Fig. 1. Frame structure of the training sequency used by sender

Fig. 2. Structure of FOE algorithm based on phase difference of training sequence and CZT

Fig. 3. Simulation platform of FTN-WDM coherent light transmission system

Fig. 4. Absolute value of the maximum FOE error of rough estimation of FTN-16QAM with accelerate factor of 0.9 and OSNR of 24 dB. (a) Under different lengths of single training block; (b) under different lengths of training sequence

Fig. 5. The 4th CZT spectrogram with carrier frequency offset of 1 GHz and accelerator factor of 0.9, where f1 is the rough estimation result and f2 is the fine estimation result. (a) Rough estimation residual frequency offset is large; (b) rough estimation residual frequency offset is small

Fig. 6. Absolute value of maximum FOE error of FTN-16QAM signal under different conditions. (a) Error under -20~20 GHz frequency offset using 4th-FFT algorithm (1024 points) and proposed method; (b) error under -1.6~1.6 GHz frequency offset with proposed method

Fig. 7. BER curves of different accelerated factors for FTN-16QAM signal under 1 GHz frequency offset with proposed method

Fig. 8. Experimental setup for back-to-back measurements for the triple-carrier 40 GBaud PM-16QAM FTN-WDM coherent systems

Fig. 9. Absolute value curves of max FOE error under different conditions. (a) Error curves under different frequency offset with different accelerated factors; (b) error curves under different frequency offset and AF using 4th-FFT algorithm (512 points) and proposed method

Fig. 10. BER curves with α_AF=1.0 (Nyquist system) and α_AF=0.9 (FTN system) under different OSNR