Xiang Zhou - AT&T Labs Research

258kB Size 2 Downloads 11 Views

Such a processing delay is greatly increased with parallel processing typically required in high-speed optical systems [1]. For quadrature phase shift keying ...
Hardware Efficient Carrier Recovery Algorithms for Single-Carrier QAM systems

Xiang Zhou

1) AT&T Labs-Research, 200 Laurel Ave South, Middletown, NJ 07748, USA, [email protected]

Abstract: This paper presents an overview on recent advancement of carrier synchronization techniques for optical systems using single- carrier quadrature-amplitude-modulation (QAM), with a special emphasis on a class of newly proposed multi-stage blind carrier recovery algorithms. ©2011 Optical Society of America OCIS codes: (060.1660) Coherent communications; (060.2330) Fiber optics communications.

Introduction Digital coherent detection coupled with the use of high-order M-ary quadrature amplitude modulation (QAM) plus polarization multiplexing has proven to be an effective method to improve spectral efficiency for future 400G and beyond transmission. To implement M-QAM with M≥16, one major challenge in the optical coherent receiver is accurate carrier phase recovery. Although the decision-directed phase-locked loop (DD-PLL) has been widely used in wireless QAM systems for carrier synchronization, its tolerance to laser phase noise is poor due to its inherent feedback processing delay. Such a processing delay is greatly increased with parallel processing typically required in high-speed optical systems [1]. For quadrature phase shift keying (QPSK), the mth-power-based, feedforward, blind phase-recovery algorithm is well known and several modified Mth-power algorithms have been proposed for M-QAM [2-5]. However, these algorithms either have lower linewidth tolerance [2,4] or can only be applied to 16QAM [3,5]. It has been shown that the minimum distance estimation (MDE) based, blind phase search method (BPS) can achieve nearly optimal laser-linewidth suppression [1]. However the implementation complexity of this method is very high. To address the complexity issue, several hardware-efficient multi-stage feed-forward phase recovery algorithms have been proposed recently [6-7]. Due to the nature of feed-forward configuration, these multi-stage algorithms exhibits much better laser phase noise tolerance than the traditional DD-PLL method while the implementation complexity is significantly lower than the single-stage based BPS method. For high-speed optical systems, prior to the phase recovery, the frequency offset between the LO and transmitter laser needs to be removed. For MPSK signals, such a frequency recovery can be achieved by using either a time-domain based ‘averaging’ method or a FFT based ‘averaging’ method after the data modulation being removed using the well known Mth-power algorithm [8]. This method can be extended to higher-order QAM systems using FFT based method [9] but the required FFT sizes increases as the modulation order increases (tens of thousands symbols needed for high-order QAM systems). Recently a new two-stage MDE based algorithm has been proposed which allow us to achieve reliable carrier frequency recovery using only tens to around 100 symbols [10]. This paper presents an overview on these multi-stage based carrier frequency and phase recovery algorithms.

Multi-stage BPS/ML phase recovery algorithm [6] For this method, a coarse blind phase search method is used in the first stage to find a rough location of the optimal phase angle. Decisions made following this rough phase estimation are then used for more accurate phase estimation through a ML phase estimate in the second stage. To further improve the accuracy of phase estimation, one can add more than one ML phase estimation stage to approach the optimal phase angle iteratively. The effectiveness of the proposed method has been verified by numerical simulation using a 38Gbuad square PDM-64QAM system as is shown in Fig. 1, where the required equivalent number of test phase angle (ENTPA) using three different feed-forward carrier phase recovery scenarios, the single- stage BPS method and the proposed two-stage and three-stage hybrid BPS/ML phase recovery schemes has been displayed. For these investigations, the laser phase noise for both the signal source and the local oscillator (LO) is assumed to be 100kHz. The received optical-to-signal noise ratio (OSNR) in 0.1 nm noise bandwidth and for a single polarization is 28 dB. From Fig. 1 one can find that, to achieve a performance that is close to the optimum, the single-stage BPS method needs to test about 64 different phase angles, while the proposed three-stage hybrid BPS/ML algorithm only needs to equivalently test 18 different phase angles (14 test phase angles used in the first coarse BPS stage plus two cascaded ML phase estimation stages), resulting in a reduction of computational effort by more than a factor of 3. Multi-stage PLL/ML Phase recovery algorithm [7] To further reduce the implementation complexity, one can use a DD-PLL for coarse phase recovery and one or two ML estimators for phase fine- tuning. Such a multi-stage configuration can reduce the implementation complexity by more than one order of magnitude as compared to the single- stage BPS method for a 64QAM system, but at the cost of slightly reduced laser phase noise tolerance when the degree of parallelization is high. [pic] [pic] (a)

(b) Fig. 2. Experimental results for a 9.4Gbaud 64QAM system, where (a) shows the BER performance for different degrees of parallelization with 23 dB OSNR and (b) shows the BER performance at different OSNR levels. PLL delay D=5. The effectiveness of the proposed multi-stage PLL/ML phase recovery algorithm has been tested in a 9.4Gbaud 64QAM (single polarization) back-to- back experiment as is shown in Fig.2a and b. Fig. 2(a) shows the impact of parallel processing on the proposed algorithms for a constant 23dB OSNR. One can observe that the proposed algorithm with two MLEs can achieve the same BER performance as the BPS method for P up to 20. The BER performance versus OSNR level for P=16 is given in Fig. 2b. The proposed algorithm can achieve performance similar to the BPS method for a wide range of OSNR levels with BER ranging from 2(10-2 to close to 10-4. Two-stage MDE based blind frequency recovery algorithm [10] For this method, the frequency offset is first scanned at a coarse step size ((10MHz) and then at a fine step size of ((1MHz), and the optimal frequency offset is the one that gives the minimum mean square error. For each trial frequency, the carrier phase is first recovered (with best effort) by using the above described BPS/ML blind phase estimation algorithm, and decisions made following this phase estimation are then used as reference signals for mean square error calculation. This frequency search method can reliably recover carrier frequency by using only tens to around 100 symbols (see Fig. 3) and therefore can achieve very fast carrier frequency recovery if it is implemented with parallel processing. For the typical coherent receiver where the carrier frequency varies much more slowly than the symbol rate, the proposed blind frequency recovery method can be implemented with a sequence processing architecture (such as the CPU in the computer) to reduce the implementation complexity. Conclusions Fast and reliable carrier recovery is one of the major challenges for coherent systems using high-order QAM formats. Several blind phase recovery algorithms have been proposed, among them, multi-stage based BPS/ML algorithm can achieve a linewidth tolerance close to the optimum but with significantly reduced implementation complexity as compared to the single- stage based BPS method. The implementation complexity can be further reduced by using a multi-stage based PLL/ML algorithm but with slightly reduced linewidth tolerance. Robust carrier frequency recovery can be achieved using a two-stage MDE based blind search method by using only tens to around 100 symbols. References [1] T. Pfau et al, JLT Vol. 27, pp.989, 2009 [2] M. Seimetz, OFC/NFOEC’08, OTuM2. [3] H. Louchet et al, ECOC’08, TU 1.E.6. [4] Y. Cao, et al, ECOC2010, We.7.A.1. [5] I. Fatadin et al, PTL 2010, pp. 631-633. [6] X. Zhou, IEEE Photonics Technol. Lett , pp. 1051-1053, 2010 [7] X. Zhou, et al, OFC’11, OMJ3, 2011 [8] A. Leven, et al, IEEE Photonics Technol. Lett. pp.366, 2007 [9] M. Selmi et al, ECOC’09, P3. 08, 2009 [10] X. Zhou et al, JLT, Vol. 29, pp.571, Feb. 2011