We investigate on the uncooled RSOA driving parameters in WDM reflective PONs, focusing on an upstream path at 1.25 Gbit/s using pure binary modulation. We show how the optimal values change using direct-detection or self-coherent receivers. In particular, for the latter, the driving point optimization allows a gain of more than 3 dB in terms of ODN loss compared to a standard On-Off Keying, generating a quasi-PSK modulation. We also address operating temperature and wavelength dependence issues.
©2012 Optical Society of America
Several previous works, such as [1–4], have clearly shown the potential advantages of WDM reflective Passive Optical Network (PON) using Reflective Semiconductor Optical Amplifier (RSOA) at the Optical Network Unit (ONU) and self-coherent detection at the Optical Line Terminal (OLT). Focusing on this scenario and in particular on the upstream path, we present in this paper our recent results on the RSOA parameters optimization. In particular, sticking with pure binary modulation (i.e. two electrical levels), we have focused on optimizing an uncooled RSOA electrical parameters using direct-detection (DD) or self-coherent (SC) receiver, while also changing temperature and wavelength. While it is well known that for a DD receiver, the RSOA optimal driving parameters are those that generate an OOK (On-Off Keying) signal with maximal extinction ratio (ER), we will show that, when using a coherent receiver, they approximate a (quasi) 2-PSK (Phase Shift Keying) modulation and this setup gives at least 3 dB of gain in terms of Optical Distribution Network (ODN) loss compared to OOK. This key idea, though already introduced in several works by Prof. Chung’s group (such as [1–3], and their following papers), was never, to our best knowledge, analyzed in details by spanning all possible RSOA driving values. Thus, our paper greatly extends the already existing literature by presenting a careful parameters optimization (bias current and modulation amplitude), also introducing temperature and wavelength dependence. The used RSOA is a low-cost commercial TO-can packaged device, without any temperature control. The paper is organized as follows: we start in Sect. 2 by describing the experimental setup, then we present in Sect. 3 the RSOA optimization in the different configurations. Finally, in Sect. 4, we draw some conclusions.
2. Experimental setup
We worked on the setup shown in Fig. 1 , where we have emulated upstream transmission in a reflective ONU architecture. A continuous wave (CW) seed signal is generated by an external cavity laser at the OLT side, then sent downstream (by a circulator, output power PFIBER) to a variable optical attenuator (VOA) used to emulate different optical distribution network (ODN) losses (LODN). The optical signals is modulated by means of an uncooled RSOA with a pure binary 1.25 Gbps input signal, generated by a pattern generator (PRBS = 215-1, a value selected in order to satisfy the constraint set by the used real-time oscilloscope in our off-line processing experiments, see later). In our measurements, we focused on optimizing the two available RSOA driving parameters, given by the peak-to-peak modulating signal amplitude Vpp and the biasing current Ib, while also changing the device temperature. The upstream signal, after going again through the VOA and the circulator, reaches two different receivers:
- • a commercial coherent receiver (by Neophotonics) using the same local oscillator laser (LO) used for the upstream CW seed;
- • an optically EDFA pre-amplified DD receiver. Even though EDFA are not used in PON receivers, we have inserted only in order to have back-to-back sensitivity comparable to the coherent receiver one (close to −52 dBm at 1.25 Gbps).
In both cases, we evaluate the resulting Bit Error Rate (BER) by off-line processing on approximately 2⋅105 bits (corresponding to the maximum memory of our real-time oscilloscope when sampling at four samples per bit). For the coherent RX, we use nowadays quite common digital signal processing (DSP) techniques (such as carrier-phase Viterbi-Viterbi estimation, LMS adaptive equalization ). For the DD case, we simply implemented by DSP a post-detection low pass filter (fourth-order Bessel filter at 75% of the bit rate) followed by amplitude decision threshold and clock optimization. The other system parameters are shown in Table 1 .
3. Experimental results
The experimental results and operating conditions are concisely presented in this Section 3, while all system comments are given in the following Section 4.
Considering that the maximum allowed LODN is used by ITU-T to determine the PON architecture “Class” , we used this parameter as the key figure of merit, by changing the RSOA parameters Vpp and Ib to evaluate the maximum LODN that gives BER≤10−3. The results, expressed in terms of LODN contour plots (@BER≤10−3) vs. Vpp and Ib, are shown in Fig. 2 for the coherent-detection receiver (a) and for the optically amplified DD receiver (b) . Both figures were measured keeping the RSOA at room temperature (around 25◦C).
In order to gain a better insight, we show in Fig. 3 the resulting scattering diagram obtained when driving the RSOA with the optimal parameters resulting from Fig. 2, i.e., those that give the maximum LODN (Ib = 75 mA, Vpp = 1.5 V for coherent detection and Ib = 75 mA, Vpp = 4 V for DD, −35 dBm optical input power). We have driven in this case the RSOA with a 500 MHz sinusoidal electrical signal. The received signals were post-processed in DSP by applying a Jones matrix polarization transformation that minimizes the Y polarization power. Due to the receiver AC coupling characteristics, the DC components is lost and both scattering diagrams have coordinate (0,0) as their average.
Other measured parameters are shown in Tab. 2 . The ER has been evaluated by means of a DC-coupled DD receiver. The degree of polarization (DOP) was measured by means of a polarization analyzer when varying the RSOA input polarization, so the minimum and the maximum DOP has been found over all possible polarization states.
To evaluate the stability of the optimal driving point, we repeated the measurements inserting the RSOA in a climatic chamber, and setting the operating temperature to 10 and 50°C. The results are shown in Fig. 4 and Fig. 5 . Finally, we report in Fig. 6 the RSOA (power) gain vs. operating temperature under static condition (DC driving current). To complete the RSOA characterization campaign, we evaluated the optical wavelength system performances dependency. The CW seed wavelength signal was swept from 1530 to 1560 nm (full C-band), while the RSOA parameters have been fixed to the best operating conditions obtained respectively for coherent and DD, showing the results in Fig. 7 .
4. Discussion on the experimental results
We observe in Fig. 2 that the optimal RSOA driving parameters are significantly different for the two receivers. In fact, we have an optimum at Vpp = 1.5 V and Ib = 75 mA when using the coherent receiver, while we have Vpp = 4 V and Ib = 75 mA for the DD receiver. Due to the relatively small ER (3.47 dB from Tab. 2) for the Vpp = 1.5 V and Ib = 75 mA modulation case, the relative scattering diagram, represented in Fig. 3(a), is similar to a 2-PSK modulation. For the other case (Vpp = 4 V and Ib = 75 mA, i.e. the optimal condition for the DD receiver) the relative scattering diagram, represented in Fig. 3(b), is a much more involved figure that, due to its higher ER (8.67 dB from Tab. 2), corresponds to an OOK modulation. The two scattering diagrams can be explained by remembering [1,8] that the RSOA has a non-negligible chirp, so that when changing its electrical input signal a combined amplitude and phase modulation is obtained. In a coherent setup, the receiver is sensitive also to phase variations, so the optimal driving condition is close to a 2-PSK modulation under the constraint of binary modulation. For a DD receiver, the phase is irrelevant, so the optimal condition is basically the one that generates an OOK signal with maximum ER. Furthermore, in the OOK case, we observe that a residual scattering diagram on Y polarization is present (see Fig. 3(b)). It is due to the large RSOA input modulation signal (4 Vpp) that introduces a spurious polarization modulation and a variable DOP (DOPMIN ≠ DOPMAX from Tab. 2). For lower input modulation signal (see Fig. 3(a)), no significant polarization modulation appears (DOPMIN ≈DOPMAX from Tab. 2).
From a system point of view, these results are very interesting and they can be commented as follows:
- • when setting the obtained optimal transmitter conditions, the RSOA ONU can be modulated by a much smaller electrical signal in the self-coherent receiver than the one required for DD receiver (1.5 V rather than 4 V in our case), allowing a reduction in ONU electrical power consumption and an increase in long-term reliability;
- • focusing on the self-coherent receiver alone, our result shows that moving from the “usual” OOK modulation applied in most RSOA experiments to the proposed “quasi-PSK” modulation, the resulting system performance is significantly improved. Looking at Fig. 2(a), the “quasi-PSK” point (1.5 V, 75 mA) gives 38 dB in terms of maximum ODN loss, while the “OOK point” (4.0 V, 75 mA) gives 34-35 dB. Thus, the advantage is more than 3 dB on ODN loss. In terms of receiver sensitivity, since the ODN loss counts twice in the upstream received power (the RSOA was working in both cases in a small signal regime, thus giving the same optical power gain in both cases), the advantage corresponds to more than 6 dB. We qualitatively interpret this result as follows. On an additive white Gaussian noise channel (well approximated in our situation by the coherent receiver), it is know that, to obtain the same target BER, 2-PSK has a 3 dB advantage in terms of average power compared to OOK, and 6 dB advantage in terms of peak power. In our system, the launched power and the RSOA gain are fixed, so that the comparison should be done in terms of peak power. The aforementioned theoretical 6 dB receiver sensitivity advantage of 2-PSK over OOK is anyway in our experiments affected by at least two different non-idealities: it is diminished by the non-perfect “quasi-PSK” condition (with 3.47 dB ER, see Table 2, rather than an ideal 0 dB ER), but at the same time it is increased by the fact that the OOK condition, corresponding to a 4 VPP driving, also generate a non-negligible spurious polarization modulation (See Fig. 3(b)), beside a non-ideal OOK ER (8.67 dB, see Table 2). These two non-idealities acts in opposite directions and have likely similar impact, thus justifying heuristically why we experimentally get 6 dB advantage in terms of receiver sensitivity.
Changing the RSOA operating temperature respectively to 10°C and 55°C, we observe that the optimal RSOA driving point (shown in Figs. 4 and 5) still remains approximately the same for every receiver. This is again an interesting result from a system point of view, since it allows to envision uncooled operation  for the RSOA (in fact, we used a TO-can packaged RSOA, that has no provision for temperature control).
While the optimal operation point remains constant, the resulting maximum allowable ODN loss changes with the temperature. Indeed, at 10°C it increases of about 1 dB (from a maximum LODN = 38 dB at room temperature to LODN = 39 dB at 10°C), while it significantly decreases of about 5 dB at 50°C. The performances variations can be explained by observing Fig. 6, which shows the RSOA optical power gain versus different operating temperatures: when moving from 10°C to 50°C the gain decreases of about 12 dB, so that the ODN loss should be reduced by half this value as shown in Fig. 4. Anyway, even at 50°C, the LODN = 33 dB value obtained with this setup is very interesting, indeed it would still be compatible with ITU Class C + which requires LODN = 32 dB (see ).
Regarding the system performances dependency versus the wavelength over the full C-band, the maximum LODN variation (to get BER = 10−3) is lower than 1 dB for the coherent receiver case, and is practically constant for the DD receiver. These results are respectively summarized in Fig. 7, fixing the RSOA parameters to the best operating conditions obtained for both the employed receivers. The same measurement is repeated at 10, 25 and 50°C. This is again an interesting result from a system point of view, since it allows driving the RSOA independently of the system wavelength.
We have shown that the use of coherent receivers at the OLT completely changes the optimal operating condition of the reflective modulator at the ONU side. We observe that for the more traditional DD receiver, the RSOA should be optimized to have high ER and low chirp. For the coherent optical receiver case instead, in order to enhance phase modulation and reduce amplitude modulation, it would be better to design the RSOA with high chirp and minimum ER. Ironically, while for the (up to now) much more common DD receiver the RSOA designers usually strive to obtain low chirp operation (see for instance ), for a coherent receiver one should try to obtain exactly the opposite, i.e. high chirp and low amplitude modulation, to obtain a quasi-PSK signal with very small electrical signal swing. We believe this is one of the interesting results of our paper, and can trigger some further component-level research. We also demonstrated that the optimal RSOA driving point remains the same when changing the RSOA temperature and this allows envisioning uncooled operation for the RSOA. We have also shown a negligible dependence on operating wavelength.
References and links
2. K. Y. Cho, Y. Takushima, and Y. C. Chung, “10-Gb/s operation of RSOA for WDM PON,” IEEE Photon. Technol. Lett. 20(18), 1533–1535 (2008). [CrossRef]
3. Y. Takushima, K. Y. Cho, and Y. C. Chung, “Design issues in RSOA-based WDM PON,” Photonics Global 2008, Singapore, C-34 – C-37.
4. B. Charbonnier, A. Lebreton, S. Straullu, V. Ferrero, A. Sanna and R. Gaudino, “Self-coherent single wavelength SC-FDMA PON uplink for NG-PON2,” OFC/NFOEC 2012, Los Angeles, (OW4B.4).
6. ITU-T Recommendation G.984.2 Amendment 2 (03/2008).
7. P. D. Townsend, A. Clarke, P. Ossieur, D. W. Smith, A. Borghesani, D. G. Moodie, I. F. Lealman, X. Z. Qiu, J. Bauwelinck, X. Yin, K. Grobe, B. T. Teipen, R. Jensen, N. Parsons and E. Kehayas, “Towards colourless coolerless components for low power optical networks,” ECOC 2011, Geneva, Switzerland (Tu.5.LeSaleve).
8. I. Papagiannakis, M. Omella, D. Klonidis, J. A. L. Villa, A. N. Birbas, J. Kikidis, I. Tomkos, and J. Prat, “Design characteristics for a full-duplex IM/IM bidirectional transmission at 10 Gb/s using low bandwidth RSOA,” J. Lightwave Technol. 28(7), 1094–1101 (2010). [CrossRef]