White Paper

WDM-PON Technologies

Contents

1- Introduction

2- WDM-PON Architectures
-Type 1 : Tuneable Laser
-Type 2 : Sliced Broadband Source
-Type 3 : Reflective Architectures
a) Injection locked Fabry-Perot Laser
b) Single polarisation RSOA
c) Polarisation independent RSOA
d) Reflective EAM

3- Experimental evaluation of a Single Polarisation RSOA in WDM-PON

4-Conclusion

5- References

6- Appendix ( RF and bias settings)

1
 1. Introduction
Passive optical networks (PONs) were originally developed in the 1980’s [1]
as a cost effective method of sharing fibre infrastructure for narrowband
telephony (TPON) to business premises. Since those early days the
application of PONs has moved on to interactive broadband networks
implemented as either BPON (Broadband PON) or Ethernet PON (EPON) and
now GPON (Gigabit PON) [2].
Optical Network
Unit #1

Central Office
(CO) ONU #2

ONU #3

Passive Optical
Power Splitter ONU #4

ONU #N

Figure 1: Basic PON diagram

All these systems are based on the same idea of time sharing the optical
medium by TDMA (time division multiplexed access). However, it has long
been realised that using wavelength division multiplexing (WDM) offers an
alternative method of sharing the capacity of a PON between multiple users
and would offer advantages in terms of capacity, low latency and service
transparency. Whilst the concept of using WDM within a PON has been
widely demonstrated within research projects it is only recently that the
enabling technology has become sufficiently mature for commercial
consideration.
Recently the interest in WDM -PONs has grown significantly, especially in
parts of Asia, and it is widely believed to be the route towards the next
generation of PONs. The technology challenge for WDM-PON has been to
avoid the need for expensive wavelength selective optical components in
each end-users optical network unit(ONU). In practice this means that it is
cost prohibitive to use the type of lasers currently available for long haul
dense WDM (DWDM) transmission within a WDM-PON. Moreover, it would be
impractical for each customer’s terminal to be built with a fixed single
wavelength laser because managing the inventory of lasers would be
complex and costly for the network operator. For the customers ONU to be
‘colourless’ either a tuneable laser is required or an alternative WDM-PON
network design based on a reflective architecture must be used. In the

2
longer term the tuneable laser approach probably offers the highest
performance WDM-PON with the greatest potential number of wavelength
channels but at present the cost of tuneable lasers is still far too high.
Moreover, a tuneable laser solution may require additional network control
and management to set and maintain wavelengths. The reflective
architecture takes a different approach since all of the individual
wavelengths are provided by a shared network resource, such as shown in
fig 2. In this scheme the upstream transmitter within the ONU only requires
a reflective optical modulator.

Optical Line
Terminal
Upstream (OLT) AWG
Shared Network ONU #1
Source (e.g.
ASE source) ONU X

λ1, λ2, ,λN Downlink RX

Downstream
Signal
R-MOD
Uplink

ONU #N

RX X

Figure 2: Reflective PON diagram

Over the past 5 years there have been a wide range of reflective WDM-PON
architectures reported in the research literature and more recently
extensive network trials have been announced. This white paper first
reviews the alternative approaches to the reflective WDM-PON architecture
and then goes onto to discuss the attributes required from the key optical
component within the ONU .

2. WDM-PON Architectures
Within this paper only the upstream path is considered because this is the
most demanding as far as the cost critical customer’s ONU is concerned. In
principle the same transmission architecture can be used for both upstream
and downstream as might be the case when a symmetric service in terms of
bit rate is required. However, this may not always be the case as higher
data rates are often required for the downstream. A further variation of the
overall architecture is when it is necessary to both wavelength-share and
time-share the PON [3]. In this situation a wavelength multiplex is used to

3
increase the capacity of the primary feeders in the access network and the
distribution PON capacity is then further shared using optical power splitters
in a classical TDMA approach. Where combinations of both time sharing and
WDM are used a greater power budget is required which can be stretching
for some types of reflective architecture PON. In the schemes outlined
below only wavelength sharing of the PON is considered for simplicity.

Type 1 Tuneable Laser Scheme

This uses a tuneable laser within the ONU and is a scheme that offers the
ultimate in terms of optical performance and flexibility. If the splitters in
the network are solely WDM devices, such as planar array waveguide grating
(AWG) devices, the number of ONUs supported will be determined by the
channel spacing of the AWGs and the tuning range of the laser. Use of
broadband splitter/combiners at the distribution point and more wavelength
selective filters at the central office would allow more channels to be
accessed, although available power budget might then be a consideration.
To design a system which used the available spectrum most efficiently and
make good use of the available power budget a combination of both WDM
splitter/combiners and broadband splitter/combiners would be used in the
distribution network.

Central Office Upstream Spectrum

A
Data Out

1530nm-1565nm
A C band

Customers
Optical Receiver AWG Splitter
ONU
Splitters Data In
(or AWG) A

Tx Spectrum

Modulator Tuneable
Laser
˜ 0.1nm

Figure 3: Type 1 architecture: WDM-PON with tunable laser at the ONU.

The problem with the tuneable approach is that a much more sophisticated
laser is required within the customer’s ONU compared to conventional EPON
and GPON systems. Tuneable lasers also usually require internal wavelength
lockers or an external network wavelength reference to ensure they operate
at the correct wavelength channel. Furthermore, to maintain a stable laser
operating regime, external modulation rather than simple direct laser
modulation is the norm. When the PON uses WDM devices for combining
signals, the constraints on wavelength control can be slightly eased because

4
only when the laser is operating at the correct wavelength can the signal
pass through the WDM, so the demands of wavelength control and blanking
during laser start up are eased.

Type 2 Sliced Broadband Source

In this case each ONU contains a broad optical spectrum source within the
transmitter, such as a superliminescent light emitting diode (SLED). The
broad spectral output of the customer’s ONU is connected to one port on a
WDM device, which could be thin film filter or AWG based. Only the optical
spectral components from the LED which can pass through the WDM channel
are transmitted through to the central office and the remaining power is
wasted. Although all the customers ONUs have identical SLEDs, because
each is connected to a different port on the WDM combiner, it is possible to
slice a different part of the available optical spectrum for each ONU.
SLEDs would normally be used as the transmitter device in the customers
ONU although, alternatively, the self-amplified spontaneous emission from a
reflective semiconductor optical amplifier (SOA) can be used instead. The
latter has the advantage of producing a greater optical output power for a
lower drive current but has the disadvantage that it can be sensitive to
optical back reflections resulting in an amplitude ripple in the output
spectra and in the worst case laser action.
When a simple LED is used as the transmitter it is only practical to have a
few upstream data channels of 155Mbit/s from this scheme. With a RSOA as
the slicing source up to 32 X 155Mbit/s has been reported [4]. The number
of channels and the data rate for each channel is determined by the excess
optical intensity noise produced by the slicing process which is itself a
function of the ratio of the bit rate to the optical bandwidth of the sliced
source.
Central Office
Upstream Spectrum
Data Out ˜ 10nm

C BA

C Customers
Optical Receiver WDM Splitter
ONU
B Data In
WDM A
(Slicer)

Spectrum from SLD

Super Luminescent Diode (SLD)

1530nm-1565nm
C band

Figure 4: Type 2 architecture: WDM-PON with broad spectrum optical source for the
uplink at the ONU.

5
Type 3 Reflective Schemes
In this class of schemes a separate sliced source is used to seed the return
path modulators within the customer’s ONU. This has the advantage over
the previous non-seeded scheme because the optical power produced by the
ONU transmitter is now all within the required spectral channel and so none
is wasted. A second advantage with some schemes is that the excess
intensity noise produced by the slicing process can be reduced by gain
saturation effects.
Four types of reflective WDM-PON are outlined below.

Type 3A Reflective : Spectral Slicing with Injection locked Fabry-Perot

Laser
SLD or EDFA
Broadband Source

Seed Spectra
Central Office

Data Out 1530nm-1565nm

C band

AWG
Seed Combiner

Customers
Optical Receiver AWG Splitter ONU
Upstream Spectrum A Data In

Return Spectrum
(with FP modes)

1530nm-1565nm
Laser with
C band
Asymmetric facet
reflectivity
˜ 1nm

Figure 5: Type 3A architecture: WDM-PON with Injection locked FP as ONU transmitter

In this scheme [5] a central broadband seeding source is used, this is

typically the amplified spontaneous emission of an erbium doped fibre
amplifier (EDFA), reflective SOA (RSOA) or SLED. The broadband seed signal
is sliced by the AWG splitter/combiner located within the network. The
sliced continuous wave (cw) seed light is used to “injection lock” a Fabry-
Perot (FP) laser transmitter within the customer’s ONU. This is not the same
as coherent injection locking of two lasers as the ONU laser is effectively
operating as a saturated resonant amplifier. In this scheme the Fabry-Perot
cavity modes which fall within the spectral passband of the AWG will be
amplified and modulated by the ONU laser and returned back through the
AWG. Resonant operation of the reflective transmitter reduces the required
drive current and can reduce the slicing noise through gain saturation. The
laser required is a little different from normal FP transmission lasers
because the laser chip should have lower than normal reflectivity on the
front facet and enhanced reflectivity on the back facet. The chip should

6
also be longer than a conventional FP laser to reduce the cavity mode
spacing to enable more laser longitudinal modes to fall within the spectral
passband of the AWG. This effect can reduce the impact of mode
competition noise. A laser of this type developed at CIP was used in the
demonstration of spectral slicing directed at very-high digital subscriber
line (VDSL) transmission over optical fibre [6]. To optimise the performance
of the injection locked scheme requires control of both the laser parameters
and the injected optical power. The optimisation of the operating point will
also be dependent on the position of the natural gain peak of the laser
compared to the injected spectral slice so is likely to vary between ONU’s.

Type 3B Reflective : Spectral Slicing with single polarisation RSOA

The reflective SOA (RSOA) is in some ways related to both the asymmetric
FP laser and the SLED, however, it also has important differences. By
reducing the front facet reflectivity close to zero it is possible to suppress
all of the natural cavity modes, but unlike the SLED it is necessary to also
ensure that there is an efficient reflection from the back facet. The basic
concept of the RSOA is not new [7] but designs have been refined since by
the use of angled stripes and mode expansion to reduce output facet
reflectivity and improve fibre coupling. RSOAs can be designed to be either
input signal polarisation dependent or polarisation independent. Generally
the polarisation dependent devices have the best temperature performance,
up to 70deg C and by trading polarisation sensitivity offer greater design
freedom in most other parameters.
The single polarisation RSOA can be used in essentially the same system
architecture as the asymmetric Fabry-Perot laser scheme 3a, since the
system uses an un-polarised seed source.
SLD or EDFA
Broadband Source

Seed Spectra

Central Office

1530nm-1565nm
C band

Data out
AWG
Combiner

Customers
Optical Receiver AWG Splitter
ONU
Data In
Upstream Spectra

Return
Spectra
1530nm-1565nm
C band Reflective Semiconductor
Optical Amplifier (RSOA)
˜ 1nm

Figure 6: Type 3B architecture: WDM-PON with single polarisation RSOA as ONU

transmitter .

7
The seeded RSOA approach has an advantage over the un-seeded SLED
approach (type2) because the optical power will be higher in the selected
spectral slice due to the optical gain of the RSOA. Moreover, the use of gain
saturation can be used to “squeeze” the excess noise produced by slicing [8].
This will allow higher figures of merit ( channels X bit rate) to be achieved
compared to un-seeded slicing and in practice 32 channels of Gigabit
Ethernet (GbE) is achievable over a 20km reach PON. The maximum speed
that the RSOA can be modulated at is determined by carrier lifetime and the
ability to provide up to 100mA of electrical drive current. When the device
is operated under gain saturation the effective lifetime is reduced by
clamping of the carrier level and in practice modulation speeds of
2.5Gbit/s are possible.
For a longer reach PONs operating over longer distances from the central
office impairments from chromatic dispersion can result when using a
filtered broadband seed source. For these longer reach systems [3] it is
better to use a comb of wavelengths generated by a bank of single-wave
lasers as the seed source, but this usually requires the use of a polarisation
independent RSOA within the customers ONU.
Type 3C Reflective: Spectral Slicing with polarisation independent RSOA
In this scheme the RSOA within the customer’s ONU has polarisation
independent gain (SOA-RL-OEC-1550) [9] . This means that the seeding
source can now be polarised and derived from a bank of single wavelength
lasers. The use of a spectrally pure seed reduces the impact of chromatic
dispersion for the return channel in long reach applications and avoids the
excess noise produced by the slicing process. In principle more wavelength
channels and at higher bit rates are possible. However, in practice another
noise source, coherent Rayleigh backscatter, [10] must also be considered in
situations where a single shared feed fibre is used for both the seed
wavelength and modulated return signal. This problem can be reduced by
using a separate fibre for the seed and return, as is shown below.
Seed Laser Bank
Seed lasers
Central Office

AWG 1530nm-1565nm
C band

Data out Seed AWG

Combiner

Customers
Optical Receiver AWG Splitter
ONU
Upstream Spectrum
Data In

Return
1530nm-1565nm Spectra
C band

Reflective Semiconductor
<0.1nm
Optical Amplifier (RSOA)

Figure 7: Type 3C architecture: WDM-PON with polarisation independent RSOA as ONU

transmitter.

8
Type 3D Reflective Architecture: Reflective EAM
The carrier lifetime modulation speed limitation when using a RSOA as the
transmitter in the reflective architecture can be eliminated by using a
reflective electro-absorption modulator (REAM) (R-EAM-1550) [9]. With this
reflective device, data modulation rates of 10Gb/s or higher (>40Gb/s) are
possible. Obviously by using a REAM rather than the RSOA configuration
there is no optical amplification at the ONU so all of the optical power
needed to overcome the loss budget must be available from the central
office seed lasers. For high bit rate systems it is most likely that lasers
would be used as the seed sources and therefore for some applications dual
fibre feeding may be necessary to reduce coherent Rayleigh back scatter
noise. For lower data rates the REAM scheme is attractive when it is
necessary to minimise the electrical power within remote equipment
because the EAM requires a low drive voltage (<2V) and has intrinsically high
electrical impedance. Low power consumption is of particular interest
where EAM is used for VADSL line extension using WDM-PON [6] and radio-
on-fibre applications [11].

Seed lasers
Central Office
AWG

Data out Seed

Customers
Optical Receiver AWG Splitter ONU
Data In

AWG
Combiner

Reflective Electro-Absorption
Modulator (REAM)

Figure 8: Type 3D architecture: WDM-PON with reflective EAM as ONU transmitter.

For the longer term, new types of devices are likely to be available for long
reach high split architectures [3]. These devices will combine the higher
modulation speed of the REAM with the gain and output power of the RSOA.

9
 Table 1: Comparison of Upstream WDM-PON Architectures
Scheme Description Channel Number Positives Negatives
Bit rate of λ
Channels
1 Tuneable High High .Power .ONU
laser ONU budget complexity
.No seed .Additional
>10Gb/s >32
source network
control
.λ
Scalability
2 Spectrum Low Low .Cheap .Poor
slicing with ONU scalability
LED based .No seed
ONU <155Mb/s <32
source

3A Reflective Medium Medium .No cooler .long term

Architecture: in ONU stability?
Spectral .Limited
Sliced ILD <2.4Gb/s 32
scalability
.Seed source
3B Reflective Medium Medium .No Cooler .Limited
Architecture in ONU scalability
:Spectral
.Seed source
Sliced with <2.4Gb/s 32
RSOA based
ONU.
3C Reflective Medium High . λ Scalabil .Cooler
Architecture ity required
>32
: Central .Rayleigh
laser bank <2.4Gb/s
noise
with RSOA
based ONU. .Seed source
3D Reflective High Low .High bit .Poor
Architecture: rates scalability
<32
Central laser .Low .Rayleigh
bank with >10Gb/s
electrical noise
REAM based power
ONU. .Seed source

10
 3. Experimental Evaluation of RSOA in WDM-PON

A CIP single polarisation RSOA [9], SOA-R-OEC-3169 was tested in a type 3B

architecture. The device offers extended temperature operation as shown in
figure 9. A stable gain at 1550nm, ~27dB, can be obtained between 10 and
70˚C, when the device bias current is varied between 15 and 80mA. The
small signal bandwidth (measured for PinSOA=-10dBm and DC bias between 20
and 70mA) was measured to be around 1GHz between 20 and 80˚C.

29
27
25
Gain (dB)

23
21
19
17
15
0 10 20 30 40 50 60 70 80 90

Temp (C)

Figure 9: Continuous Wave (CW) Gain at 1550nm of RSOA versus Temperature. The
RSOA bias is adjusted at each temperature to keep the gain constant.
Figure 10 shows the Amplified spontaneous emission (ASE) at each
temperature and bias current settings. It can be observed that, as the
temperature increases, the ASE broadens and its peak shifts to longer
wavelengths.

80mA_80C 80ma_70C 45mA_60C 35mA_50C

30mA_40C 25mA_30C 19mA_20C 15mA_10C
-20
-23
-26
-29
ASE (dBm)

-32
-35
-38
-41
-44
-47
-50
1460
1470
1480
1490
1500
1510
1520
1530
1540
1550
1560
1570
1580
1590
1600
1610
1620
1630
1640

wavl(nm)

Figure 10: ASE of RSOA at different temperature and drive settings.

11
As a consequence of the different operating condition of the RSOA when the
temperature is varied, the saturated input and output powers of the device
change as well. The 1550nm results are shown in figure 11.

In Out
0 15
13
-5 11

Psat_out (dBm)
Psat_in (dBm)

-10 9
7
-15 5
3
-20 1
-25 -1
-3
-30 -5
0 10 20 30 40 50 60 70 80 90
Temp (C)

Figure 11: CW Saturated Input and Output power versus Temperature. The device bias
varies at each temperature.
The system architecture is shown in figure 12. At the central office, the un-
polarised broadband ASE source for the upstream signal, is provided by a 2
stage EDFA amplification. The 5nm bandwidth filter in between, centred at
1550nm, is needed to concentrate the second EDFA output power at
1550nm. The attenuator is used to vary the power reaching the ONU which
is monitored via the 10% tap. Between the CO and the ONU, a 100GHz AWG
slices the EDFA spectrum. Different lengths of SMF-28 fibres can be inserted
after the AWG. At the ONU, the RSOA is DC and RF biased via a Bias Tee.
The device drive was varied depending on the temperature setting of the
device. During the experiment the DC drive was set in order to minimize
the Bit error rate at the Receiver and optimise the sensitivity. The RF
voltage (AC coupled) was set to optimise the extinction ratio with the eye "0
level" down to real zero (i.e. with the RSOA switched off). The upstream
signal, after passing through the AWG, is sent to the pre-amplified receiver.
The system sensitivity is measured with the attenuator A. The signal is then
amplified and filtered and detected using a Lightwave converter which
output is filtered using a low pass 3GHz filter. The electrical signal is
evaluated using an oscilloscope for eye quality and extinction ratio and
analysed for Q and BER via a BER detector.

12
 Figure 12: Reflective WDM-PON experimental setup with RSOA at the the ONU.
The system performance in a back to back configuration at 1.25 Gb/sec was
measured. Sensitivity, Q and BER were assessed versus RSOA operating
temperature and RSOA Optical Input power. Measurements were carried out
with both 10km and 20km of SMF-28 fibre between the CO and ONU.
For the back to back experiment at 1.25Gb/s, the RSOA RF drive was a PRBS,
231-1 NRZ pattern (detailed voltage and bias settings can be found in the
appendix). System performance was assessed for 2 possible configurations:
with the circulator and receiver as in figure 12, at the CO, or with the
circulator placed between the AWG and the ONU. In the later scenario we
investigated any possible penalty coming from the fact that the signal that
reaches the receiver, which includes the RSOA ASE, doesn’t get filtered at
the AWG. Results are shown in figure 13. The two plots are very similar and
show the beneficial effect of noise “squeezing” at high input optical power.
For example at 20˚C, the sensitivity and Q value are stable for Pin_SOA>-9dBm
(at this power the SOA is well into saturation, figure 11). In fact, at low
input powers, when the RSOA is not into deep saturation, the system is
limited by the noise from the broadband ASE source that limits the BER and,
hence the sensitivity. The benefits of intensity noise suppression by the
saturated RSOA are also visible in the BER curves of figure 15. Only for
Pin_SOA≥-17dBm the BER curves are not limited by a noise floor and low BER
(1*10-11) is achievable: the BER curve gradient increases with increased
Pin_SOA. While 20 and 40˚C results are comparable, the system performance
is slightly poorer at higher RSOA temperatures. Nevertheless the system was
still error free at 80˚C. As the RSOA unsaturated gain is over 30dB at room
temperature, the device has still a reasonable gain when in the “noise-
squeezing” regime, as shown in figure 14.

13
 Sens_20C Sens_40C Sens_60C Sens_70C
Sens_80C Q_20C Q_40C Q_60C
Q_70C Q_80C
-20 24
-22 22
-24 20
-26 18
-28 16
S (dBm)

Q(dB)
-30 14
-32 12
-34 10
-36 8
-38 6
-40 4
-42 2
-44 0
-27 -24 -21 -18 -15 -12 -9 -6 -3 0 3 6
Pin_SOA (dBm)

Sens_20C Sens_40C Sens_70C Sens_80C

Q_20C Q_40C Q_70C Q_80C
-20 24
-22 22
-24 20
-26 18
-28 16
S (dBm)

Q(dB)
-30 14
-32 12
-34 10
-36 8
-38 6
-40 4
-42 2
-44 0
-27 -24 -21 -18 -15 -12 -9 -6 -3 0 3 6
Pin_SOA (dBm)

Figure 13: Sensitivity and Q at 1.25Gb/s for the back to back system. The top plot is for
the system with the circulator near the RSOA , while the bottom plot is the for the
scheme with the circulator at the CO.

20C_50mA 40C_50mA 60C_60mA

70C 70mA 80C 70mA
30
26
22
Gain (dB)

18
14
10
6
2
-2
-27 -24 -21 -18 -15 -12 -9 -6 -3 0 3 6
Pin_SOA_TOTAL(dBm)

Figure 14: Gain (under 1.25Gb/s modulation) of the RSOA when inserted in the WDM-
PON 1.25Gb/s system at different temperatures and bias settings.

14
 "" B2B_-25dBm_50mA_20C
B2B_-17dBm_50mA_20C B2B_-15dBm_50mA_20C
-4 B2B_-7dBm_50mA_20C B2B_-13dBm_50mA_20C
B2B_-11dBm_50mA_20C B2B_-9dBm_50mA_20C
B2B_-7dBm_50mA_20C B2B_-5dBm_50mA_20C
B2B_-2dBm_50mA_20C B2B_0dBm_50mA_20C
Log(Log(BER)) -5 B2B_2dBm_50mA_20C B2B_3-7dBm_50mA_20C
B2B_-19dBm_50mA_20C

-6

-7

-8

-9

-10

-11
-12
-44

-42

-40

-38

-36

-34

-32

-30

-28

-26
Received Power (dBm)

Figure 15: BER curves (at 1.25Gb/s) versus Received power at different RSOA input
powers. RSOA at 20˚C.

Figure 16: 1.25Gb/s eye diagrams. Top row, from left to right: Pin_SOA=-19, -15 and
+1.4dBm and RSOA kept at 20˚C and DC current of 40mA (RF drive of 4.7V). Middle row,
from left to right, Pin=-15 and +1.4dBm and RSOA kept at 70˚C, and DC current of
65mA (RF drive of 6.5V). Bottom row, Pin=0dBm, RSOA kept at 80˚C, DC drive of 70mA
and RF drive of 6.5V.

15
This noise filtering is also visible in the eye diagrams of figure 16: while at
low input power the eye shows a significant amount of noise in the “1 level”
of the eye, the eye opens up at higher injected powers. Extinction ratios
(XR) were all pretty good, between 10.3 and 12dB at the different
temperatures. The improvement in performance at high injected power
comes also from the enhanced electrical to optical bandwidth of the device
when in saturation.
For the transmission experiment (over 20km of SMF-28) at 1.25Gb/s, the
RSOA RF drive was a PRBS, 27-1 NRZ pattern. Results are shown in figure 17.
The figure shows as well the back to back results for an easy comparison of
the data. Due to the 5dB insertion loss of the fibre, there is less optical
power available to saturate the RSOA. Therefore a reduced “noise
squeezing” beneficial effect will be available. Nevertheless, the results are
good and low BER (1*10-11) were obtained as shown in the BER plots of figure
19. The eye diagrams (fig. 18) show a significant amount of noise at low
injected power on the “1” level but the eyes open as the SOA gets more into
saturation. Eye XRs varied between 10.5dB and 12.3dB.
The plot of figure 20 shows the minimum amount of RSOA input power
required at the different temperatures to get a BER of 1*10-9 for the back to
back and transmission cases. The two cases’s required powers are very
comparable. The discrepancy at 70 degree could be due to a better DC and
RF driving of the RSOA in the transmission case.
Sens_20C Sens_40C Sens_70C Sens_80C
Q_20C Q_40C Q_70C Q_80C
-20 24
22
-24 20
18
-28 16
S (dBm)

Q(dB)

14
-32 12
10
-36 8
6
-40 4
2
-44 0
-21 -18 -15 -12 -9 -6 -3 0 3
Pin_SOA_TOTAL(dBm)

Sens_20C Sens_40C Sens_70C Sens_80C

Q_20C Q_40C Q_70C Q_80C
-20 24
-22 22
-24 20
-26 18
-28 16
S (dBm)

Q(dB)

-30 14
-32 12
-34 10
-36 8
-38 6
-40 4
-42 2
-44 0
-21 -18 -15 -12 -9 -6 -3 0 3
Pin_SOA (dBm)

Figure 17: Sensitivity and Q at 1.25Gb/s for the back to back (top plot) and 20km
transmission (bottom plot) systems

16
 .

Figure 18: 1.25Gb/s eye diagrams for the transmission system. From left to right:
T=20˚C & Pin=-19dBm, T=20˚C & Pin=-3.7dB, T=70˚C & Pin=-20dBm, T=70˚C & Pin=-
3.7dBm

"" B2B_-10dBm_42mA_20C_20km
B2B_-5dBm_42mA_20C_20km B2B_-3.7dBm_41mA_20C_20km
-4 B2B_-15dBm_42mA_40C_20km B2B_-10dBm_42mA_40C_20km
B2B_-15dBm_42mA_20C_20km B2B_-5dBm_42mA_40C_20km
B2B_-3.7dBm_42mA_40C_20km B2B_-15dBm_66mA_70C_20km
B2B_-10dBm_66mA_70C_20km B2B_-3.7dBm_66mA_70C_20km
B2B_-20dBm_66mA_70C_20km B2B_-5dBm_66mA_80C_20km
-5 B2B_-19dBm_42mA_20C_20km
Log(Log(BER))

-6

-7

-8

-9

-10

-11

-12
-42

-40

-38

-36

-34

-32

-30

Received Power (dBm)

Figure 19: Transmission over 20km of SMF-28. BER curves (at 1.25Gb/s) versus Received
power at different RSOA input powers and Temperature settings.

17
 20km Back to Back
0

Pin for BER=1*10-9 (dBm)

-5

-10

-15

-20

-25
0 20 40 60 80 100
Temp (C)

Figure 20: Minimum RSOA input power to have BER of 1*10-9 in the back to back and
transmission case at 1.25Gb/s versus RSOA temperature.

4. Conclusion

WDM-PON is becoming of great interest to operators as a means of

increasing the capacity of optical access networks. WDM-PON also offers the
possibility of both sharing fibre plant and at the same time providing a
virtual point-to-point link from the central office to the end customer which
gives both flexibility and service transparency not possible with a classical
TDMA PON. The reflective WDM-PON architectures avoid the need for
wavelength specific components within the customers ONU and gets round
the usual inventory problem of WDM systems. There have been significant
advances in the reflective components required to enable the colourless
ONU and there are now RSOA and REAM devices available for a wide range
of possible architectures .

5. REFERENCES

1. TPON-a passive optical network for telephony,

J R Stern et-al, ECOC 1988
2. ITU-T Recommendation G.984.1: “General characteristics for Gigabit-
capable Passive Optical Networks”
3. Progress in PON research in PIEMAN and MUSE, Davey R, NOC2006
4. Demonstration of DWDM-PON Employing Spectrum-sliced RSOA, S B park,
et-al, ECOC 2006

18
5. Fiber-to-the-home services based on wavelength division multiplexing
passive optical network, S J Park, Journal of Lightwave Technology,
Vol22, no 11, November 2004
6. WDM-based optical feeder for VDSL with electrical powering from the
customer premises”, A. Borghesani et-al, ECOC2006.
7. US Patent US5015964 published 1987
8. Spectral slicing WDM-PON using wavelength-seeded reflective SOAs”, C
Ford et-al, Electron. Lett., Vol 37, No 19, 2001
9. see data sheets www.ciphotonics.com
10. Rayleigh backscattering impairments in access networks with centralised
light source, G. Talli et-al, Electronics Letters, vol. 42, no. 15, 2006
11. Passive picocell:a new concept in wireless infrastructure, D Wake et-al,
Electronics Letters, 33 no5, 1997.

6. Appendix ( RF & Bias Settings)

ƒ RF and DC bias settings for 1.25Gb/s experiment (back to back

configuration).
Temperature ˚C DC Bias Current (mA) RF Voltage (peak to peak) (V)
20 40 4.7
40 40 4.7
60 50 5.6
70 65 6.5
80 70 6.5

ƒ RF and DC bias settings for 1.25Gb/s experiment (transmission over

20km of SMF-28).
Temperature ˚C DC Bias Current (mA) RF Voltage (peak to peak) (V)
20 42 4.7
40 50 4.7
70 66 7.2
80 66 7.2

19