Analysis of narrow-FSK downstream modulation in colourless WDM PONs I. Garce ´s, J.C. Aguado, J.J. Martı ´nez, A. Lo ´pez, A. Villafranca and M.A. Losada The performance of a colourless WDM PON architecture based on frequency-shift keying (FSK) by direct modulation of a DFB laser diode for the downstream and intensity modulation by means of a reflective semiconductor optical amplifier for the upstream channel is assessed. The operation of the architecture is demonstrated, showing excellent performance for distances up to 25 km. Introduction: Wavelength division multiplexing (WDM) PONs are a promising technology for the next generation access networks, and are being extensively studied as a means to increase the number of users using a well mature technology. A key issue in current research is to study architectures with a colourless ONU design where the optical network unit (ONU) is wavelength seeded from the central office (CO) using spectral sliced sources or CW lasers which are modulated in the upstream channel by means of reflective semiconductor optical amplifiers (RSOAs) [1]. In some of these proposals the CW seeding wavelength is also modulated in FSK [2], or DPSK [3] formats, but these proposals need external modulators or special lasers to achieve a good performance. Recently, we proposed a new method for FSK modulation of the downstream channel based on the frequency shift caused by direct modulation of a DFB laser [4], which was termed narrow-FSK modulation [5]. The main interest of this approach is its simplicity, combined with a high spectral efficiency compared to similar schemes based on DFB direct modulation, because here the spectrum of the modulated signal is only slightly wider than the CW signal. Moreover, the narrow-FSK modulation increases the effective optical bandwidth of a purely CW downstream signal, decreasing the efficiency of effects such as Brillouin and Rayleigh scattering [6] and allowing transmission of higher optical powers from the optical line terminal (OLT). Network architecture: The network architecture that we propose is depicted in Fig. 1. It is based on a set of directly modulated lasers (DMLs), each one serving 16 to 64 users with a dynamic TDMA request=grant approach (as, for example, in EPON), or any other suitable method to obtain collision avoidance. Fig. 1 Proposed WDM-PON architecture The downstream channel is narrow-FSK modulated taking advantage of the laser adiabatic chirp in the frequency range where this is the dominant effect. Other chirp effects, such as transient and thermal chirp [7], however, restrict the usable frequency range. The laser modulation must be low enough to grant that the intensity modulation (IM) ratio of the downstream signal is negligible compared to that of the upstream signal, but sufficient to ensure that the optical carrier frequency of the laser is slightly modulated by its own adiabatic chirp. This frequency excursion depends on the absolute power levels for ‘1’ and ‘0’ and thus a low extinction ratio value is required. The downstream data is carried from the OLT to the ONUs, where the FSK signal is demodulated using an athermal filter and then directly detected. This same signal is then intensity modulated and amplified in the ONU using a RSOA to carry the upstream data. Experiment: Fig. 2 shows the experimental setup used to test the architecture. Both upstream and downstream channels were emitted simultaneously. For the downstream channel we used a 1544 nm DFB laser source (CQF935=27 from JDS) emitting þ7 dBm optical power with a linewidth enhancement factor of 3.6 and an adiabatic chirp of 3 GHz=mW. Fig. 3 shows the optical spectra of the laser in CW and when the FSK modulation is applied, measured with a high resolution optical spectrum analyser (BOSA-C from Aragon Photonics) that allows resolving the optical frequency separation between ‘1’ and ‘0’ optical power levels. At the reception end, the FSK signal is demo- dulated using an athermal Mach-Zehnder differential interferometer (MZDI) filter from Optoplex Inc. with 2 dB insertion loss, 2.5 GHz free spectral range and 0.3 GHz=  C thermal behaviour in all the C-band. The insertion loss of this device against wavelength is also shown in Fig. 3. Fig. 2 Experimental setup for characterisation of proposed architecture using Gb Ethernet channels in up and downstream channels Fig. 3 Measured optical spectra of CW DFB (dark grey), narrow-FSK modulated signal (black) and insertion loss of demodulator (dashed) In the upstream channel we used a RSOA (SOA-RL-OEC-1550 from the Centre for Integrated Photonics, UK) with þ5 dBm optical satura- tion power and 20 dB optical gain at a bias current of 500 mA. Polarisation dependence of the RSOA was lower than 1.5 dB. Fig. 4 BER and eye diagram of downstream narrow FSK signal back-to- back and with 25 km standard SMF A GbE analyser (Advisor J3446 from Agilent) with two GbE outputs is used to feed both upstream and downstream channels simultaneously. Data traffic is real Ethernet traffic at 100% channel load. BERs presented in this Letter from now on are obtained as the number of frame check sequence (FCS) errors over the total number of transmitted bits, which is a good way to measure errors provided that there is no loss of Ethernet frames. This way, and due to the codification of the Ethernet signal, we obtain a nearly pure random signal but without the ELECTRONICS LETTERS 12th April 2007 Vol. 43 No. 8 Downloaded 09 May 2007 to 155.210.207.77. Redistribution subject to IET licence or copyright, see http://ietdl.org/copyright.jsp