-km-long SMF transmissionand for the back-to-back. The optical receivedpower put into the PD was fixed to be －16.0dBm. The BER floor did not appear. The


Measured optical spectrumFig.6

Photodetected signals (a) before themixer and (b) after the mixer

Fig.7

52

minimum RF input power to get the BER of10–9 after the 25-km-long SMF transmissionwas also 6.3 dBm. Compared with the back-to-back case, no serious power penalty wasobserved.

4 Conclusion

A novel mothod for optical heterodynedetection in 60-GHz-band radio-on- fiber sys-tems has been described. The key to detectionwas to photodetect the radio-on-fiber signal byusing a dual-mode local light source at the

receiver side. The principle and mathematicaldescription for the detection moethod has beentheoretically described. The free-running dual-mode local light was used to detect a radio-on-fiber signal at a 60-GHz band . When the pho-todetected signal and reference componentsmultiplied each other, the same amount of thelaser phase noise in the components were dif-ferentially subtracted. Therefore, our methodwas in principle free from the laser phasenoise. This scheme was theoretically immunefrom the fiberdispersion effect, which causessignal fading because only two components ofthe optical signal are selected out by the locallight to demodulate themselves. This immuni-ty persisted even if the transmitted optical sig-nal was in the DSB format. The theoreticallimit of the system performance has also beenderived. Compared with the conventionalIM/DD system using optical amplifiers, weclarified that the proposed detection scheme issuperior to the conventional one. To confirmthe principle of the proposed scheme, the 25-km-long fiber-optic transmission and the opti-cal heterodyne detection of a 59.6-GHz,155.52-Mb/s DPSK radio-on-fiber signal hasbeen successfully demonstrated. No seriouslaser phase noise and fiber-dispersion effectswere occurred.


BER: (a) vs. optical signal power and(b) vs. RF power

Fig.8

References1 K. Kitayama, "Architectural considerations of fiber-radio millimeterwave wireless access systems,"

Fiber and Integrated Optics, Vol. 19, pp. 167–186, 2000.

2 T. Kuri, K. Kitayama, A. Stöhr, and Y. Ogawa, "Fiber-optic millimeterwave downlink system using 60

GHz-band external modulation," J. Lightwave Technol., Vol. 17, No. 5, pp. 799-806, May 1999.

3 T. Kuri, K. Kitayama, and Y. Ogawa, "Fiber-optic millimeter-wave uplink system incorporating

remotely fed 60-GHz-band optical pilot tone," IEEE Trans. Microwave Theory and Tech., Vol. 47,

No. 7, pp. 1332-1337, July 1999.

4 T. Okoshi, K. Emura, K. Kikuchi, and R. Th. Kersten, "Computation of bit-error-rate of various het-

erodyne and coherent-type optical communications schemes," J. Opt. Commun., Vol. 2, No. 3, pp.

89–96, Sep. 1981.

5 D. Novak, Z. Ahmed, G. H. Smith, and H. F. Liu, "Techniques for millimeter-wave optical fiber trans-

mission systems (invited)," in Topical Meeting on Microwave Photonics (MWP’97) Tech. Dig., Duis-

brug, Germany, TH1-3, Sep. 1997, pp. 39–42.

6 R.-P. Braun, G. Grosskopf, D. Rohde, and F. Schmidt, "Fiber optic millimeter-wave generation and

53Toshiaki KURI and Ken-ichi KITAYAMA

bandwidth efficient data transmission for broadband mobile 18–20 and 60 GHz-band communica-

tions," in Topical Meeting on Microwave Photonics (MWP’97) Technol. Dig., Duisbrug, Germany,

FR2-5, pp. 235–238, Sep. 1997.

7 G. P. Agrawal, Nonlinear optics, Second Edition, Secs. 2 to 4, Academic Press, 1995.

8 M. Suzuki, Y. Noda, and Y. Kushiro, "Characterization of a dynamic spectral width of an

InGaAsP/InP electroabsorption light modulator," Trans. IEICE, Vol. E69, No. 4, pp. 395–398, Apr.

1986.

9 R. M. Gagliardi and S. Karp, Optical communications, Second Edition, Sec. 1, John Wiley & Sons,

Inc., 1995.

10 R. Gross, R. Olshansky, and M. Shumidt, "Coherent FM-SCM system using DFB lasers and a phase

noise cancelling circuit," IEEE Photon. Technol. Lett., Vol. 2, No. 1, pp. 66–68, Jan. 1988.

11 T. Ohno, S. Fukushima, Y. Doi, Y. Muramoto, and Y. Matsuoka, "Application of uni-travelling-carrier

waveguide photodiode in base stations of a millimeter-wave fiber-radio system," in Tech. Dig.

MWP’99, Melbourne, Australia, F-10.2, pp. 253–256, Nov. 1999.

12 M. Ogusu, K. Inagaki, and Y. Mizuguchi, "60 GHz-band millimeterwave generation and ASK data

transmission using 2-mode injectionlocking of a Fabry-Perot slave laser," in Tech. Dig.

TSMMW2000, Yokosuka, Japan, P-12, pp. 181–184, Mar. 2000.

13 K. Shimoda, H. Takahashi, and C. H. Townes, "Fluctuation in amplification of quanta with application

to maser amplifiers," J. Phys. Soc. Japan, Vol. 12, No. 6, pp. 686–700, June 1957.14 H. Ishio, K. Nakagawa, M. Nakazawa, K. Aida, and K. Hagimoto, Optical amplifier and its applications,

Sec. 2, Ohm, 1992 (in Japanese).

15 S. Betti, G. D. Marchis, and E. Iannone, Coherent optical communications systems, Sec. 4, John

Wiley & Sons, 1995.

54 Journal of the Communications Research Laboratory Vol.49 No.1 2002

appendixA Intensity Modulator

If the intensity modulation is performed,then MOD[eRF (t)] is expressed as[8]

where mIF is the intensity-modulation index asa function of VRF , andγrepresents the intensitymodulation with the chirp effect, which isdefined as

αc is the chirp parameter of the optical modu-lator. We carried out the Fourier expansion ofMOD[eRF (t)] by using the Binominal theorem.As a result, we got an, expressed as

whereΓ(•) is the Gamma function, which hasthe following relationship.

If the chirp is negligible (α= 0), then an issimplified to

where (2n＋1) !! ≡ 1・3 … (2n＋1), (2n) !! ≡2・4 … (2n), 0 !! ≡ (－1) !! ≡1, and (－2n－1) !! ≡ (－1) n/(2n－1) !! . Here, we used thefollowing formula for the natural number of n.

If mIF is small, then the Fourier coefficients are

approximately given by

To verify the above calculation, we also com-pared α±n for mIF = 1 with the following ana-lytical solution.

As a result, an numerically agreed well withthe analytical solution, ân.

B SNR for Proposed Detection Tech-niqueThe photodetected signal and local compo-

nents are represented by Eqs. (25) to (28). Thenoise spectral density at the output of the PD,η, is generally written as

where [= e・α] is the responsivity of thephotodetector. The responsibity is assumed tobe the white noise, which in order, consists ofthe shot noises for the signal and the locallight, the dark current of the photodetector(ID), and the thermal noise (2kBT/RL). For thethermal noise, kB, T, and RL are the Boltzmannconstant, the temperature, and the load resist-ance, respectively. When the shot noise isdominant, the noise spectral density,η,isapproximated as

s(t), l(t), and n(t) are assumed to be statistical-ly independent of each other, and their aver-ages are all zero. For the square-law detectorin PNC1, the self-correlation becomes

55

Here, E [•] represents the ensemble average,and Rx,y (τ) is the correlation function of x (t＋τ) and y(t) and is defined as

Rn2,n2(τ) is neglected because it is much small-er than 4 {Rs,s(τ)＋ Rl,l(τ)}・Rn,n(τ). In thiscase, our interest terms are the down-convert-ed component, 4Rs,s(τ)・Rl,l(τ), and the noisecomponents, 4 {Rs,s(τ)＋Rl,l(τ)}・Rn,n(τ).They become

where δ(•) represents the delta function.From the above calculation, under the shot-noise limited environment, the SNR for theproposed detection becomes

where BIF is the bandwidth of BPF12. If theinput put into the multiplier is optimally band-width-limited, then the noise power is halved.Therefore, the SNR is rewritten as

Under the local shot-noise limited environ-ment (Pc2≫Pc1), moreover, the theoretical limitof SNR is expressed as

In the same manner, the SNR for PNC2 isderived and acquires the same result as theabove SNR.

C SNR for IM/DDTwo transmitting optical signals are

assumed to have no data (θ(t) ≡ 0) andexpressed using Eqs. (32) and (33), respec-tively. In this case, because the photocurrent isgiven by Eq. (34), the signal power, Ps,becomes

Here, G is the gain of an optical amplifier witha noise figure of nsp, and RL is the load resist-ance. On the other hand, the noise from thephotodetector is expressed as[13][14][15]

In the above equation, the components are inorder the shot noise for the signal and local,the dark current, the amplified spontaneousemission (ASE) shot noise, the (signal＋local)×ASE term, the ASE×ASE term, and the rel-ative intensity noise (RIN), respectively.When the signal and local shot noises aredominant, Pn is approximated as


56

From the above, the SNR for the conventionalIM/DD is approximated as

If the gain of the optical amplifier is largeenough to suppress the other noise compo-nents (G≫1), then the theoretical limit ofSNR is given as


Toshiaki KURI, Ph. D.

Senior Researcher, OptoelectronicsGroup, Basic and Advanced ResearchDivision

Optical communication systems

Ken-ichi KITAYAMA, Dr. Eng.

Professor, Department of Electronicand Information Systems, Osaka Uni-versity

Photonic network

4-2 optical heterodyne detection for 60- ghz-band radio-on … · 2013. 11. 21. · 4-2 optical...

Documents