06/30/2005,Jacob Cai,
Sinclair Li, Douglas Cheng
Fiberxon Inc.
Introduction
With the development
of optical access network
s technology,Passive Opti
cal Networks (PON) is mak
ing rapid progress in the
world's leading technology-adopter markets. Accompany with the
Request For Proposal (RFP) of PON equipments from U.S. RBOCs an
d NTT, the promise of delivering last-mile optical connections t
o residential customers becomes more realizable. As the deployme
nt of Gigabit Ethernet PON (GEPON) systems becomes increasingly
important, the performance evaluation of the corresponding syst
em is crucial. The relative intensity noise (RIN) is one of the
limiting factors in the transmission of analog or digital signa
ls over fiber and must be characterized. It may reduce the Signa
l-to-Noise ratio (SNR) and cause an obvious performance degradat
ion of GEPON system. This paper presents a detailed analysis of
the RIN in component level and system level respectively. The a
uthors also describe an optimum measurement method of RINxOMA i
n GEPON system in great detail.
Theoretical analysis of RINlaser and RINsystem
The noise at the receiver output in GEPON system mainly resu
lts from three fundamental contributions: laser intensity nois
e, thermal noise and photonic shot noise.
Laser intensity noise, NL, refers to the noise generated by
the laser. The intensity of semiconductor lasers fluctuates duri
ng operation due to the quantum nature of the lasing process. In
tensity noise is caused by optical interference between the lase
r signal and spontaneous emission within the cavity. In additio
n, operating conditions, such as bias level and modulation frequ
ency, also affect the noise level directly. The presence of exte
rnal feedback or reflection will increase the Laser intensity no
ise as well.
Thermal Noise, Nth, limits the sensitivity of the receiver a
nd restricts the transmission distance of both analog and digita
l signal in GEPON system.
Shot noise, Nq, is produced by the quantum nature of photon
s arriving at the detector, and related detection statistics. Th
e noise level is directly related to the amount of incident ligh
t on the photodetector.
Thermal noise and laser intensity noise vary with frequency. Sho
t noise is a function of the intensity of incident light on the
photodiode.
The total system noise, NT(f), is the linear summation of these
three noise terms:
(1)
where NL(f) is the laser intensity noise power per Hz; Nq i
s the photonic shot noise power per Hz; Nth(f) is the contributi
on of thermal noise power per Hz.
RIN describes the amplitude fluctuations in the optical field. I
t is defined as the ratio of mean-square optical intensity nois
e to the square of the average optical power, the unit is dB/H
z:
(2)
where is the mean square intensity fluctuation spectral de
nsity of the optical signal and P is the average optical power.
As the ratio of optical powers squared is equivalent to the rati
o of the detected electrical powers, RIN can be expressed in ter
ms of detected electrical powers:
(3)
where N is the power spectral density of the photocurrent a
t a specific frequency, and Pave is the average power of the pho
tocurrent.
If the dominant noise source in the system is laser intensit
y noise, the maximum realizable SNR increases as RIN decreases:
(4)
where m is the index of modulation and B is the noise bandwidt
h.
Distinguish between RINlaser and RINsystem
RIN is the ratio of noise power to average power. Equations
(1) and (3) can be used to distinguish the value of RINLaser an
d RINsystem.
(5)
And RINLaser is only the ratio of laser intensity noise powe
r to average power:
(6)
Figure 1. RINlaser calculated from the measured RINsystem, v
ersus average power.
Figure 1 describes RINlaser against RINsystem to show the ef
fect of subtracting the thermal and shot noise with varying aver
age power and constant thermal noise (8 dB NF) and responsivity
(0.8 A/W into 50 ohms). The figure shows that when the total me
asured noise is greater than the thermal and shot noise terms b
y about 5 to 10 dB, the values of RINsystem and RINlaser are ess
entially equal.
RINxOMA
(a) for various levels of RIN (b) with reduced extinction ratio
Figure 2 BER curves for various RIN
The curves presented in Figure 2(a) is achieved by assuming
the laser is fully modulated. In order to achieve good performa
nce in GEPON systems, designers prefer to operate the laser at r
educed modulation amplitude. By reducing the optical modulation
amplitude (OMA) while maintaining the same average power, the la
ser dynamic response characteristics is generally improved. The
Fibre Channel specification is specifically written to permit la
sers to be operated at reduced modulation (reduced extinction ra
tio). Instead of specifying minimum average power and extinctio
n ratio, a specification is only given for the minimum amount o
f transmit OMA. This gives the designer freedom to select the op
timum extinction ratio for best transmit eye quality. However, t
ransmit RIN plays an important role in determining how low an ex
tinction ratio can be used.
Figure 2(b) illustrates the impact of RIN on BER performanc
e when the same laser employed in Figure 2(a) is operated at a r
educed extinction ratio. Since the Extinction Ratio is less, th
e required average power for a given BER is higher as seen in th
e left most curve.
To permit the operation of optical sources at levels other t
han full modulation, but not experience the added RIN degradatio
n in Figure 2(b), IEEE 802.3ah adopts a more challenging specifi
cation for transmitter RIN identified as RINxOMA. It requires th
e transmitter RIN ratio to be calculated using the same RMS modu
lation amplitude of the laser instead of an average DC level. RI
NxOMA can be expressed as:
(7)
where
RINxOMA is the Relative Intensity Noise referred to optical
modulation amplitude measured with x dB reflection;
PN is the Electrical noise power in Watts with modulation off;
PM is the Electrical power in Watts with modulation on;
BW is the Low pass bandwidth of apparatus - high pass bandwidt
h of apparatus due to DC blocking capacitor [noise bandwidth of
the measuring system (Hz)].
Equation 7 shows that as one attempts to reduce the extincti
on ratio of the laser by decreasing the modulation, the value o
f PM will decrease proportionally and result in an increment of
RINxOMA.
Relationship between RINxOMA and RIN
Comparing the two definitions of RINxOMA and RIN, we can der
ive the following relationship between these two parameters:
(8)
where PER is the extinction ratio penalty given by
(9)
Where P1 and P0 are optical power for “1” and “0” levels of
the optical signal respectively.
Measurement of RINxOMA
RIN can be measured by capturing the electrical output of a
high-speed photodetector on an electrical spectrum analyzer. Al
ternatively, RIN can be measured using one lightwave signal anal
yzer only. A setup for measuring RIN in GEPON is shown in Figur
e 3. It contains the transmitter under test, optical splitter, a
ttenuator, polarization rotator, Optical to Electrical converte
r O/E and measuring equipment.
Figure 3 RINxOMA measurement setup
The optical path and detector combination must be configure
d for a single dominant reflection with an optical return loss.
The length of the fiber is not critical but should be in exces
s of 2 m.
The polarization rotator is capable of transforming an arbit
rary orientation elliptically polarized wave into a fixed orient
ation linearly polarized wave. The purpose is to remove the pola
rization dependence of the devices, which used in the test syste
m, like splitter and O/E converter.
The optical splitter permits the laser output to be monitore
d with the measurement equipment, while at the same time generat
ing a controlled amount of reflection back to the laser.
Various link physical interface standards require that a por
tion of the light transmitted in the forward direction be reflec
ted back towards the laser. This is to insure the transmitter RI
N is being characterized under a more stressful, but realistic,
link condition. Reflections can significantly degrade laser RI
N performance as a portion of the light can couple back through
the optics and into the laser die itself. This can create insta
bilities that appear RIN like in nature and increase the effecti
ve noise output of the laser. The potential exists in fiber opti
c links for back reflections to be present.
Both the OMA power and noise power are measured by AC coupli
ng the O/E converter into the electrical power meter. If neede
d, an amplifier may be used to boost the signal to the power met
er. A low pass filter is used between the photo detector and th
e power meter to limit the noise measured to the passband approp
riate to the data rate of interest. In order to measure the nois
e, the modulation to the DUT is turned off.
The upper –3 dB limit of the measurement apparatus is as sp
ecified for the transmitter optical waveform test. The bandwidt
h used in the RIN calculation takes the low-frequency cutoff of
the DC blocking capacitor into consideration. The low-frequency
cutoff is recommended to be less than 1 MHz. The filter should b
e placed in the circuit as the last component before the power m
eter so that any high-frequency noise components generated by th
e detector/amplifier are eliminated. If the power meter used ha
s a very wide bandwidth, care should be taken to ensure that th
e filter does not lose its rejection at extremely high frequenci
es.
The RMS electrical power meter should be capable of being ze
roed in the absence of input optical power to remove any residua
l noise.
Equipment selection
Various measurement equipments can perform the task of RINxO
MA test. Below is one recommended solution:
O/E converter:
Agilent 83440C, 81000FI
Polarization controller
Agilent 8169A
Varible refelector
Agilent 8163B, 81570A, 81000FI x2, 81000BR, 81000UM
Spectrum analyzer:
Agilent E4440A
It is worth to point out that Agilent 71400C or 71401C is co
mbined with the spectrum analyzer and O/E converter. They are th
e preferred equipment in test platform setup.
Test procedure
Use the following procedure to test RIN optical modulation a
mplitude:
a) With the DUT disconnected, zero the power meter;
b) Connect the DUT, turn on the laser, and ensure that the lase
r is not modulated;
c) Operate the polarization rotator while observing the power me
ter output to maximize the noise read by the power meter. Note t
he maximum power, PN;
d) Turn on the modulation to the laser using the pattern specifi
ed for the PMD type and note the power measurement, PM. It may b
e necessary to change or remove the effective reflection to obta
in an accurate reading;
e) Calculate RIN from the observed electrical signal power and n
oise power using equation 7.
RINxOMA parameter Requirement in IEEE 802.3ah
IEEE 802.3ah specifies the value of RINxOMA in detail, as follow
ing table describes:
As a globally leading vendor of GEPON transceiver, Fiberxon
apply very strictly requirement of the measurement on RIN15OMA
of GEPON OLT/ONU transceivers. Readers can refer to the produc
t datasheets for more detail.
Conclusion
The level of relative intensity noise reflects the amplitud
e fluctuations in the optical field. These fluctuations propagat
e over fiber and limit SNR at the receiver. Fiberxon provides GE
PON OLT/ONU transceivers with excellent RIN performance by stric
tly product design, manufacture and measurement. All of the rela
ted products are fully compliant with the requirement of GEPON a
pplication.