Assessment of improvement of extension of reach of 10 Gbps PONs by using APDs

Transcription

1 MS.C. DISSERTATION ON ELECTRICAL AND COMPUTER ENGINEERING, JULY 05 Assessment of improvement of extension of reach of 0 Gbps PONs by using APDs João A. R. Mourato Optical Communications and Photonics Group of Instituto de Telecomunicações, Department of Electrical and Computer Engineering, Instituto Superior Técnico, Universidade de Lisboa, joao.mourato@tecnico.ulisboa.pt Abstract The 0 gigabit per second passive optical network (XG-PON standard was proposed to be employed in optical fibre telecommunication systems due to the increase in binary rate, possibility of more users in the same PON and non-intrusive deployment on the existing optical access network. The PON systems can use the Positive-Intrinsic-Negative diode (PIN or the Avalanche Photodiode (APD as photodetector. However, the high bitrate increases the effect of the mode partition noise (MPN. Therefore, the study of the photodetector performance in the presence of the MPN is of special concern. In this dissertation, the models that characterize the APD and MPN are described and their impact on the performance of 0 Gbps PON is assessed through numerical computation. In addition, for optimizing the performance of the APD on the XG-PON system, for non-null extinction ratio, an expression for the APD sensitivity is proposed. The XG-PON, comprising the use of multi-longitudinal mode (MLM lasers or single longitudinal mode (SLM lasers at the optical line termination (OLT and the optical network unit (ONU, is evaluated. For a dispersion parameter value of 8.4 ps/(nm.km in the downstream and a dispersion parameter value of 3.68 ps/(nm.km in the upstream, MLM lasers cannot be used in either direction. Moreover, the effect of the MPN on the signal from SLM lasers is negligible, provided that the side-mode suppression ratio (SMSR is high (SMSR 30 db. Furthermore, the average power level emitted by the ONU imposes a limit in the splitting ratio and maximum distance. Index terms XG-PON, avalanche photo-diode, mode partition noise, multi-longitudinal mode laser, single-longitudinal mode laser. I. INTRODUCTION The evolution of telecommunications technologies and its rapid spread led to a remarkable growth in the number of applications and services available to the users, resulting in dramatic increase of demanded bandwidth. Thus, it was necessary to come up with a different approach in all segments of the network, in order to satisfy such demand. Focusing on the first mile (also known as the access network, and in the new generation network technologies, the natural course of action is to migrate (when possible from copper solutions towards fibre solutions, enabling the network to meet the demand. With that purpose in mind, the concept of Passive Optical Networks (PON emerged in the mid 90s when the Full Service Access Network (FSAN group started working on fibre to the home (FTTH architectures. In 00, the FSAN started the development of an enhanced standard that could support multiple services in their native form, improve the total bandwidth and its efficiency, upgrade the security and management mechanisms in an evolutionary form. The result was the G.984 recommendation, also known as Gigabit passive optical network (GPON. Recently, extensions to the standards were made, resulting in the XG-PON (or 0G-PON as the ITU-T G.987 recommendation, on which this work will be focusing. However when increasing the system rate to a 0 Gbit/s PON, some limitations arise. This limitations come in many forms, however in the case where the system s transmission rate is increased, an effect with great influence on the system is the Mode-Partition Noise (MPN. Also, the sensitivity of the receiver itself has performance issues when there is such an increase in the data rate. Thus, the need for an evaluation of the receiver, which can be from a common PIN diode receiver to an Avalanche photo-diode (APD receiver. The main objective of this work is to assess the reach extension of the XG-PON by using APD receivers in the presence of MPN, through numerical simulation. Additionally, the APD is optimized and a model for MPN in SLM lasers is proposed. This paper is structured as follows. In section II, the XG- PON is briefly introduced. In section III, the APD receiver is analysed and expressions for the sensitivity and the optimum gain are proposed. In section IV, the MPN is analysed and a model for evaluating BER in SLM lasers is proposed. In section V, the assessment of XG-PON reach is presented. Finally, in section VI, the final conclusions of this work are drawn. II. XG-PON XG-PON was designed based on the existing GPON system, as a kind of an improvement to the previous generation. The XG-PON system inherits: the TC layer principles; the dynamic bandwidth allocation; QoS and traffic management; the remote operation of ONU (optical network unit through OMCI (redefined on G.988. This recommendation also included improvements to the existing system, namely: enhanced power saving options; synchronizing options enabling mobile back-hauling applications; upgrading the performance monitoring; the optical distribution network (ODN; the security mechanisms []. On the PMD layer there is a difference on the downstream/upstream rate combination. There are two standards, 0/.5 Gbps (asymmetric on XG-PON and 0/0 Gbps (symmetric on XG-PON. The wavelengths chosen

2 MS.C. DISSERTATION ON ELECTRICAL AND COMPUTER ENGINEERING, JULY 05 were nm for the downstream and nm for the upstream, allowing the coexistence with the previous generation and RF overlay video []. GPON was class B+ (allows for loss up to 8 db, and the coexistence of the two systems implicates the use of a filter, which almost certainly will introduce additional loss. Also, some deployed systems were designed with a bit more loss than required, for commercial reasons, therefore two nominal budgets were introduced: nominal that goes up to 9 db; the nominal, for the over designed deployed systems, that goes up to 3 db. In the GPON system an extended loss budget was developed that had two major features: 4 db more loss than the nominal budget, and ONU specifications that were unchanged from the nominal budget. After consideration, the same design features were reused in XG-PON leading to two extended budgets of 33 and 35 db. There are no definitions on the receivers because both PIN and APD have their specific advantages on the system. The decision is left to be ruled by the market over the years. XG-OLT G-OLT RF Video Central Office WDM Video Downstream: 550 nm GPON Downstream: 490 nm XGPON Downstream: 577 nm :N G-ONT G-ONT GPON Upstream: 30 nm XGPON Upstream: 70 nm XG-ONT G-ONT XG-ONT Fig.. GPON and XG-PON coexistence example, assuming every ONU has a WDW filter. III. APD IMPACT ON THE SYSTEM PERFORMANCE The optical power, p i, at the receiver is related with the optical current, I as p i = I p /(R λ M where R λ corresponds to the unity gain responsivity and M to the average avalanche gain. The relation between power of bit 0, p i,0, and the power of bit, p i,, is the extinction ratio r = p i,0 /p i,. ITU-T established that the maximum value of the extinction ratio that can be used is r = 0.5 [3]. A. Noise characterization There are two main noise mechanisms that lead to fluctuations in the current regardless of the incident optical signal having a constant power, the shot noise and the thermal noise. The variance of the thermal noise σ c is given by σ c = (4k B T/R L F n B e,n ( where F n represents the factor by which thermal noise is enhanced by the various resistors used in pre and main amplifiers, B e,n is the effective noise bandwidth, the bandwidth of noise in hertz over which the noise is considered, T is the temperature, and k B is the Boltzmann constant. This noise is exactly the same in both PIN and APD, it does not depend on the photo-diode type since it originates in the electrical components of the receiver. The shot noise variance σ s is defined by [4] σ s = qm F A (M(R λ p i + I d B e,n. ( where I d is the dark current. In expression F A (M is the excess noise factor of the APD and is given by F A (M = k A M + ( k A ( /M (3 where k A represents the ionization coefficient ratio for the APD, which in general increases with M and is in the range 0 < k A <. This value should be as small as possible in order to achieve the best performance from an APD [5]. In the case of an PIN receiver, M = which makes F A (M =. B. APD receiver sensitivity The bit-error ratio (BER is the performance criterion for digital systems. BER is defined as the probability of incorrect identification of a received bit. Hence, the receiver sensitivity is then defined as the minimum average received power required by the receiver to operate at a certain BER. BER is given by BER = ( Q erfc exp( Q / Q, Q > 3 (4 π where the parameter Q is the quality parameter given by [6] Q = I I 0 σ + σ 0 (5 where I and I 0 are the average currents, and σ and σ 0 are the square root of variances of bits and 0, respectively. In the denominator part of Eq. 5, there are the square root of variances of bits 0 and, which are defined by (σ σ 0 = s,0 + σ c (6 σ = (σ s, + σ c. (7 From Eqs. 6 and 7, one can conclude that the square root variance of a given bit is composed by the thermal noise contribution and the related bit shot noise contribution. Using expression, the shot noise variance for bits 0 and can be expressed in terms of their average powers, p i,0 and p i,, respectively, as follows σ s,0, = qm F A (M (R λ p i,0, + I d B e,n. (8 where average power p i,0 is then defined as and p i, is given by p i,0 = p i, = p ir ( + r, (9 p i ( + r. (0

3 Sensitivity (dbm MS.C. DISSERTATION ON ELECTRICAL AND COMPUTER ENGINEERING, JULY 05 3 Therefore, the sensitivity, p i, for an APD with non-null extinction ratio is given by ( Q(r + p i = MR λ (r QqF A (MMB e,n (r + + [ (qf A (MM B e,n I d +.σ c (r + r(qqf A (MMB e,n ]. ( The sensitivity expression can be applied to the PIN receiver by considering the PIN as an APD with unitary gain (M =. Also, if one considers the simple case of a null extinction ratio (r = 0, and also neglect the contribution from the dark current (I d = 0, the well known expression presented in [6] is obtained p i = Q ( qf A (MQB e,n + σ c. ( R λ M A comparison of the APD with the PIN receiver can be seen in Fig.. There can be an improvement of several dbs by using an APD. Sensitivity improvement (db X: Y: 7.77 X: 0 Y: 7.40 X: 3 Y: Ge InGaAs Avalanche Gain (M Fig. 3. Sensitivity variation with the avalanche gain for Ge and InGaAs APDs. when APDs with higher values of responsivity are employed. Also, the sensitivity increases with the decrease of the effective noise bandwidth. The excess noise factor F A (M is also multiplying terms that appears throughout the sensitivity expression. The sensitivity decreases with the increase of the excess noise factor F A (M, which directly proportional to the ionization rate k A. Lastly, in Fig. 4 is shown the influence of the extinction ratio r in the sensitivity. The extinction ratio Fig. 4 is in db (R = 0 log 0 (/r. Values exceeding 30 db (r = /000 have no physical meaning, so it was the limit value in Fig M = 5 M = 0 M = 0 M = 40 r = 0.05 r = 0. r = Avalanche Gain (M Sensitivity (dbm Fig.. Sensitivity improvement by using APD instead of a PIN diode for three values of extinction ratio for k A = C. Analysis of sensitivity variation The expression for the sensitivity depends on many parameters. Thus, it becomes essential a better understanding of how each one of those parameters influences the sensitivity level. A set of typical values for all variable parameters, for the three most used materials in APD receivers, are shown in Table I. The considered materials were chosen for comparing purposes. Fig. 3 shows the dependence of the sensitivity on the APD gain M. It is possible to see that there is an optimum gain M for which the maximum sensitivity is achieved. TABLE I SET OF TYPICAL VALUES FOR APD RECEIVERS Material Q r R λ [A/W] B e,n [GHz] σ c [A ] I d [na] k A Ge InGaAs From this point forward, the following analysis is based on the InGaAs APD parameters, since it would be a suitable choice for a XG-PON APD material. By analysing, the performance of a system using APD optical receivers is better R (db Fig. 4. Sensitivity dependence on the extinction ratio value for k A = 0.45 and B e,n = 8 GHz. The effect of the extinction ratio increase (or decrease in the case of r is more perceptible for lower levels rather than with higher levels of R. The phenomenon present in the plot, with the avalanche gain M = 5 and M = 0 leading to similar results, is explained by the optimum gain, which is near M = 0. Thus, the performance increases till that optimum avalanche gain and decreases afterwards. D. Optimum avalanche gain The avalanche gain is important since, in the case of the excess noise factor, the avalanche gain has a direct influence in the level of noise introduced on the system. From Eq. an approximated expression for the optimum avalanche gain was achieved and it is given by M Optimum = [ a σc a (r (3 + a(k A (r + r B e,n Qq]

4 4 MS.C. DISSERTATION ON ELECTRICAL AND COMPUTER ENGINEERING, JULY 05 where a = B e,n Qqk A (r +. Expression 3 is merely an analytical re-engineered expression based on an approximation expression for the sensitivity given by Q(r + [ p i,apx = MR λ (r Q(r + qf A (MMB e,n + ] (4 σc (r. Thus, Eq. 3 can be compared with Eq., within the valid extinction ratio range (i.e. 0 r 0.5, to assess its validity. In Figs. 5 and 6, is shown the plot of expressions 4 and for the extinction ratio values of r = 0.05 and r = 0.5, respectively. 3 4 wavelengths incurs in amplitude and phase noise, simultaneously [8]. A SLM laser, is normally described by one main Fig. 7. Random power spectrum at different time instants t, t, for illustrating the partition noise. mode and a side-mode. The relation between the main-mode and the most dominant side-mode is described by the sidemode suppression ratio (SMSR, defined as SMSR = P mm P sm (5 dbm X: 3 Y: 9.3 p i p i,apx where P mm is the average power of the main mode and P sm is the average power of the most dominant side mode. SMSR should be above 30 db for a good SLM laser [6], which is the minimum side-mode suppression ratio recommended by ITU-T [9]. 30 X: 3 Y: Avalanche Gain (M Fig. 5. Sensitivity versus sensitivity approximation, with extinction ratio r = A. MPN in MLM lasers The MPN effect in MLM lasers can be estimated by adding a noise term to the total noise at the receiver. This additional noise term is added to the receiver noise so that Q is determined by [0] Q = [( + r r ] σ0 + σ + MR λ P [ σmp ] N (6 dbm X: Y: 8.07 X: Y: Avalanche Gain (M p i p i,apx where σ MP N is the square root of the variance of MPN, σ 0 and σ are the square root of the variances of bits 0 and, respectively, P is the average received signal power at the receiver, r is the extinction ratio, M is the average avalanche gain of the receiver and R λ is the responsivity. The standard deviation of MPN, σ MP N, is then given by [0] σ MP N = k MP N [ exp( β ] (7 Fig. 6. Sensitivity versus sensitivity approximation, with extinction ratio r = 0.5. In Figs. 5 and 6 it can be seen that near the optimum gain region, the approximate expression gives a value with an error of less than 0.5 db. Therefore, it can be concluded that expression 4 is a good approximation. IV. MPN IMPACT ON THE SYSTEM PERFORMANCE The MPN is related with the laser source. The lasing modes compete at slightly different wavelengths leading to fluctuation in the relative portion of the modal powers, though the total output power remains essentially constant [8]. The optical power fluctuation at different longitudinal-mode wavelengths interacting with the chromatic dispersion of the fibre results in a random profile of the output pulse intensity. The random nature of source power distribution among longitudinal-mode where k MP N is the mode-partition coefficient and β is a dimensionless parameter given by β = πbd λ Lσ (8 where B is the bitrate, D λ is the dispersion parameter, L is the distance and σ is the laser r.m.s. spectral width. The power penalty induced by MPN is related to the received power that is necessary to maintain a constant SNR seventh. Let us consider Q n given by Q n = Q Q σ MP N (9 where Q is determined using expression 4. Then, expression is the base used to derive the sensitivity expression. Thus, the power in the presence of the MPN, P n, can be obtained

5 MS.C. DISSERTATION ON ELECTRICAL AND COMPUTER ENGINEERING, JULY 05 5 by using expression using Q n instead of Q. Moreover, the power in absence of MPN, P 0, can be obtained using using Q. Therefore the power penalty (in decibels is given by α MP N = 0 log 0 ( Pn P 0. (0 The MPN has a critical importance when designing an optical transmission system. For example, if the desired BER is 0 which corresponds approximately to Q = 7, the maximum value of σ MP N for the power penalty to be under db is σ MP N B. MPN in SLM lasers There is a major difference between multi-mode and (nearly single mode lasers, which is the statistics that characterize the mode-partition fluctuations. In MLM lasers, the side modes are typically above threshold and, therefore, are well described by a Gaussian probability density function whereas, in a SLM laser, side modes are typically below threshold. SLM side modes follow an exponential distribution given by [6] p(p sm = P sm exp [ P ] sm, ( P sm where P sm is the average value of the power of the side mode P sm. For better understanding the effect of side mode fluctuations on system performance, an ideal receiver is considered (i.e. no dark current nor thermal noise and 00% quantum efficiency [6]. Let τ = DL λ be the relative delay between the main mode and the side mode, where λ is the mode spacing. Let us assume that τ > /B which implies that BLD λ >. This is the same to say that it is assumed the relative delay is long enough that the side mode does not reach the bit slot in time. Let the decision threshold be at Pmm /, being P mm the average power of the main mode. There will be an error if the transmitted bit is 0 and the receiver detects a value above the threshold or if the transmitted bit is and the receiver detects a power below the threshold. Also, it is assumed that the total power remains constant, i.e. the two modes are anti-correlated so that when main mode power drops below threshold, side mode power will exceed it. Since 0 and have can be considered equally probable, BER is defined by [] BER = + P mm/ ( p(p sm dp sm = exp SMSR ( where SMSR corresponds to the side-mode suppression ratio defined in equation 5. If a BER of the 0 is considered, leads to a minimum SMSR 7.4 db. The main mode current i 0, (t, for bit 0 or, can be considered a random variable with an associated statistical distribution. If the two mode laser is considered, the total current after the optical receiver is then given by i T0, (t = i 0, (t + i s (t (3 where i s (t corresponds to the current after the optical receiver resulting from the side-mode. It is known [3] that the total laser current i T0, (t follows a Gaussian distribution. However, we may define the total current as i T0, (t = ĪT 0, + δi T0, (t, where only δi T0, (t is a random variable and corresponds to the fluctuations of the total current. The variable ĪT 0, corresponds to the total average current for bit 0 or. Hence, the probability density function for δi T0, (t is given by p δt0, (x = σ T0, π exp ( x σt 0,. (4 Considering the probability of bit 0 and to be equal, the BER is given by BER = [ P (δit0 (t + i G0 (t > I D ĪT 0 + P (δi T (t + i G (t < I D ĪT ] (5 where i G0, (t corresponds to the Gaussian noise current for bits 0 or, and I D the decision threshold. The resulting BER expression is given by BER = erfc ( Q (6 where Q is given by Q = and I D is given by σt + σ (Ī0 + Īm SMSR I D = Ī Ī0 σ T0 + σ 0 + σ T + σ + σ T0 + σ 0 + σ T + σ σ T 0 + σ 0 (Ī + Īm SMSR (7 (8 The variance of bit 0, σ 0, and the variance of bit, σ, are given by expressions 6 and 7, respectively. The variance of the total current after the optical receiver for bit i, σ T i, is given by σ T i = σ mm,i σ sm,i (9 where the variance of the laser intensity noise σmm i is given by σmm i = R λ MP i r l (30 where P i is the power of bit i and the parameter r l is a measure of the noise level of the incident optical signal. The parameter r l is considered as the inverse of the SNR of the light emitted by the transmitter [6]. Typically, the transmitter SNR is better than 0 db and r l < 0.0 [6]. The side-mode noise variance is given by ( σsm Rλ MP i i = (3 SMSR Since bit 0 and bit are equally likely to occur, the average current of the main mode Īm may be defined as Ī m = Ī0 + Ī. (3.

6 Power penalty (db Power penalty (db 6 MS.C. DISSERTATION ON ELECTRICAL AND COMPUTER ENGINEERING, JULY 05 C. Penalty for using MLM lasers When a MLM laser is employed, the power penalty may be calculated using equations 7 and 0. A common used value for the mode-partition coefficient is k MP N = 0.5 [0] and in [] it is suggested that k MP N is within the range. So, for the purpose of the analysis, several values of k MP N were considered. In Fig. 8, it can be seen how the penalty is influenced by the dispersion parameter β, for a value of Q = 7. For a BER level of 0 (Q = 7, the power penalty k MPN = parameter Fig. 8. Power penalty when using a MLM laser as a function of β parameter, for various values of k MP N for Q = 7. of the MPN in a MLM laser increases greatly with β. For high values of k MP N, the power penalty may become infinite (i.e. the defined BER of 0 is not achievable for β values above a certain point. For experimental values of k MP N, in the previous specified range of , the power penalty becomes infinite for values of β above 0.5. Therefore, considering the XG-PON bitrate of 0 Gbps and using expression 8, the distance is limited to L < /(0 0 9 πdσ. Considering D = 7 ps/(nm/km and σ = nm [], the distance is limited to L < 0.94 km. In Fig. 9,it is represented the power penalty for a BER level of 0 3 (Q = 3. It can be seen that, for a higher BER, the mode-partition coefficient k MP N has less influence on the power penalty. When designing an optical system, the parameter k MPN =.0 Fig. 9. Power penalty when using a MLM laser as a function of β parameter, for various values of k MP N for Q = 3. β parameter is not the used metric. In fact, light-wave systems are usually designed such that [] BDLσ < 0., which as per Eq. 8 means that β/π < 0. β < 0.π In this case, the power penalty can be as low as a negligible 0.5 db or infinite, depending on the value of k MP N. To render the penalty to negligible values ( 0.5 db, independently of the k MP N value, one must design a system where β 0.336, approximately. Such imposition means that BDLσ < 0.. Therefore, if the usual measure of [] BDLσ < 0. is cut down by one half, the MLM MPN power penalty is not noticeable. The σ and D are fixed parameters, intrinsic to the laser and fibre used, respectively. Thus, there is a trade-off between transmission speed B and covered distance L. D. Penalty for using SLM lasers In SLM lasers, the evaluation of the power penalty induced by MPN is dependent on the value of Q, defined in expression 7. Since BER is given by Eq. 6, the BER value is set by the value of Q. Expression 7 can be analytically manipulated to be expressed in terms of power value. The power penalty is then given by ( P0 P MP N = 0 log 0 (33 P MP N where P MP N is the power in the presence of MPN and P 0 is the power in the absence of MPN (i.e. SMSR = +. Unless otherwise stated, the parameter values on Table VI were used to obtain the numerical results. Considering a BER target of 0 (Q = 7 and an extinction ratio r = 0., it is shown in Fig. 0 the relation between the side-mode suppression ratio and the increase of the power at the receiver, for the values of M = and M = 0. Increase of received power (db M = 0 M = BER = SMSR (db Fig. 0. Relation between the increased power at the receiver and the sidemode suppression ratio, considering a reference BER of 0 and r = 0., for M = and M = 0. If we consider a greater avalanche gain (M = 0, as shown in Fig. 0, it can be seen that the increase in the avalanche gain decreases the needed power. However, the influence of the avalanche gain is low and thus, the decrease in the power penalty is low. Also, it can be seen that, for values of modesuppression ratio SMSR 30 db, which is the minimum required by ITU-T, the power penalty is negligible. In Fig., it is shown the power penalty using the extinction ratio values of r = 0.0, r = 0. and r = 0.5 (maximum

7 MS.C. DISSERTATION ON ELECTRICAL AND COMPUTER ENGINEERING, JULY 05 7 allowed by ITU-T, for an avalanche gain of M = 0. The power penalty decreases with the extinction ratio decrease. However, the power penalty variation is low and is negligible for values of SMSR used in practice (SMSR 30 db. Increase of received power (db r = 0.0 r = 0. r = 0.5 BER = SMSR (db Fig.. Relation between the increased power at the receiver and the sidemode suppression ratio, considering a reference BER of 0 and M = 0, for r = 0.0, r = 0. and r = 0.5. V. ASSESSMENT OF XG-PON REACH IMPROVEMENT A. Link budget of the XG-PON system The operation of a XG-PON system requires that the emitted power at the OLT (or ONU reaches the ONU (or OLT at a level that the optical receiver can correctly identify the bits sent, with a desired BER. The assessment of XG-PON reach is accomplished considering a BER target of 0, when no Forward Error Correction (FEC is employed, and a BER of 0 3 for FEC. It was considered an extinction ratio of r = 0.. In the XG-PON system, the typical optical budget follows the equation given by [6] P r P e A link > P i (34 where P r corresponds to the power at the receiver input, Pe is the power emitted at the transmitter output and A link is the total attenuation. Typically, it is considered a system margin M s given by [6] M s db = P e A link P i P i (D λ L max (35 where P i (D λ L max is the maximum path penalty due to dispersion in the XG-PON system transmission, and is set to a maximum of db. The system margin M s should cover a safety margin (due to unexpected losses M saf = 3 db and other margins due to ageing and variations on the environment and temperature. A minimum of M s = 6 db is required for the system margin [6]. The path penalty due to the dispersion is given by P i (D λ L db = P i (D λ L L db + P i (D λ L M db + P i (D λ L MP N db (36 where P i (D λ L M db is the power penalty associated with the modulated bandwidth, P i (D λ L L db is the power penalty associated with the source linewidth and P i (D λ L MP db N is the power penalty associated with the MPN. The attenuation due to the transmission and passive components, A link, is given by A link = N c A c + N s A s + α f L + N sp A sp (37 where N c is the number of connectors used and A c is the connector attenuation, N s is the number of splices and A s is the splices attenuation, N sp is the number of splitters and A sp is the splitter attenuation, α is the fibre attenuation per km and L is the distance in km. Thus, the optical power budget is then given by P e P i + M s + A link + P i (D λ L max (38 By using expressions 38 and 37, it is obtained the maximum distance L that can be accomplished. The maximum distance is given by L P e P i M s A passive P i (D λ L max α f (39 where A passive is given by A passive = N c A c + N s A s + N sp A sp. (40 Unless otherwise stated, the parameters used for the XG-PON system are shown on Tables II, III, IV and V. TABLE II XG-PON SYSTEM OPERATION WAVELENGTHS. Down λ [nm] Up λ [nm] TABLE III TYPICAL PARAMETERS OF XG-PON SYSTEM USED FOR OBTAINING NUMERICAL RESULTS. k MP N A s [db] A c [db] M s [db] r BER / 0 TABLE IV TYPICAL G.65 FIBRE PARAMETERS USED FOR OBTAINING NUMERICAL RESULTS [3]. λ 0 [nm] D λ [ps/(nm.km] α [db/km] TABLE V PON SPLIT RATIOS AND CORRESPONDING LOSSES [4]. Split Ratio A s [db] : 3.63 :4 7. :8 0.7 : : : :8 4.9

8 8 MS.C. DISSERTATION ON ELECTRICAL AND COMPUTER ENGINEERING, JULY 05 The APD receiver is compared against a PIN receiver in order to assess the reach extension when using an APD. Unless otherwise stated, the parameters used for the receiver are shown in Table VI. The receiver sensitivity in dbm P i = 0 log 0 ( pi mw is computed based on expression 7. The PIN receiver is considered by setting the avalanche gain M =. The downstream FEC code is RS(48, 6 whereas the upstream FEC code is RS(48, 3 []. The effective noise bandwidth, B e,n, increases with the FEC code ratio. Taking into account the values shown in Table VI, the TABLE VI RECEIVER PARAMETERS USED FOR OBTAINING NUMERICAL RESULTS. R λ [A/W] B e,n [GHz] σ c [A ] I d [na] k A M sensitivity values for a PIN receiver and an APD receiver, with FEC and without FEC, were computed and are shown on Table VII. Let us assume that the transmitting powers for the OLT TABLE VII SENSITIVITIES USED FOR OBTAINING NUMERICAL RESULTS. FEC coding Q BER P i APD [dbm] P i PIN [dbm] RS(48, RS(48, Without FEC TABLE VIII EMMITING POWERS FOR THE OPTICAL SOURCES OF OLT AND ONU USING MLM AND SLM LASERS. Tx OLT [dbm] Tx ONU [dbm] 0.5 and ONU lasers are those in Table VIII. The XG-PON system can be constituted by only one splitter or several, if cascading is applied (e.g. using a splitter of : and one of :4 instead of applying a splitter of :8. For simplicity, only one splitter is considered. Thus, let us consider two connectors, one at the OLT and one at ONU, two splices in the splitter and a splice every.36 km. The XG-PON has two maximum distances 0 km and 40 km [], the following sections use those maximum distances. B. MLM lasers Downstream: Consider the use of a MLM laser as an optical source for the OLT. In the downstream direction the chromatic dispersion parameter is D λ = 8.4 ps/(nm.km, for the operating wavelength λ 0 = 577 nm, and the bitrate is 0 Gbps. Assuming a spectral half-width of σ = nm [] for the MLM laser and using expression 0, the power penalty for L = 0 km due to MPN is P i (D λ L MP N = for both the scenario without FEC and the scenario with FEC. Therefore, the MLM laser cannot be used as the optical source at the OLT. Upstream: Consider the use of a MLM laser as an optical source for the ONT. In the upstream direction the chromatic dispersion parameter is D λ = 3.68 ps/(nm.km, for the operating wavelength λ 0 = 70 nm. The bitrate is 0 Gbps if the XG-PON is designed to have a symmetrical bitrate or.5 Gbps if the XG-PON has an asymmetrical bitrate. Assuming a spectral half-width of σ = nm [] for the MLM laser, a bitrate of 0 Gbps and using expression 0, the power penalty for L = 0 km due to MPN is P i (D λ L MP N = for both the scenario without FEC and the scenario with FEC. For the case of where a.5 Gbps bitrate is used, the power penalty for L = 0 km due to MPN is P i (D λ L MP N = for the scenario without FEC and P i (D λ L MP N =.9 < db for the scenario with FEC. Let us consider the case where FEC is employed and evaluate the penalty associated with the source linewidth P i (D λ L L db and the penalty associated with the modulated signal bandwidth P i (D λ L M db. The power penalty P i (D λ L M db is given by [6] P i (D λ L M ( db= 5 log 0 ( 8αc β LB + (8β LB where α c is the chirp parameter and β is given by [6] (4 β = D λλ 0 πc. (4 In the upstream, if conventional bulk lasers are considered, the chirp parameter is α c 6 [6]. Therefore, the power penalty associated with the modulated signal bandwidth is P i (D λ L M db = 0.08 db. The power penalty P i (D λ L L db is given by [6] P i (D λ L L db= 5 log 0 [ (4BDλ Lσ λ,l ] (43 where σ λ,l is the r.m.s. spectral width and it is given by [6] σ λ,l = λ L (44.35 where λ L is the laser linewidth. For MLM lasers, the linewidth is typically λ L 5 nm [6]. Therefore, assuming λ L = nm, the power penalty associated with the source linewidth P i (D λ L L db =.08 db. Consequently, the total dispersion penalty P i (D λ L db db. Therefore, the MLM laser cannot be used in either the OLT or the ONU. C. SLM lasers Downstream: In the downstream direction, the power penalty is negligible for the considered mode-suppression ratio of SMSR = 30 db. The penalties consider the bitrate of 0 Gbps and a distance of L = 0 km. Using expression 7 and expression 33, the power penalty is P i (D λ L MP db N db for the APD. The obtained values for the MPN penalty are similar whether FEC is employed or not. Let us evaluate the penalty associated with the source linewidth P i (D λ L L db and the penalty associated with the modulated signal bandwidth db for the PIN and P i (D λ L MP N db P i (D λ L M db. The typical linewidth of the SLM laser (in Hz units is υ 00 MHz [6]. If we consider υ = 50 MHz, λ = λ 0 υ c m. Thus,

9 MS.C. DISSERTATION ON ELECTRICAL AND COMPUTER ENGINEERING, JULY 05 9 the power penalty associated with the source linewidth is P i (D λ L L db = 0 db. In the XG-PON system, the lasers at the OLT are externally modulated lasers (EML [6], which means that the chirp parameter is α c 0 [6]. Thus, the power penalty associated with the modulated signal bandwidth is P i (D λ L M db = 0.3 db. Therefore, the total dispersion penalty P i (D λ L db db. If we consider the distance of L = 40 km, P i (D λ L M db =.03 db and P i(d λ L L db = 0 db. Therefore, the total dispersion penalty P i (D λ L db db. Thus, the dispersion does not limit the link of the XG- PON in the downstream direction. It is assumed the value P i (D λ L max = db, the maximum distance imposed by the optical budget is obtained from expression 39. Using expression 35 for obtaining the system margin, it is shown in Table IX the distances imposed by the optical power budget for both the PIN and the APD receiver for every splitting ratio and system margin for a distance of L = 0 km as a function of the splitter ratio, when FEC is not employed. TABLE IX MAXIMUM DISTANCE IMPOSED BY THE POWER BUDGET AND SYSTEM MARGIN FOR L = 0 KM, FOR BOTH PIN AND APD RECEIVERS AS FUNCTION OF THE SPLITTER RATIO WITHOUT FEC, IN THE DOWNSTREAM DIRECTION. PIN APD Splitter ratio L B [km] M s [db] L B [km] M s [db] : : : : : : : When FEC is not employed, the PIN receiver is limited by the optical budget to a splitting ratio of :6, whereas the APD receiver supports splitting ratios up to :64. The commonly used splitting ratios are between :8 and :64 [6]. Thus, the APD covers the use of all common splitting ratios for L = 0 km and up to :3 for L = 40 km. In the case where FEC is employed, the results are in Table X. TABLE X MAXIMUM DISTANCE IMPOSED BY THE POWER BUDGET AND SYSTEM MARGIN FOR L = 0 KM, FOR BOTH PIN AND APD RECEIVERS AS FUNCTION OF THE SPLITTER RATIO USING FEC, IN THE DOWNSTREAM DIRECTION. PIN APD Splitter ratio L B [km] M s [db] L B [km] M s [db] : : : : : : : The use of a PIN at the ONU allows a splitting ratio up to :3, and is limited by the optical budget for higher splitting ratios. On the other hand, the APD receiver supports a splitting ratio up to :8. If we consider a distance of L = 40 km, the PIN receiver only supports a splitting ratio up to :8 whereas the APD receiver supports up to :64 splitting ratio. Again, the use of an APD covers the use of all common splitting ratios at both L = 0 km and L = 40 km. Upstream: In the upstream scenario, for both.5 Gbps and 0 Gbps bitrate, the power penalty due to MPN is the same used in section V-C. Let us evaluate the penalty associated with the source linewidth P i (D λ L L db and the penalty associated with the modulated signal bandwidth P i (D λ L M db. If we consider υ = 50 MHz, λ = λ 0 υ c m. Consider the distance L = 0 km, the power penalty associated with the source linewidth is P i (D λ L L db = 0 db for both 0 Gbps and.5 Gbps bitrate. In the XG-PON system, the lasers at the ONU are directly modulated lasers (DML [6], which means that the chirp parameter is α c 6 [6]. The power penalty associated with the modulated signal bandwidth is P i (D λ L M db = 0. db for.5 Gbps and P i (D λ L M db =.67 db for 0 Gbps. Therefore, the total dispersion penalty P i (D λ L db db. If we consider the distance of L = 40 km, P i (D λ L M db =.87 db and P i (D λ L L db = 0 db for a bitrate of 0 Gbps. The XG-PON system is dispersion limited for the distance L = 40 km at a 0 Gbps bitrate. However, for a.5 Gbps bitrate, P i (D λ L M db = 0.4 db and P i(d λ L L db = 0 db. Therefore, the total dispersion penalty P i (D λ L db db for an upstream bitrate of.5 Gbps. Consider the case where FEC is not employed, using expression 35 for obtaining the system margin, M s, it is shown in Table XI the distances imposed by the optical power budget for both the PIN and the APD receiver for every splitting ratio and system margin for a distance of L = 0 km as a function of the splitter ratio. When the PIN receiver is employed at TABLE XI MAXIMUM DISTANCE IMPOSED BY THE POWER BUDGET AND SYSTEM MARGIN FOR L = 0 KM, FOR BOTH PIN AND APD RECEIVERS AS FUNCTION OF THE SPLITTER RATIO WITHOUT FEC, IN THE DOWNSTREAM DIRECTION. PIN APD Splitter ratio L B [km] M s [db] L B [km] M s [db] : : : : : : : the OLT, the distance L = 0 km is not achievable. Under the conditions tested and using an APD receiver, the XG-PON system is limited to a splitting ratio of :4 when FEC is not employed for a distance of L = 0 km. The results for the

10 0 MS.C. DISSERTATION ON ELECTRICAL AND COMPUTER ENGINEERING, JULY 05 case when FEC is employed are shown in Table XII. The use TABLE XII MAXIMUM DISTANCE IMPOSED BY THE POWER BUDGET AND SYSTEM MARGIN FOR L = 0 KM, FOR BOTH PIN AND APD RECEIVERS AS FUNCTION OF THE SPLITTER RATIO USING FEC, IN THE UPSTREAM DIRECTION. PIN APD Splitter ratio L B [km] M s [db] L B [km] M s [db] : : : : : : : of a PIN at the OLT limits the XG-PON system to : splitting ratio. On the other hand, the APD receiver supports a splitting ratio up to :8. The distance of L = 40 km is only achievable if an APD receiver is employed and a splitting ratio of : is used. To use the common splitting ratios FEC is mandatory, the OLT receiver needs to be an APD and, in the case of the splitting ratios :3 and :64, the emitted power at the ONU has to be increased. Under the conditions studied, the XG-PON system is limited by the optical budget in the upstream direction, which leads to a maximum distance of L = 0 km, for the commonly used splitting ratios. Also, the splitting ratio of :8 is not achievable. VI. CONCLUSIONS In this dissertation, the assessment of the reach improvement by using APD receivers in the presence of MPN was performed. Both MLM and SLM laser sources were considered. The fundamentals of APD were given, with focus on the factors that influence the APD performance, the APD sensitivity was analyzed and a model for the APD sensitivity, for any given extinction ratio, was proposed. The ionization rate and the excess noise factor have great impact on the APD performance. The gain of an APD over a PIN receiver was presented numerically. Also, an expression for the optimum avalanche gain was proposed. In addition, an expression for the power penalty due to MPN in MLM lasers was presented and it was found that, at the XG-PON bitrate of 0 Gbps, there is a great limitation on the achievable distance. Also, a model for BER related to SLM lasers when MPN is present, based on the Gaussian approximation for the total current after the receiver [3] and the model that describes the SLM laser of [3], was proposed and analyzed. The avalanche gain of the receiver and the extinction ratio have low influence on the power penalty due to MPN, and the power penalty is negligible for values of SMSR above 30 db (minimum imposed by ITU- T. Moreover, it was presented an assessment of the effects of the MPN and the enhancement in the XG-PON system by using an APD receiver as opposed to a PIN receiver. The MLM lasers cannot be employed due to the high dispersion suffered by the signal emitted by an MLM laser. Furthermore, it was analysed the use of the SLM lasers in the downstream direction and in the upstream direction. In the downstream direction, the use of an APD instead of a PIN at the ONU allows an improvement of reach and higher splitting ratio. Moreover, the use of an APD receiver allows the use of the common splitting ratios, between :8 and :64, for distances of L = 0 km and L = 40 km. However, under the conditions studied, to achieve the common splitting ratios in the upstream direction, the use of an APD receiver at the OLT is mandatory, FEC needs to be employed and the emitting power of the ONU transmitter has to be increased. Moreover, under the conditions studied, the splitting ratio of :8 was not achieved due to limitations in the upstream direction. REFERENCES [] 0-Gigabit-capable passive optical network (XG-PON systems: Definitions, abbreviations and acronyms, ITU-T recommendation G.987, 0. [] 0-Gigabit-capable passive optical network (XG-PON systems: General Requirements, ITU-T recommendation G.987., 00. [3] Optical interfaces for equipments and systems relating to the synchronous digital hierarchy, ITU-T recommendation G.957, 006. [4] G. Keiser, Photodetectors, Optical Fiber Communications, 3rd ed., Singapore: McGraw-Hill, pp , 009. [5] Photonics Online, Tutorial : Avalanche Photodiodes Theory And Applications. Available (005, Dec. 5. [6] G. Agrawal, Fiber-Optic Communication Systems, 3rd ed., John Wiley & Sons, Inc., 00. [7] G. Agrawal, P. Anthony, and T. Shen, Dispersion penalty for.3- m lightwave systems with multimode semiconductor lasers, IEEE/OSA Journal of Lightwave Technology, vol. 6, no. 5, pp , 988. [8] Transmission systems and media, digital systems and networks, ITU-T recommendation G.959, 0. [9] K. Pettermann, Laser Diode Modulation and Noise, st ed., Springer, 988. [0] W. T. Tsang, Semiconductors and Semimetals. Volume : Lightwave Communications Technology : Part C, Semiconductor Injection Lasers, II Light-Emitting Diodes, st ed., Academic Press, Inc., 985. [] S. Bottacchi, Noise and Signal Interference in Optical Fiber Transmission Systems: An Optimum Design Approach, st ed., John Wiley & Sons, Inc., 008. [] M. Mori, Y. Ohkuma, and N. Yamaguchi, Measurement of mode partition characteristics of laser diodes modulated by M-sequences, IEEE/OSA Journal of Lightwave Technology, vol. 7, no. 7, pp. 5-9, 989. [3] Prysmian, G.65 Series. Available markets/markets/fibre/products/single-mode/g65-series/ (05, Jul. 0. [4] M. Hajduczenia, S. Pato, Channel insertion loss for x64 and x8 split EPONs, IEEE80.3 Plenary Meeting, Dallas, 006. [5] Thorlabs, Single-Frequency Lasers. Available newgrouppage9.cfm?objectgroup id=4934 (05, Jul. 0. [6] F. Bourgart, Optical Access Transmission: XG-PON system aspects, FTTH Conference 00, 00.