1024

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. LT-2, NO.

6 , DECEMBER 1984

Degradation of Bit-Error Rate in Coherent Optical Communications Dueto Spectral Spread of the Transmitter and the Local Oscillator
KAZURO KIKUCHI,TAKANORI OKOSHI, FELLOW,IEEE, MASATO NAGAMATSU, AND NAOYA HENMI, STUDENT MEMBER, IEEE

Ahstruct -In coherent optical communication schemes, the FM noises of the transmitter and/or the local oscillator seriously deteriorate the receiver sensitivity. A theory is developed describing the effect of the FM noise on the bit-error rate (BER)performance of various heterodyne and homodyne schemes. To verify the theory, the BER of a phase-shift keying (PSK) heterodyne system is measured. The feasibility of the PSK scheme using semiconductor lasers is also discussed on the basis of analysis and experimentation.

signal is homodyne-detected by using a local laser, which is phase-locked to the transmitted carrier. The baseband modulation signal is directly restored from the homodyne mixer. The heterodyne coherent detection scheme and the homodyne scheme are usually called generically coherent optical communication schemes. Theoretically, the coherent schemes have better receiver sensitivity than the noncoherent schemes [3], [4], if the FM noises of the I. INTRODUCTION transmitter and/or the local oscillator can be neglected. HE optical communication systemsnowbeingused However, experiments show [ 5 ] , [6] that when we use employ the intensity-modulation/direct-detection semiconductor lasers, which have large FM noise, the scheme. This scheme has a great advantage in system receiver sensitivity degrades from the predicted by the simplicity and low cost, whereas the good temporal simple theory. However, no theory has been presented to coherence of lasers is not utilized. give the degradation quantitatively. On the other hand, heterodyne and homodyne optical The paper first presents a theory on the BER of coherent communications are beingdevelopedto improve the ;e- optical communication systemswith a transmitter and a ceiver sensitivity up to the shot-noise limit [l], [2]. In these local oscillator having FM noises. Next, to verify the schemes, any of the amplitude, frequency, or phase of the theoretical prediction, the bit-error rate (BER) heterodyne carrier light from the transmitter laser can be modulated. coherent of a phase-shift keying (PSK) system is measured In the heterodyne system [l], the transmitted signalis asafunction of the S / N ratio at the IF stage.In the optically heterodyne-detected by using a local oscillator experiment, a controlled amount of phase fluctuation is (LO) laser. The frequency difference between the trans- given to the signal. Finally, the feasibility of a PSK scheme mitter and LO lasers is kept constant by an AFC scheme. using semiconductor lasers as the transmitter and the local Finally,the intermediate frequency (IF) signal thus ob- oscillator is discussed on the basis of analysis and experitained is demodulated electronically. mentation. To demodulate the IF signal and restore the baseband modulation signal, two methods are available [2]: 1) the 11. COMPUTATION OF BIT-ERROR RATE IN VARIOUS noncoherent (envelope) detection scheme and 2) the coherCOHERENT OPTICAL COMMUNICATION SCHEMES ent detection scheme. In the noncoherent detection scheme, the good temporal coherence of lasers is not utilized, because the power of the IF signalisused for the level A . Formulation of the Problem discrimination. On the other hand, in the coherent scheme, In the following, the IF (or baseband, in homodyne case) the phase information of the IF signal is indispensable in signal and noise components are all expressed asthe equivthe demodulation process. alent input currents at the IF amplifier input. We consider The homodyne optical communication systemuses a first the heterodyne scheme. The IF signal (including noise) more sophisticated detection technique [2]. The transmitted can, generally, be expressed as [3]
Manuscript received November 15, 1983; revised April 13, 1984. The authors are with the Department of Electronic Engineering, University of Tokyo 7-3-1 Hongo, Bunkyo-ku, Tokyo 113, Japan.

i(t)=S(t)cos(2rfIFt+6(~)+~n(~))+x(t)cos2~fIFt

+ y ( t ) sin2rfIFt

(1)

0733-8724/84/1200-lO24$01.00 01984 IEEE

KIKucm et al.: DEGRADATION OF BIT-ERROR RATE

1025

signal is also expressed as (6). However, in this case A+*( t ) denotes the phase difference between the transmitted carrier and the LO output. A phase tracking technique such as optical PLL (phase-Tocked loop) or injection loclung must be employed to suppress the phase error A c $ ~ , ( ~below ) a 'prescribed value. We should note that the bandwidth required for the homodyne detection is half of that required for the heterodyne detection. Therefore, the receiver sensiwhere P,(t) denotes the optical signal power, PL the local tivity is improved by 3 dB as compared with the heterooscillator power, 7 the quantum efficiency of the photode- dyne scheme [3]. tector, v the optical frequency, e the electron charge, and The variance : a in both the heterodyne and homodyne h the Planck's constant. The noise components x(t) and schemes is closely related to the spectral width of lasers. y ( t ) canbe regarded as whte Gaussian noiseswith a This relation will be given in Section 111. In this section, we variance a calculate BERs of various coherent schemes for given values of a+, as a function of the S / N ratio y defined as ( x ( t ) > = ( Y ( t ) > = 0, (3) S2 ( x ' ( t ) > = ( y ' ( t ) > = a2. (4) where S ( t ) denotes the amplitude signal, B ( t ) the phase signal, +n(t)the phase fluctuation, ,f,, the intermediate frequency, and x( t)and y ( t ) noise components. The amplitude S ( t ) is expressed, in terms of the optical signal, as [ 3 ]
y = . ,

When the shot-noise-limited condition is achieved, the variance of the equivalent input noise current is given as [3] a2=2e-PLB ell

2a

B. Heterodyne Coherent ASK and Homodyne ASK Schemes

We consider a two-level ASK scheme with amplitudes 0 and So. The demodulator outputs in the mark and space states are given from (6), respectively, as

hv

where B is the receiver bandwidth. In the heterodyne scheme, two methods are available to demodulate the IF signal coherently. In the first method, SocosA+fi(t)+x(t), in themarkstate we use a reference signal synchronized with the IF carrier. u(t)/K= in the space state. This can be generated by a phase-tracking circuit. However, when the I F signal has a large FM noise, it is (11) difficult for the phase-tracking circuit to generate a reference signal synchronized precisely with the IF carrier. Ifwe assume that the discrimination level is D,the BER , " in the mark state is given as Therefore, a phase error A+n(t) is induced between the IF P carrier and the reference signal. In this case, the demodulaP," = Prob(u(t)/K < D ) tor output u ( t ) is expressed as [7]
u ( t ) =KS(t)co~(B(t)+A+~(t))+Kx(t) (6)

where K denotes the transfer impedance from the input current to the demodulator output. The phase errorAGn(t) can beconsidered as a zero-mean Gaussian noise with a variance:a
(A4%(t)) = 0 ,

(7)

(8)
The variance u : is determined by the bandwidths of the phase fluctuation +Jt) and the phase-tracking circuit. In the second method, which can be applied only to the PSK scheme, the signal is demodulated by the differential phase detection. This method uses the IF signal itself as the reference, but with a delay equal to the signal bit cycle T. Therefore, the phase error is expressed as

where we define and use

f(x)=

x /1 x

e-x2dx

00

b=so
y =7

(normalized discrimination level) (14)

si?

2a

( S / N ratio).

(15)

A G J d =+&>-GA

-0 .

(9)

This canalso be considered as a zero-mean Gaussian noise. Its variance is determined from the bandwidth of the phase fluctuation & ( t ) and the bit cycle T. On the other hand, in the homodynescheme, the LO laser output is used as the reference signal. The received
i

On the other hand, the BER P," inthe space state is expressed as

1026

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. LT-2, NO.

6 , DECEMBER 1984

10
Y

20
(SN ratio) [dB1

30

Y (SN ratio) [dB1 Fig. 2. Theoretical BER curves for a PSK scheme.

Fig. 1. Theoretical BER curves for a ASK scheme.

final BER P, is given as When the mark and the space states have equal probability, the BER P, is given as

P,= -(PT+ P,"). 2

(17)

Fig. 2 shows the BER curves calculated from (19) and (20).

Fig. 1 shows the BER calculated from (17) as functions of y with u+ as a parameter. In the calculation we assumed that b = 0.5, which is optimum for u+ = 0 (noise-free case) [3]. When a+ # 0, the optimum value of b is slightly less than 0.5. However, the difference is negligible. C: Heterodyne Coherent PSK andHomodyne PSK Schemes The BER of a two-level PSK scheme with phases 0 and 7~ is calculated. We assume that the mark and the space states have equal probability. The demodulator outputs are given as
u ( t)/K =

D. Heterodyne Coherent FSK Scheme

In a two-level FSK scheme with a frequency deviation f,, the IF signal (equivalent input current) is expressed as

We assume that the IF signal is filtered by two bandpass filters in parallel, and that the filter center frequencies are (18) f I F + f , and fIF. The filter outputs are then demodulated by using reference signals having frequencies of f I F f , Theoptimum discrimination level D is zero, because and flF, respectively, generated by two independent signal levels in the mark and the space states have opposite phase-tracking circuits. Finally, the demodulator outputs signs. Therefore, the bit error in the mark state occurs are subtracted from each other. The output thus obtained when u(t) < 0 is given as

S,cosA@,(t)+x(t), - S,cos A@n( t ) + x ( t),

in themark state in the space state.

u ( t)/K =

SocosA~n((t)+x'(t), - S,cos A@, ( t ) + x ' ( t ),

in the mark state in the space state (22)

Since BERs in the mark and the space states are equal, the

where x'(t) is a zero-mean Gaussian noise having a variance of 2u2. From (18) and (22), we find that the signal level in the FSK scheme is equal to that in the PSK scheme, whereas the noise power is twice that in the PSK scheme. Therefore, the BER curvesfor in the coherent FSK scheme are obtained by shifting those in Fig. 2 by 3 dB to the right.

KIKUCHI et

d.: DEGRADATION OF BIT-ERROR RATE

1027

E. Heterodyne Differentially CoherentPSK (DPSK) Scheme

In the DPSKscheme, the IF signal itself delayed by a bit cycle T is used as the reference. With respect to the phase of the reference bit, we may express the IF signal i(t) as i ( t ) = { S(t)cosA@,,(t)+x(t)} (2Tf1,t .cos .sin(2rfIFt where x ( t ) = x r ( t ) c o s + n ( t - T ) + y ( t ) s i n + n ( t - T ) (24)
Q

a ,

+ +,, ( t - T ) )
y(t)} (23)

a , u U
L 0 L

- { S(t)sinA+n(t)+

+ +*(t- T ) )

* . . . +

k lo-?

y(t)=-x(t)sin+,,(t-T)+y(t)cos+,,(t-T)
(25)
0
Y (SN

10

20

and A+,(t) is defined by (9).

On the other hand, the delayed version of the IF signal,
i(t - T ) , which is used as the reference, can be written as

r a t i o ) [dB1 Fig. 3. Theoretical BER curves for a DPSK scheme.

where

i(t-T)= {S(t-T)+x(t-T)} -cos {2.rrf1,t + +,,(t - T ) }

- y(t - T)sin(2nfrFt

,
(26)

(32) (33)

+ +,,(t - T ) }

where we assume that f I F T= n (integer).Therefore,the u ( t ) algebraic is given, small after a demodulator output manipulation, as 1 u ( t ) / K = -( R; - R;) 8 (27) where R;= [ S ( t ) c o s A + , , ( t ) + x ( t ) + S ( t - T ) + d ( t - T ) ] 2

(34)
a : = u2 = 2u2.

(35)

By performing the integral in (30) and averaging over we obtain the final expression for P, as

A@,,,

+ [ S ( t ) sin A+,,( t )+ y( t ) + y( t - T ) ] + [S(t)sinA$,,(t)+ y(t)-y(t

--T)I2.

P,=-J4 0 3 e - ~ z J o , ~ o [ 2 ~ ~ ~ e - ~ * - ~

(28)

J;;

-m
*

R;= [ S ( ~ > C O S A + , , ( ~ ) + X ( ~ ) - S ( ~ - T ) - X ~ ( ~ - T ) ] ~
(29) where

lm(I0

[2@,$]e-52-pd ( de dt

(36)

The bit error occurs (1) when S(t) = S(t - T ) and u(t) < 0, and (2) when S ( r ) = - S ( t - T ) and u ( t ) > 0. If the mark and space states have equal probability, the bit-error rates in these cases are equal. Therefore, P, is expressed as

( i = 1,2) (31)

Fig. 4 shows the P, versus v+ characteristics calculated from (39).

1028

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. LT-2, NO,

6,

DECEMBER

1984

Phase d e t e c t o r

Reference

signal

i
2 0 L o -l
~

KO ",(s)= s v

Fig. 5. Block diagram of the electronic PLL circuit.

0.25

0.50
oD

0.75 [rad]

1.0

electric field is expressed in terms of the above variance as (44)

Fig. 4. Theoretical BER curve in the limit of infinite S / N ratio for PSK and FSK schemes as a function of u+.

On the other hand, in the ASK scheme, the of the BER ishalf of (39). This isbecause amplitude in the space state is zero regardless hence the BER in the space state isalways y -+a.

lower limit the signal of u+, and zero when

Equations (43) and (44) yield

R ( T )= exp(- nSfp).

(45)

111. RELATION BETWEEN THE VARIANCE OF THE PHASE ERROR AND THE FM NOISE OF LASERS

The field spectrum S( f ) is obtained as the Fourier transform of (45). Thus the spectrum S( f ) has a Lorentzian line shape with a full width at the half maximum (FWHM) equal to Sf,

BER's in various coherent optical In Section 11, the communication schemes have been calculated in terms of a+. I n t h s section we discuss how u+ is related to the FM noise of lasers.

where f is measured from the center oscillation frequency of the laser.

B. Analysis of Phase-Tracking Circuit

A. FM Noise and the Spectral Width o f the Laser

The phase fluctuation anof laser output is related to the instantaneous angular frequency fluctuation w I as [8]

The power spectrum of oIhas been measuredfor GaAlAs lasers by Yamamoto et al. [9], Daino et al. [lo], and Kikuchi and Okoshi [ll]. The power spectrum of a, in the frequency range 1 MHz 5 f 5 1 GHz, which determines dominantlythe line shape of semiconductor lasers, has been found to bealmost frequency independent. Therefore, in the following we assume that the frequency fluctuation w I exhlbits a whte Gaussian noise. The autocorrelation function of w1 then becomes (W/(t)W/(t - TI)
=

2nSf,W

(41)

where S f i is a constant and S(T) denotes Dirac's delta function. Thepower spectrum of a/, { l Q , ( f > l 2 ) , is given as

In the coherent detection, we need a referencesignal whose phase is synchronized with the IF carrier. The reference signal is generated in a phase-tracking circuit such as a phase-locked loop (PLL). The phase error A+,, between the IF carrier and the reference signal depends on the FM noise of the IF signal and the bandwidth of the PLL circuit. We derive a simple expression for the phaseerror variance u+, based on a linearlized model of a PLL circuit. Fig. 5 shows the linearlized model of a PLL circuit [12]. The Laplace transforms of the phase of input IF signal, that of reference signal, and the phase error between them, are denoted by O,(s), O r ( $ )and O,(s), respectively. The phase detector detects the phase difference between the IF and reference signals. The output is then led to a loop filter whose transfer function is F ( s ) . The output voltage of the loop filter is given as

(152, ( f ) I ')

= Sf,/277

(two-sided spectrum).

(42)

The variance of the phase deviation induced in a time interval T is given, from (40) and (411, as

{a@:(
On theotherhand,

T))

= 2778fiT.

(43)

the autocorrelation function of the

where K , denotes the conversion factor of the phase detector. The phase error between the IF carrier and the reference signai generated in a PLE circuit originates from the shot noise as well as the FM noise of the IF carrier. However, the phase error due to the shot noise is much smaller than that due to the FM noise, when the S / N ratio of the IF

KIKUCHI et

al. : DEGRADATIONOF BIT-ERROR RATE

1029

signal is sufficiently high ( > 10 dB), and the locking bandwidth of the PLL is smaller than the signal bandwidth. These conditions are satisfied in an actual system. Therefore, in the following, we neglect the shot noise effect for the simplification of the analysis. The voltage V, controls the frequency of a voltage-controlled oscillator (VCO). The VCO then generates the reference signal. The phase of the reference signal is given as

where 1 / T is the data transmission rate; in such a case, we find from (49) that the PSK waveform is not deteriorated by PLL.
C. Phase- Error Variance in Heterodyne Coherent Scheme

The variance of thephase error inthe steady state is expressed as

0 ; =

+M

IO,( j 2 n f ) l 2 d f .

(571

Or(s)

= -V,(S)
S

K O

-02

where K O is the conversion factor of the VCO. Thus, we have the relation between 0,and Oiin the closed loop as (49) where
C = K,K,.

Therefore, from (54) and (56), we obtain the variance of the phase error A+,, between the input IF carrier and the reference signal

Equation (58) is the relation to be obtained for the heterodynz coherent schemes. (50) D. Phase -Error Variance in DPSK Scheme In the DPSK scheme, the phase error A c $ ~ ( ~is ) given by (9). From (9) and (51), we obtain

Inthe following, we assume that F ( s ) = 1 (first-order loop). The instantaneous angular frequency of the IF signal is given as
WIF

= W T - W,

(51)

A+,,(t) = / r + T ~ I F ( dt. t)
f

(59)

where wT and w L are instantaneous angular frequencies of the transmitter laser and LO laser, respectively. The autocorrelation function of wIF is obtained from (41) and (51) as
(WIJ?(t)WIF(t

Equations (51) and (59) lead to

If we define the signal bandwidth f, as

= 2gr(6fT+ s f L ) s ( 7 )

(52) the variance of the phase error is expressed as

where S f T and Sf, denote spectral widths of the transmitter laser and LO laser, respectively. From (46) and (52), we find that the IF signal has a Lorentzian line shape with an FWHM of SfT + S f L . The power spectrum of the frequency fluctuation of the input IF .signal is given as (two-sided spectrum) (53)

This is the relation to be obtained for the DPSK scheme. E. Phase - Error Variance in Optical PLL Homodyne Scheme An optical PLL system or an injection-locking technique can beused for the phase-tracking in the homodyne scheme. Fig. 6 shows the block diagram of an optical PLL system 1131,[14]. In t h s system, a photodetector functions as the phase detector and the LO laser as the VCO in the electronic PLL in Fig. 5. When a semiconductor laser is used as the LO laser, the frequency can be controlled by changing its bias current. After calculations similar to those led to (49), we obtain the phase error O,(s) between the transmitted carrier and the LO laser as
O,(s)
=

Therefore, the power spectrum 10e(j27rf)l ofthe phase difference between the input IF signal and the reference signal is expressed, from (49) and (53), as

where

f,. = - (locking bandwidth of phase-locked loop).

(55) ThePLL circuit trackes the phase fluctuation within the 1ocking.bandwidth f , . On the other hand, we assume that 1/T
f c

C 277

sOi(s)+i2L(s)
s+c

(56)

where C is a quantity proportional to the loop gain of the optical PLL. The FM noise term of the LO laser Q,(s) is

1030

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. LT-2, NO.

6 , DECEMBER 1984

Varlable delay

+%
Fig. 6 . Block diagram of the optical PLL circuit.
Transmitted signal

Attenuator

Local oscll lator power

I

Fig. 8.

Simulation model of a PSK heterodyne optical communication system.

G. Comparison of Various Schemes

Note ,the resemblance among (58), (62), (64), and (68); the only difference is found in the definitions of the denominators. IV. EXPERIMENT

Fig. 7. Block diagram of the injection-locking scheme

now added to the numerator, because the FM noise generated by the LO laser can not be neglected. By using (56) : as and (63), we obtain the phase error variance a

where

F. Phase - Error Variance in Injection - Locking Homodyne Scheme

Fig. 7 shows the block diagram of the second homodyne scheme, the injection-locking technique [15]. The LO laser is phase-locked to the transmitted signal by injection. The transmitted signal light is homodyne-detected with the LO laser output. From the Van der Pol equation, we may readily derive a differential equation governing the phase error in the injection-locked state [15]

where 8, is the phase error, Bi the phase of the transmitted signal, (b,! the amplitude of the transmitted signal, (b,l the amplitude of the LO laser, K the loss of the resonator of the LO laser, and w L the instantaneous angular frequency of the LO laser. The Laplace transform of (66) gives

To verify the theoretical prediction givenby (19) and (20), the BER of a PSK systemhas been measured byusing a simulation model [6]. Fig. 8 shows the simulation model of the PSK heterodyne system used in the experiment. A single-mode GaAlAs laser oscillating at 840 nm is used as a light source. An optical isolator is inserted in front of the laser to remove the spectral instability induced by the returning beam. A part of the laser light is used as the local oscillator power. The other part, the signal light, is frequency shifted by 40 MHz by using an acoustooptic modulator. The phase of the light is also modulated between 0 and a by switching the phase of the modulatordriving voltage by a PCM signal of 1 Mbit/s. By changing the opticalpath difference, we can add acontrolled amount of FM noise to the IF signal, because the GaAlAs laser has a spectral spread. The receiver consists of a Si p-i-n-photodiode and front-end circuit with a tuned amplifier [16]. The receiver output is led to a PSK demodulator shown in Fig. 9. We use an open-loop phase-tracking circuit instead of the PLL described in Section I11 [17]. The bandwidth of the phase-tracking circuit islimitedto 30 kHz in this experiment. Therefore, when the optical path difference is increased, the phase-tracking error increases. We can thus obtain a desired value of a+. The value of u+ can be estimated from the measured voltage fluctuation of the demodulator output. The demodulator output in the mark state is given as

u ( t ) = KS,COS AGn

(70)

when the signal power is sufficient. Assuming that A+,, has aGaussian probability distribution, the average u,, and the variance u,, of u ( t ) can be expressed as Thus we obtain an expression for
; a =

02 from (64) as
(68)
Uvar =

Sf, + SfL
2f.

- ( K S J 2 ( 1- exP

(69) well In the injection-locking scheme as scheme, we assume that l / T > > f,. as the PLL

From (71) and (72), we have (73)

KIKUCHI

et a[.: DEGRADATION OF BIT-ERROR RATE

1031
TABLE I

I F siqnal

~~

Phose tracking circuit

REQUIREMENTS FOR VARIOUS PHASE-TRACKING SCHEMES WHEN THE SPECTRAL WIDTH OF THE LASER IS 10 MHz.
~~

Heterodyne PSK Heterodyne DPSK Optical PLL Homodyne

Bandwidth of PLL B >25OMHz Signal bit rate l/T>3Gbit/s Bandwidth of PLL Bo>250MHz

Fig. 9.

Block diagram of a demodulator of PSK signal.

Injection-Locking Homodyne Signal power Ps>-37dEm

The dotted curve in Fig. 10 shows a theoretical BER in an ASK envelope-detection scheme. The measured BER (dots) are found to be about 2 dB higher in y-values. It is foundthat the receiversensitivity of the coherent PSK scheme in the FM-noise-free state (u+ = 0) is better by 5 dB ascompared with the ASK envelope-detection scheme. (The difference should theoretically be 6 dB [3], [4].)
V.
0
Y (SN

DISCUSSION

10

20

30

A . Requirements in Various Phase -Tracking Schemes

We consider the feasibility of the PSKschemeusing semiconductor lasers.Figs. 2 and 3 show that the phase error u+ must be smaller than 0.2 rad to suppress the receiver-sensitivity degradation at BER = below 3 dB. On the other hand, we have the relation

r a t i o a t I F stage) (dB1

Fig. 10. Measured and theoretical BER of PSK and ASK systems: 0.34 (measured) 0.30 (measured) x - u = Orad, A - u+ = , u - u+= 9 0.28 (fitted) 0.35'(fitted) 0.47 (measured) 0-u+= 0.47 (fitted)

On the other hand, the variance u, can be determined from the measured average power P , dissipated in a ap: can propriate load resistance by u ( t ) . Thus the variance u be determined by using (73). The BER has been measured as a function of the S / N ratio y for four values of u+. Note that in the heterodyne scheme, the S / N ratio y defined by (10) represents the S / N ratio at the IF stage. Crosses, triangles, squares, and circles in Fig. 10 show the measured values of the BER. For comparison, the measured BERs in an ASK envelope-detection scheme are also plotted by dots. Small crossed along solid curve in Fig. 10 show the BERs measured in the FM-noise-free state, i.e., u+ = 0. The dash-dotted curve is the theoretical curve for a+ = 0. Approximately 3-dB difference is found in y between measured and theoretical characteristics. This difference is .probably mainly due to the distortion in the PCM waveform. Solid curves and @ are the theoretical curves calculated by using (19) and (20), where u+ is assumed so as to give best fit with experiment. Furthermore, all these curves are shifted to the right by 3 dB to be consistent with result for a+ = 0 (curve The values of u+ estimated from the voltage-fluctuation measurement (see (73)) and those from the above fitting, both shown in the caption of Fig. 10, showgoodagreement,suggesting the validity of the theory.

where fx stands for 2f , in ( 5 8 ) and (64), f , in (62), and 2 f , in (68), corresponding to various types of the phase-tracking scheme. If we assume that 6f T = SfL( = S f ) and u+ 0.2 rad, we have

Sf < 0.O2fx.

(75)

The typical spectral width S f of a GaAlAs laser is 10 MHz when the output power is 10 mW [18].Therefore, f x must be greater than 500 MHz to keep u+ -= 0.2 rad. Table I shows the requirements for various phase-tracking schemes for realizing f , > 500 MHz. The details will be described in the following three subsections.
B. Feasibility of Heterodyne P S K or DPSK Schemes

0, 0,

a).

In the heterodyne PSK scheme, the requirement f , = 2f, > 500 MHz implies Be > 250 MHz where B, denotes the PLLbandwidth. In the heterodyne DPSK scheme, the requirement f , = f , > 500 MHz leads to (1/T) > 3 Gbit/s. In such systems the intermediate frequency must practically be comparable to or greater than 10 GHz to satisfy Be > 500 MHz or ( 1 / T ) > 3 Gbit/s. At present this requirement is difficult to achieve because of the response limitation of the photodetector to be used as the mixer. Therefore, some spectral purification techniques [19] must be completed before the realization of such systems.

1032

JOURNAL OF LIGHTWAVE

TECHNOLOGY, VOL. LT-2,NO. 6,

DECEMBER

1984

is 10 mW and K = 5 X lo1 s-. This requirement can not practically be satisfied because the received signal level of In the optical PLL-type homodyne scheme, the require- the PSK homodyne scheme for BER = is of the order ment is again expressed as the PLL bandwidth Bo > 250 of magnitude of -60 dBmwhen the signal bit rate is 1 MHz. This seems to be feasible because the bandwidth of Gbit/s. the frequency modulation of a semiconductor laser by varying bias current exceed 1 GHz. However, the frequency VI. CONCLUSION deviation depends strongly on the modulation frequency. Therefore, a phase compensation is necessary in the PLL The B E R s of various coherent optical communication circuit to remove the circuit instability. schemes have been calculated. The FM noise of the transMoreover, we should note that the controlling bias curmitter and LO lasers deteriorates the BER performance rent inevitably induces an additional AMnoise composeriously as shown in Figs. 1-4. In particular, in the PSK nent. In the following, we estimate the magnitude of this scheme, the phase error between the received carrier and AM noise. the reference must be kept within 0.2 rad to suppress the The parameter C of the PLL defined in (63) is given, receiver sensitivity degradation at BER = loe9 below 3 dB. corresponding to (50), as The theory presented has been verified by the BER meaC = 2.rr{K, (75) surement in the PSK heterodyne scheme. On the basis of the analysis and the experiment, the where 5 denotes the ratio of the frequency increase of the feasibility of various PSKschemesusing semiconductor laser to the bias-current increase, and K c denotes the lasers has been discussed. It has been demonstrated that bias-current increase induced due to a phase error of a the technical requirements are rather severe as far as the ( I ) laser spectral spread remains as wide as 10 MHz. unity radian. The rms of the bias-current fluctuation vr-r is given, from (64), (65) and ( 7 9 , as
C. Feasibility of Optical P L L Homodyne Scheme
REFERENCES

This bias-current fluctuation induces a photodetector-curwhich is given as rent fluctuation

= 2(

m,

;)( 2 ) m

(77)

where E is the slope of the laser power-versus-current characteristics. On the other hand, the photocurrent fluctuates inevitably due to the shot-noise effect, whose rms is given as

From (77) and (78), we obtain the ratio of the two photocurrent fluctuations as

If we assume that 5 =lo* Hz/A [20], c = 0.2 W/A (for GaAlAs CSP lasers), 17 = 0.7, P , =lo- W, and Sf = lo7 Hz, R is calculated to be 15. Therefore, the residual AM noise of the LO laser predominates the shot noise. The former is often hazardous inthe system performance. Some AM-noise reduction technique such as the balanced-mixer-type detection scheme [21] will have to be employed to remove its effect.
D. Feasibility of Injection - Locking Homodyne Scheme In this case, the requirement is given as f , = 2f,> 500 MHz, and this leads to P,> - 37 dBm when the LO power

coherent optical fiber communications: Recent progress, IEEE Truns. Microwave Theory Tech., vol. MTT-30. pp. 1138-1149,1982, D. J. Malyon, T. G. Hodgkinson, D. W. Smith, R. C. Booth, and B. E. Daymond-Fhn, PSKhomodyne receiver sensitivity measurement at 1.5 pm, Electron. Lett., vol. 19, pp. 144-146, 1983. T. Okoshi, K. Emura, K. Kikuchi, and R.Th. Kersten, Computarate of various heterodyne and coherent-type tion of bit-error Commun., vol. 2, pp. optical communication schemes, J . Opt. 89-96, 1981. Y. Yamamoto, Receiver performance evaluation of various digital optical modulation-demodulation systems in the 0.5-10 pm wavelength region, IEEE J . Quantum Electron., vol. QE-16, pp. 1251-1259, 1980. F. Fame and D. Le Guen. Effect of semiconductor ohase noise on BER performance in an optical DPSK heterodyne-typk experiment, Electron. Lett., vol. 18, pp. 964-965, 1982. K. Kikuchi, T. Okoshi. M. Nagamatsu, and N. Henmi, Bit-error rate of PSK heterodyne optical communication system and its degradation due to spectral spread of transmitter and local oscillator. Electron. Lett.. vol. 19. DU. 417-418. 1983. S. Stein and J. J. Jones, Mbh.&n Communicution Principles. New York: McGraw-Hill, 1965. M. Sargent 111, M. 0. Scully, and W. E. Lamb, Jr., LuserPhysics. London: Addison-Wesley, 1974. Y. Yamamoto. S. Saito, and T. Mukai. AM and FMauantum noise in semiconductor lasers-Part 11: Comparison of theoretical andexperimental results for AlGaAs lasers, IEEE J . Quuntum Electron, vol. QE-19, pp. 47-58, 1983. B. Daino, P. Spano, M. Tamburrini. and S. Piazzolla, Phase noise and spectral line shape in semiconductor laser, IEEE J . Quantum Electron., vol. QE-19, pp. 266-270, 1983. K. Kikuchi and T. Okoshi, Measurement of spectra of and correlation between FM and AM noises in GaAlAs lasers, Electron. Lett., vol. 19, pp. 812-813,1983. F. M. Gardner, Phuse Lock Techniques. New York: Wiley,1979. W. R. Leeb, H. K. Phillip. A. L. Scholtz. and E. Bonek, Frequency synchronizationand phase locking of CO, lasers, Appl. Phys. Lett., vol. 41, pp. 592-596, 1982. R. C. Steele, Optical phase-locked loop using semiconductor laser diodes, Electron. Lett., vol. 19, pp. 69-71, 1983. S. Kobayashi and T. Kimura, Optical phase modulation in an injection locked AlGaAs semiconductor lasers, Electron. Lett., vol. 18, pp. 210-211, 1982. K. IClkuchi, T. Okoshi. and K. Emura, Achievement of nearly shot-noise-limited operation in a heterodyne-type PCM-ASK optical communication system. presented at the 8th Europ. Conf. Opt. Commun., Cannes, France, 1982.

T. Okoshi, Heterodyneand

KIKUCHI

et a/. : DEGRADATION OF BIT-ERROR RATE

1033 Bell Laboratories, Inc., Murray Hill, NJ, where he was engaged in research on electron guns. In 1972 he joined the Technical University of Munich on a temporary basis as a Guest Professor. In January 1977 he became a Professor at the University of Tokyo. At present, his main fields of interest are optical fibers, optical-fiber communications, microwave planar (two-dimensional) circuits, and optical memories. He has written twelve books including three in English: Three-Dimensional Imaging Techniques (New York: Academic Press, 1976), Optical Fibers (New York: Academic Press, 1982), and PIanarCircuits-forMicrowavesand Lightwaves (Heidelberg: Springer, to be published in 1984). He holds 13 Japanese patents and 5 U.S. patents. Dr. Okoshi is also a Guest Research Fellow of Radio Research Laboratory of JapaneseGovernment, and the Secretary and Commission-D Chairman of JapaneseNational Committee for URSI. He is presently serving as the Chairman of Tokyo Chapter of the IEEE MTT Society since January 1983. He has been awarded twelve prizes from three Japanese academic institutions.

1171 K. Sam Shanmugam, Digital and AnalogCommunication Systems, New York: Wiley, 1979. [18] M.W. Fleming and A. Mooradian, Fundamental line broadening of sinele-mode (GaA1)As diode lasers. Aool. .1 Phvs. Letf.. vol. 38. pp. 511-513, 1981. [19] K. Kikuch and T. Okoshi, Simple formula giving spectrum-narrowing ratio of semiconductor-laser output obtained by optical feedback, Electron. Lett., vol. 18, pp. 10-12, 1982. 1201 S. Kobayashi, Y. Yamamoto, M. Ito, and T. Kimura, Direct frequency modulation in AlGaAs semiconductor lasers, IEEE J . Quantum Electron., vol. QE-18, p ~ 582-595, ; 1982. [21] H. P. Yuen and U. W. S. Chan, Noise in homodyne and heterodyne detection, Opt. Lett., vol. 8, pp. 177-179, 1983.
,

Kazuro Kikuchi was born in Miyagi Prefecture, Japan,on March 6, 1952. He received the B.S. degree in electrical engineering from University of Tokyo, Tokyo, Japan, in 1974, and the M.S. and the Ph.D. degrees both in electronic engineering from the same university, in 1976 and 1979, respectively. In 1976 he joined the Department of Electronic ~ ~ ~ i ~ of Tokyo, ~ ~ is r at University and now an Associate Professor, He is currently involved in research for coherent optical communications and noise of semiconductor lasers. He is a. member of theInstitute of Electronics andCommunication Engineers of Japan, the Japan Society of Applied Physics, and the Physical Society of Japan.

Masato Nagamatsu was born in Oita,Japan, on July29, 1959. He received the,B.S. and M.S. degrees in electronic engineering both from University of Tokyo, i ~ Japan, ~ in 1982 and 1984, respectively. He is now with Toshiba Research and Development Center in Kawasaki, Japan He is a member of the Institute of Electronics and Communication Engineers Of Japan.

Takanori Okoshi (S56-M60-SM81-F82) was born in Tokyo, Japan, on September l,6, 1932. He received the B.S., M.S., and Ph.D. degrees, all from the University of Tokyo, Tokyo, Japan, in . 1955, 1957, and 1960, respectively, all in electrical engineering. In 1960 he was appointed an Instructor, and in 1961, became an Associate Professor in the Department of Electronic Engineering, University of Tokyo. From 1963 through 1964, on leave of absence from the University of Tokyo, he joined

Naoya Henmi (s84) was born in Tokyo, Japan, on February 27, 1961. He received the B.S. degree in electronic engineering in 1983 from University of Tokyo, Japan, and is now studying toward the M.S. degree at the same University. He is a member of the Institute of Electronics and Communication Engineers of Japan.