Training Sequences Design for Simultaneously Transceiver IQ Skew Estimation in Coherent Systems

Transceiver in-phase/quadrature (IQ) skew due to device imperfections is one of the major factors restricting the transmission performance of coherent systems, especially for the signal with high baud rate and high-order modulation format. Although two 4×4 real-valued multiple-in multiple-out (MIMO) equalizers can deal with the transceiver IQ skew impairments to some extent, it leads to high computational complexity in the receiver side digital signal processing (DSP). Since the transceiver IQ skew is a relatively static impairment, the precise skew pre-compensation is an attractive alternative if the transceiver IQ skew values are available. In this paper, a novel low-complexity and precise skew estimation method based on a specially designed training sequence is proposed for coherent systems, which can simultaneously estimate the transmitter side and receiver side IQ skew (Tx/Rx-skew) by a simple single-shot measurement in the frequency domain. The designed training sequence exhibits the following features: 1) concatenated by the pure real and imaginary sequences in the time domain, and 2) composed of four identical frequency blocks in the frequency domain. The former feature achieves the separation of the Tx-skew and Rx-skew since the pure real and imaginary signals pass through the I and Q channels individually at the transmitter side, while the latter one ensures a low-complexity acquirement for the skew value by the conjugate multiplication and sum (CMS) operation between adjacent frequency blocks. The effectiveness of the proposed method is verified both in simulation and experiment, the results show that the estimation error is within ±0.2ps, and the measurement range covers -2 T_s to 2 T_s , where T_s is the symbol period. The transmission of 50GBd dual-polarization 16/32QAM signal is experimentally demonstrated to validate the proposed IQ skew estimation and pre-compensation method. After transceiver skew calibration with the proposed method, the required optical signal-to-noise ratio (OSNR) can be reduced by about 2dB at the hard-decision forward error correction (HD-FEC) threshold of 3.8×10 ^-3 when existing 1ps skew at the transceiver for 50GBd 16QAM signal.