You are on page 1of 9

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 9, NO.

IO, OCTOBER 1991 I361

PLL Propagation Delay-Time Influence on Linewidth


Requirements of Optical PSK Homodyne Detection
Seiji Norimatsu and Katsushi Iwashita

Abstract-PLL propagation delay time influence on optical linewidth requirement. In near future, the solitary LD’s
homodyne detection based on a decision-driven phase-locked are expected to be used in the real systems for their reli-
loop is unavoidable in the real systems. The loop delay-time
influence is investigated both theoretically and experimentally. ability and their compactness. At that time, the relations
Applying the Pad6 approximation, which is often used in the between required laser linewidth and the loop delay time
control system, to the calculation of the phase-error variance, must be studied in detail.
high accuracy analytic expression of it is obtained. The line- In this paper, the Pade approximation [9], which is
width requirement with the nonnegligible loop delay time for often used in the control system, is adopted in calculation
PSK homodyne detection is obtained as 6 u = 2.04 x 10-‘/~
where 6 u (hertz) is beat linewidth and 7 (seconds) is the loop of the phase-error variance with the nonzero loop delay
delay time. The linewidth requirement with small delay time time. With this method, the analytic expression of the
approaches 6 u = 6.2 x R,, where Rh (bits-per-second) is phase-error variance can be obtained. In Section 11, the
the system bit rate. Results are confirmed by a lO-Gb/s optical analysis methods of PLL based on a decision-driven loop
PSK homodyne detection experiment by using external cavity
is briefly reviewed. In Section 111, the accuracy of the Pad6
LD’s. Receiver sensitivity degradations due to loop delay time
and beat linewidth are in good agreement with theoretical re- approximation is studied theoretically and the relations
sults. between laser linewidth and the loop delay time are ob-
tained. The experimental results on the influence of the
loop delay time of PLL are reported in Section IV.
I. INTRODUCTION
11. BRIEFREVIEWOF PLL BASEDON A
0 PTICAL phase-shift keying (PSK) homodyne detec-
tion offers the best sensitivity of any binary signaling
technique and requires only the same electrical bandwidth
DECISION-DRIVEN LOOP
The construction of the PSK homodyne receiver based
as the bit rate. While it is suitable for multigigabit coher- on decision-driven PLL is reviewed roughly in this sec-
ent transmission systems, it requires extremely narrow tion [ 11. The block diagram of the decision-driven PLL
linewidth lasers. Laser linewidth requirements depend on type receiver is cited in Fig. 1. Here, ~ $ ~ is
( t the
) phase
the configuration of the phase-locked loop (PLL). A de- of the received signal and is given by
cision-driven PLL can tolerate a wider linewidth require-
ment, e.g., 6.2 X IOp4 times the bit rate [ I ] , than a bal- dS(f) = + d ( t ) + +NT(r) (1)
anced PLL which is 5.9 x l o p 6 times the bit rate [ 2 ] . where +NT(t)is the phase noise of the received signal and
Therefore, linewidth requirement can be tolerated by in- ~ $ ~ ( depends
t) on the transmitted data d

i
creasing the bit rate. However, the linewidth requirement 0, ifd = +1
is also restricted by PLL propagation delay time. Some = (2)
papers have considered loop propagation delay theoreti- T, i f d = -1.
cally [ l ] , [3]. However, only the region where the loop +LO(r) is the phase of the local oscillator’s signal and is
propagation time is negligible is clarified in [ 11 and only
given by
the conditions for two types of linewidth were obtained in
[3]. In these works, there were no general conditions for +LO(f) = +c@)+ +NLO(f) (3)
linewidth requirement with nonzero loop propagation where &LO(r) is the phase noise of the local oscillator’s
time. signal and & ( t ) is the controlled phase determined by the
Recently, several results on multigigabit homodyne de- VCO input controlled voltage and is given by
tection experiments using LD’s have been reported [4]-
[8]. Although the required beat linewidth at 10 Gb/s with
a decision-driven PLL is 6.2 MHz, these LD’s were op-
C$c(t)= KVCO j‘ -m
V c ( 0 dt’ (4)
erated by narrowing spectral purity with the external cav-
ities, because the loop propagation time restricts the where Kvco(rad/(sV)) is the VCO gain.
The output voltages of ac-coupled optical detectors are
Manuscript received January 8. 1991; revised May 13. 1991.
The authors are with NTT Transmission Systems Laboratories, 1-2356 V l ( t )= ~ R ~ J- (ks)(l
I - kLo) JPsP,od
Take Yokosuka-shi, Kanagawa 238-03, Japan.
IEEE Log Number 9102076. . cos [+rl(t> - 4c(t)l + n,(t) (5)
0733-8724/91/1000-1367$01.00 0 1991 IEEE
1368 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 9. NO. IO, OCTOBER 1991

LOCAL OSCILLATOR
SIGNAL
I'pro exp UMW

m(t)

mS (Wd(I)+ (1) @,,Lo(f)


+Ldt)='c(th Fig. 2. A linearized model of Fig. 1

Fig. I . Block diagram of optical decision-driven PLL receiver.


As for the power spectral density (PSD), the double-
sided PSD is adopted. Therefore, the PSD's of the phase
VAt) = 2Rr& d sin [4,(t) - 4&)] + n2(t) noise which is caused from 4,(t) and that of the shot noise
are given by
where R(A/ W )(= v e / h v ) is the detector's responsivity,
7 is the quantum efficiency of the photo detector, e ( A s)
is the electron charge, h (J s) is the Planck constant, v
(hertz) is frequency of light, r ( Q ) is the detectors' load
resistance, Ps (W) is the received signal power, PLO (W)
is the local oscillator power which is used as the voltage where 6 v (hertz) is the beat linewidth (FWHM) of the
controlled oscillator (VCO), &(t) represents the total transmitter's light and the local oscillator's light. To eval-
phase noise of the received light and the local oscillator's uate the phase-error variance, it is necessary to integrate
light ( E 4 N T ( t ) - &.,LO(t)), n l ( r ) , and n2(t) represent the the PSD by frequency f from -a to m.
shot-noise processes, k, is the received power splitting The phase-noise and the shot-noise processes are
ratio to Q-arm, and kLO is the local oscillator's power independent from each other, therefore the PLL per-
splitting ratio to Q-arm. Two lines of electrical circuits formance is characterized by the phase-error variance
following the two optical detectors are called I(1n phase)- defined by
arm and Q(Quadrature phase)-arm, as in Fig. 1. The de-
o2 E E { & ( t ) } = E{[+,,(t)- 4c(t)I2}
cision circuit output signal dT is considered to be the same
as the data d delayed by one bit duration T. Therefore, E O$N + o& (rad2) (14)
the Tdelay line is inserted to Q-arm. Then, the multiplier
output signal is where c$,(t) represents the phase error (E 4,(t) - 4c(t)),
E{ } represents the infinite time average, o'pNrepresents
m(t) = dT V2(r - T ) . (7) the phase-error variance due to the phase noise and oiN
Substituting (6) into (7), we obtain represents the phase-error variance due to the shot noise.
The phase noise and the shot noise are considered inde-
m(t) = KPDsin [&(t - T ) - &,(t - T ) ] + n2(t - T ) pendently in analyzing the PLL performance. From Fig.
(8) 2 , the following relation can be obtained:
where KpD is the phase detector gain and defined as
KpD = 2Rr& (V/rad). (9)
Here, n2(t - T ) dT is replaced by n2(t - T ) because dT is
+ 1 or -1 with the same ratio, and n2(t - T ) offers the
same impact as n2(t - T ) dT does. Suppose that the VCO's where K (hertz) is the total loop gain and defined by K =
phase is in lock to the received signal phase, it can then KpDKVCO and T (seconds) is the loop delay time.
be assumed that First we set the phase noise to zero
14,1(1)- 4c(t>I << 1. (10) 4fAO = 0. (16)
In this case, (8) can be linearized by using sin x = x and Substituting (15) with the condition of (16) into (14), the
becomes phase-error variance due to the shot noise is given by
m(t) = KPD[4At - T ) - 4dt - 731 + n2(t - T ) .
(1 1)
oiN =
KPD
Sm

--OD
I H ( f ) I 2 df
K'pD
(rad2) (17)

The linearized PLL model is shown in Fig. 2. The ex- where H ( f ) is the closed-loop transfer function and B,, is
planation of notations will be given below. the noise bandwidth. The closed-loop transfer function
NORIMATSU d nl.: PLL PROPAGATION DELAY-TIME INFLUENCE 1369

H( f ) is given by the following expression: mation [9], which is often used in the control system. The
K . F ( f ) . e-12sfr (1,l)th and (2,2)th Pad6 approximation of e - " . (In this
paper we call the 1st Pad6 approximation and the 2nd Pade
j2r.f (18) approximation, respectively.) are given by
H(f) = K . F ( f ) . e-/2sfr'
I + 1
j2r.f 1 - - U
e-" =
2
Next we set the shot noise to zero
~

(27)
1
1+-a
n2(t) = 0. (19) 2
Substituting (15) with (19) and (18) into (l4), the phase-
e-' =
12 - 6~ + U'
error variance due to the phase noise is given by 12 + 6u + a*
nm
respectively. If these approximations are applied to (25),
(26), the integrands become the rational expressions and
the integrals can be done easily and the analytic expres-
sions of the phase error variance can be obtained as in
Appendix.
= 27r . Sv . A, (rad2). (20)
111. THEORETICAL
CONSIDERATIONS
WITHTHE PADE
As for the loop filter, the first-order active filter is APPROXIMATION
adopted and its transfer function F( f ) is denoted by
In this section, the accuracy of the Pad6 approximation
is considered theoretically and the linewidth requirement
of PSK homodyne detection with this approximation is
obtained.
Substituting (2 1) into (1 8), (1 8) is given by We consider the accuracy of the Pad6 approximation
through the method in [3], which estimates the degrada-
tions of the phase-error variance with nonzero loop delay
time due to both the phase and shot noise from the phase-
and 1 - H(s) to be error variance due to both the phase and shot noise with
zero loop delay time, which are expressed as [ 11
S2
1 H(s) (23)
- =
s
2
+ G(sr2 + l)e-sr
where s = j27rf and G = K / r l . G, r , and 7 2 are real
numbers, therefore, there is the relation given by

H(s)* = H ( - s ) (24)
where
where * represents the complex conjugate. By using s, A,,
and B, are represented by x = Gr;. (31)
1 - H(s) The expressions of the phase and shot noise parts of the
phase-error variance with the 1st Pad6 approximation are
given by (see (A8) and (A9))

ds .
= -
1 xy(2 - y ) ( x - I ) + 4x + 4 (33)
(26)
&I
x + 1 X~(-
Y 2) + 4(1 - y ) U:NO
The analytic solutions for (25) and (26) have not yet where
been obtained, except for r = 0 [I], [3]. However, to
7
apply phase error variance to, for instance, PLL optimi- y = -. (34)
zation with nonzero loop delay time, these solutions are 72

needed. It is better to approximate e -.'' and calculate (25), Those with the 2nd Pad6 approximation are given by (see
(26) analytically. One method is using the Pad6 approxi- (A10) and (A1 1))

144 - 3&y2 72xy - x2y4 + 6x2y' + - 12x2y2 2


uiN2 =
144 - 144y - 12xy3 + 3&y2 - 72ry - x'y' + 6x2y3 - 12x2y2 (35)
I370 JOURNAL OF LIGHTWAVE TECHNOLOGY. VOL. 9, NO. IO. OCTOBER 1991

The phase-noise degradations ( u i N /uiNo


l and u ; N ~ /
4- 7
utNo) due to the normalized delay time w , , (U,?:
~ the loop
natural frequency) are shown in Fig. 3(a) and those of the
and
2nd P a d j A shot noise (a&, / U & and U&,/.&)) in Fig. 3(b). Those
exact obtained through the exact calculation are also plotted in
both figures. However, the exact curves and the 2nd PadC
approximation’s ones almost overlie. The degradation
values with the 2nd PadC approximation are good agree-
/ , , , , I
ment with the exact ones because the difference at w,,r =
0.1 0.2 0.3 0.4 0.5 0.6 0.6 (& X y = 0.6) is about 0 . 4 % . The 1st Pade approx-
NORMALIZED DELAY TIME WnT imation curves separate from the exact curves about more
(a) than w , = ~ 0.3 in these figures. In these figures the case
when the damping factor is 1 /& is shown, however, the
same calculations with the 2nd Pade approximation for
the damping factor 10 have also been done. The result is
that in both cases the 2nd Pade approximation is appli-
cable up to q d B 7 i 1.2 within 1 % error. Here w~~~ is
=w I exact /../ the 3-dB loop band width with zero delay time and it is
defined as

NORMALIZED DELAY TIME WI t


(b)
Therefore, it is confirmed that the phase-error variance
Fig. 3. Degradation of phase-error variance with nonzero loop delay time
from that with zero loop delay U ’ ( T > O ) / U ’ ( T = 0) obtained through the can be with high accuracy by the 2nd
1st PadC approximation, the 2nd Pad6 approximation, and exact calculation Pade approximation Up to ~ 3 r I d 1.2
~ within 1 % error.
versus normalized loop delay time U , ,7 . (a) Phase-noise part. (b) Shot-noise The allowable beat linewidth considering the loop de-
part.
lay time is obtained by the 2nd Pad6 approximation, which

These have relations with the loop natural frequency U,, can be applied up to w~~~ r = 1.2. The case that the damp-
and the damping factor { as follows [lo]: ing factor { is 1 /& is considered. Therefore, the expres-
sion of the phase-error variance is a sum of oiN2 and
U, ’ 7 = 67 = A y (37) u ; N ~ (see (41) and (42)). The standard deviation of phase
r
error U with respect to the loop natural frequency w , ~for
several loop delay times 7 is shown in Fig. 4. The mini-
mized standard deviation of phase error U with respect to
These expressions seem to be very complicated, however, the loop delay time 7 for several beat linewidths are shown
in case of the 2nd Pad6 approximation with the damping in Fig. 5 . The minimized standard deviation of phase-
factor is set to be 1/& (x = 2), the phase-error variance error increases gradually by increasing the loop delay time
due to the phase noise and that due to the shot noise are and the beat linewidth. The loop natural frequency w, de-
expressed as pendence on the loop delay time 7 which minimizes the
standard deviation of phase error is shown in Fig. 6 and
2 2+2y-y2
upNI = (39) closes to 0.34/7. The optimized loop natural frequency
w, decreases as the loop delay time increases. Fig. 6 shows
6+2y-y2 that the optimizations are operated within the area in
which the 2nd Pad6 approximation can be applied with a
uiNI =
6 - 12y 3y2 + small error, w, 7 < 0.6. In this calculation, the total power
+ 36y - 30y2 + 6y’ - y4
36 penalty is assumed to be 1 dB. The power splitting ratio
=
36 - 72y + 6y2 - y4 (41)
uPN2 UiNO to the @arm holds ks = kLO = 0.10875, which cause a
0.5-dB power penalty. The other 0.5 dB is caused by the
36 + 12y - 18y2 + 6y3 - y4 phase error variance.
= (42)
36 - 72y + 6y2 - y4
uSN2 USNO.
The normalized permitted maximum beat linewidth
NORIMATSU cf U/.: P L L PROPAGATION DELAY-TIME I N F L U E N C E 1371

k =k =0.10875
s LO

L . . ......1 . .......I . , ..d . , . I . . . . I . .....,.I


'105 106 io' 108 lo9 10'O

LOOP NATURAL FREQUENCY w, [rad/sec]

Fig. 4. Standard deviation of phase-error a versus loop natural frequency


U,, for several loop delay time T .
...
NORMALIZED DELAY TIME zn
-
Fig. 7 . Normalized beat linewidth 6v/R,,, which permits I-dB power pen
alty at bit error rate 10- "I, versus normalized loop delay time TR,,.

The power penalty under the nonnegligible loop delay


time T is investigated. The loop delay time T with fixing
the linewidth 6 v and the beat linewidth 6 v with fixing the
loop delay time T are changed in vicinity of 6 v / R b = 1.02
x I O p 6 or TR,, = 2 x 10' in Fig. 7. This result is shown
in Fig. 8.
In the above calculation, the power splitting ratio to the
Q-arm of the signal ks and the local oscillator k,, are
LOOP DELAY TIME 'T [Sec]
fixed. In the next calculation, the signal power splitting
Fig. 5 . Minimized standard deviation of phase-error a for various line- ratio to the Q-arm is changed with a 1-dB total power
widths at 10 G b / s versus loop delay time.
penalty. The requirement for 6v x T is shown in Fig. 9.
There is an optimum k, which maximizes the requirement
for 6v x 7. This is understood as follows. If the signal
splitting ratio to the @arm k, is decreased, allowable
standard deviation of phase-error is increased and is sat-
urated at 10.8'. Then, the permitted 6v X 7 increases. If
the signal power splitting ratio to the Q-arm k, is de-
creased further, the phase-error variance due to the shot
noise becomes nonnegligible (see (30)). Therefore, the
value of 6v x r should be decreased.
It is assumed above that the thermal noise is negligible.
However, if the thermal noise is not negligible, this sit-
uation is changed because we cannot decrease power to
10-4 4 the @arm so much. To analyze influence of the thermal
lo-3 10-2 lo-l 1 IO 1 ~ 2 lo3 noise, (30) is changed as
NORMALIZED LOOP DELAY TIME zd6v. Rb 2kT
Fig. 6. Normalized loop natural frequency verbus normalized
loop delay time ~m. The solid line is for when the phase-error vari-
ance is minimized. The dotted line is for U,, r = 0.6.
where k (J/K) is the Boltzmann constant and T (K) is
with respect to the normalized loop delay time is plotted absolute temperature. This change can be done when the
in Fig. 7. When 7 = 0, the required beat linewidth cor- thermal noise is assumed to be white noise, which is the
responds to the results given in [ I]. The propagation time, same as the shot noise. In this case, assuming that T is
more than I / & , is not negligible. When the loop delay 300 K, local oscillator power is 3 dBm, bit rate is 10
time T is not negligible, the required beat linewidth 6 v is Gb/s. beat linewidth is 10 kHz, loop delay time 7 = I O
not proportional to the bit rate but limited by the condition ns and 100 ns, the power penalty from ideal detection with
of respect to the local oscillator's power splitting ratio to
@arm kLO is shown in Fig. 10(a) and (b) for a power
6v = 2.04 x W " T . (44) penalty due to the signal power splitting ratio to Q-arm ks
This condition ( 4 4 ) gives I-dB power penalty. of 0.5 and 0.1 dB, respectively. There are k,, which min-
I372 JOURNAL OF LIGHTWAVE TECHNOLOGY. VOL. 9, NO. IO. OCTOBER 1991

15 - 5

Power Penalty due to kS 0.5 dB


I
F - 4
1OGbiUs
s m
s Beat Linewidth 10 kHz
> 10 . >
5
a 5 3
z
W
a T = lOOnsec
n
z
W
n
U U 2
5 5 .
5
g B
0
0
0.001 0.01 0.1 1
LOCAL POWER SPLllTlNG RATIO
Fig. 8. Power penalty from ideal detection due to bur under significant TO Q-ARM kL0
influence of loop delay time r.

5
2.6X103 Power Penalty due to ks 0.1 dB

2.4X10-3 6 4 Beat Linewidth 10 kHr


a
I

>
2.2~16~
I-
iz 3
w
2.0x10-~ n
Lo E 2
5
1.8XlO~
B 1
1.6X103

n
1.4X10.3
0.001 0.01 0.1 1
10-5 104 1i310.' io' 1

SIGNAL POWER SPLITTING RATIO LOCAL SPLITTING RATIO TO Q-ARM k~o


TOQ-ARM % (b)
Fig. 9. Signal power splitting ratio to Q-arm, which minimizes the bit er- Fig. IO. Power penalty of I O Gb/s from ideal detection at bit error rate
ror rate, versus normalized loop delay time d u r . IO-"' with respect to the power splitting ratio to Q-arm of local oscillator
k,.o in case of the thermal noise is not negligible. (a) The power penalty
due to the signal power splitting ratio to Q-arm k, is 0.5 dB. (b) The power
penalty due to the signal power splitting ratio to Q-arm k,y is 0. I dB.
imize the power penalty for given beat linewidth 6 v and
the signal power splitting ratio to the Q-arm ks and loop
delay time 7. I Pulse Panern
Generator 1
Self Heterodyne
Linewidth
IV. PSK HOMODYNE DETECTION EXPERIMENTS Measurement

The homodyne detection experimental set up is shown


in Fig. 11. The transmitter and the local oscillator are
1.55-pm wavelength DFB-LD's with external cavities. I-arm

The output power of the local oscillator is 2.44 dBm. A


fiber-pigtailed traveling-wave electrode Ti : LiNb03 opti- Optical
Hybrid
cal phase modulator is used to produce a lO-Gb/s PSK
signal.
The optical 90" hybrid shown in Fig. 11 is composed
Polarization
Controller - Q-arm
1 Bit

*(-) I
I

of a 1 : 1 half mirror and two polarization beam splitters.


In front of the optical hybrid, there are two polarization Additional
Electrical
controllers that control the polarizations of the transmitted Line

light and the local oscillator's light to be linear and cir- Fig. 1 I . Experimental setup of PSK homodyne detection
cular, respectively [ 111. The four outputs of the 90" hy-
brid are detected by two dual-p-i-n detectors. Ideally, the by changing the direction of the polarization of the trans-
polarization of the local oscillator is circular, so power mitter. In this experiment, the power splitting ratio to the
splitting ratio to the I- and Q-arm is 1 : 1. However, the I- and Q-arm of the local oscillator and that of the trans-
power splitting ratio to the I- and Q-arm can be changed mitter were fixed at 1 : 4 and 2 : 1, respectively. The I-arm
NORIMATSU et al.: PLL PROPAGATION DELAY-TIME INFLUENCE 1373

W
I-
4:
a 10.~
i
- A 440kHZ

a
E 10.’ -
t
10-9 -
0 100 200 300 400 500
r
lo-llL-40 -30
BEAT LINEWIDTH 6v [kHz]
Fig. 13. The power penalty at bit error rate IO-’ versus beat linewidth.
RECEIVED OPTICAL POWER [dBm] (Solid line denotes one obtained through the 2nd Pade approximation.)
(a)

show the calculated degradation considering the propa-


Homodyne gation delay by the 2nd Pad6 approximation mentioned in
Beat Linewidth
Section 111. The experimental results are good agreement
0 180kHz with the calculated ones.
W
I- A 220kHr
a
P: 10-5 V . CONCLUSIONS
K
The loop delay time influence, which is unavoidable in
i?
a the real systems, on optical homodyne detection based on
W 10-7 decision-driven phase-locked loop was investigated the-
c
m Delay Time
4\ oretically and experimentally.
The Pad6 approximation was applied to the calculation
10-9

10 -1’ I
29nsec h\ \ I
of the phase-error variance, which was calculated to
within 0.4% error. Loop natural frequency U,, and line-
-40 -30 width requirement with the nonnegligible loop delay time
RECEIVED OPTICAL POWER [dBm]
for PSK homodyne detection was obtained as U,, <
0.34/7 and 6 v < 2.04 x 10p3/7, where 6 v (hertz) is beat
(b)
linewidth and 7 (second) is the loop delay time. It was
Fig. 12. Error rate performance of PSK homodyne detection for various
linewidths at IO Gb/s. (a) The loop delay time is 13 ns. (b) The loop delay
confirmed that the linewidth requirement becomes 6.2 X
time is 29 ns. l o p 4 Rh (which is already obtained by Kazovsky with the
negligible loop delay time), where Rb (bits-per-second) is
the system bit rate within small delay time limit.
signal is regenerated by a decision circuit (DEC) and the Results were confirmed by lO-Gb/s optical PSK ho-
Q-arm signal is delayed 1 bit. The two signals are multi- modyne detection experiment by using external cavity
plied by an EX-OR and the phase control signal is ob- LD’s. Error-rate performance was measured by changing
tained. The control signal is passed through the loop filter the loop delay time and beat linewidth. Receiver sensitiv-
and applied to laser diode current. The loop filter is the ity degradations were in good agreement with theoretical
first order Lag-lead filter and its parameter in these ex- results. If the PLL propagation time is reduced to 1 ns, it
periments were 71 = 6.8 x lop4 and r2 = 2 X is not negligible more than 1 Gb/s. Therefore, the line-
Since the PLL loop gain depends upon the received power, width requirement of multigigabit PSK homodyne detec-
the dc loop gain K , increased as the received power in- tion system can be limited by the propagation time. The
creased. The dc loop gain K, at the received power -34.3 methods employed here can be used for the design various
dBm was 7 x lop8(hertz). PLL’s, such as balanced PLL’s, heterodyne PLL’s and so
The error-rate performance with various linewidths of on.
lO-Gb/s PSK homodyne detection are shown in Fig.
12(a). The measured loop delay time is 13 ns. To show APPENDIX
degradation with loop delay time, a 16-ns delay line is Here, the solutions of the following integrations are
added in this loop. The error-rate performance with var- given.
ious linewidths and 29-ns loop delay time of lO-Gb/s PSK
homodyne detection is shown in Fig. 12(b). The power
penalty at lo-* error rate is shown in Fig. 13. The lines
I,, = -
I374 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 9, NO. IO. OCTOBER 1991

where ACKNOWLEDGMENT
I1 -I The authors wish to thank Dr. K. Nosu for his encour-
A(s) =
i=O
c a; . s‘ (A2) agement and guidance. They also thank K. Noguchi for
suppling the Ti :LiNb03 optical phase modulator.

B(s) = C
k=O
bk . s k . (A3) REFERENCES
[ I ] L. G . Kazovsky. “Decision-driven phase-locked loop for optical
Through the method in [12], the solutions for y1 = 1, 2, homodyne receivers: Performance analysis and laser linewidth re-
3, 4 are written by quirements,” J . Lightwave Techno/., vol. LT-3, no. 6, pp. 1238-
1247, Dec. 1985.

121 L. G . Kazovsky. “Balanced phase-locked loops for optical homodyne


By using these formulas, the expressions of the phase- receivers: Performance analysis. design considerations, and laser
error variance can be obtained. The expression of the linewidth requirements,” J . Lightwave Techno/.. vol. LT-4. no. 2,
phase-error variance due to the phase noise with nonzero pp. 182-195, Feb. 1986.
loop delay time by the 1st Pad6 approximation and that 131 M. A. Grant, W. C. Michie, and M. J. Fletcher, “The performance
of optical phase-locked loops in the presence of nonnegligible loop
due to the shot noise are propagation delay.” J . Lightwave Techno/.. vol. LT-5, no. 4, pp.
592-597. Apr. 1987.
[4] J. M. Kahn. B . L. Kasper, and K . J . Pollock, “Optical phaselock
receiver with multigigahertz signal bandwidth,” Electron. Lett.. vol.
25, no. 10, pp. 626-628, May 1989.
[5] J . M. Kahn, “ I Gbit/s PSK homodyne transmission system using
phase-locked semiconductor lasers,” lEEE Photon. Techno/. Lett..
2
e
~_ 1 xy(2 - y)(x - 1) + 4x 4 + 649)
vol. 1. no. 10, pp. 340-342. Oct. 1989.
OSNI = [6] J . M. Kahn. A. H. Gnauck. J . J . Veselka, S. K. Korotky, and B. L.
4Rk,P,T 272 XY(Y- 2) 4( 1 - y ) + Kasper. “4-Gb/s PSK homodyne transmission system using phase-
locked semiconductor lasers,” IEEE Photon. Techno/. Lett., vol. 2,
from (20), (23), (25), (27), (A6) and from (22), (26), (27), no. 4, pp. 285-287, Apr. 1990.
[7) S. Norimatsu, K . Iwashita, and K. Sato. “PSK optical homodyne
(A6), respectively. In the same way, these by the 2nd Pad6 detection using external cavity laser diodes in Costas loop.’‘ IEEE
approximation become Photon. Techno/. Lett., vol. 2. no. 5 . pp. 374-376, May 1990.

a.6v 144 - 3 6 + ~72xy~ x2p4


-~ + 6x’y’ 1 2 ~ ’ ~ ’
-
OPN2 = -
144 - 144y - 12xy3 + 36xy2 - 72xy - x2y4 + 6x2y’ 12x2y2 -

144 + 36xy2 - 72wy + + +


144x - x2y4 6x2y’ - 48x2y2 72x2y - x3y4 + 6x3y’ - 12x’y’
(A1 1)
X
144 - + ~ ’72xy - x’y4 + 6x2y’ - 12w2y2
1 4 4 ~- 1 2 ~ ~ 3’ 6 ~ -

from (20), ( 2 3 ) , (25), (28), ( ~ 7and


) from (22), (26), (2g), [SI S. Norimatsu. K. Iwashita, and K . Noguchi, “ l o Gbit/s optical ho-
modyne transmission experiments using external cavity DFB LDs.”
(A7), respectively. In these expressions, x and y are de- Electron. Lett., vol. 26, no. IO. pp. 648-649, May 1990.
fined [9] J. G . Ttuxal, Automatic Feedback Control System Synthesis. New
York: McGraw-Hill, 1955.
x = G7; (A12) [IO] F. M. Gardner, Phaselock Techniques. New York: Wiley, 1979.
[ I l l W. Leeb. “Realization of 90 and 180 hybrids for optical frequen-
7 cies,” Electron. Commun.. AEU37, no. 516, pp. 203-206, 1983.
y = --. (A13) 1121 H. M. James, N . B. Nichols, and R. S. Phillips. Theory of Serwo-
72 mechanisms. New York: McGraw-Hill, 1947.
NORIMATSU PI U / . : PLL PROPAGATION DELAY-TIME INFLUENCE 1375

Seiji Norimatsu was born in Ehime. Japan, on Katsushi Iwashita was born in Kumamoto, Ja-
December 24. 1961. He received the B . S . and pan, on March 6, 1955. He received the B . S . .
M.S. degrees in physics from Osaka University. M.S.. and Dr.Eng. degrees in electrical engineer-
Osaka, Japan. in 1985 and 1987. respectively. ing from Kyusyu University, Hukuoka. Japan, in
In 1987 he joined the NTT Laboratories, Kan- 1977, 1979, and 1987, respectively.
agawa. Japan. His current research interests are in I n 1979 he joined the Yokosuka Electrical
the area of high-speed optical coherent commu- Communication Laboratory, Nippon Telegraph
nication systems. and Telephone Corporation, Kanagawa. Japan.
Mr. Norimatsu is a member of the Institute of where he has been engaged in research on optical
Electronics, Information and Communication fiber transmission systems.
Engineers of JaDan and the Phvsical Societv of Dr. fwashita is a member of the Institute of
Japan. Electronics, Information and Communication Engineers of Japan, and he
received the Young Engineers Award from IEICEJ.

You might also like