1. Introduction

With the development of large-bandwidth and high-speed electronic analog-to-digital converters (ADCs) and photo detectors (PDs), once again, coherent detection with digital signal processing (DSP) has been attracting a great deal of interest in recent years [1–5]. Thanks to DSP technology, impairments in optical transmission can be compensated by equalization in electrical domain [2]. Recently, homodyne detection has been discussed and investigated a lot and already applied in commercial coherent communication systems for 100G, 400G, 1T or beyond [3–5]. For homodyne detection in polarization division multiplexing (PDM) systems, in-phase/quadrature (I/Q) components of each polarization state should be separated in optical domain with full information. Thus, four balanced PDs with dual-hybrid structure and four-channel time-delay synchronized ADCs are required.

By up-converting I and Q components to the intermediate frequency (IF) at the same time, not only can heterodyne coherent detection reduce the number of the balanced PDs and ADCs of the coherent receiver into half [6], but also it’s no need to consider the delays between I and Q components in the PDM signal. Therefore, the four output ports of the optical hybrid can be also reduced into half. However, this technique is restricted by the bandwidth of PDs, while the external downconversion of the IF signals increases the complexity of the system. Furthermore, compared to homodyne detection, there is extra 3-dB signal-to-noise ratio (SNR) impairment for heterodyne detection. That’s why all the commercial products for coherent transmission systems utilize the homodyne detection technique.

However, nowadays, the ever increasing speed and bandwidth of ADCs and PDs give a possibility to commercialize the simplified heterodyne detection. With large-bandwidth PDs and ADCs, downconversion of the IF, I/Q separation of quadrature signal, and equalization for the PDM and nonlinear effect can all be realized in digital domain after ADCs. The report for heterodyne detection in transmission system is a limited 5-Gb/s 4-ary quadrature amplitude modulation (4QAM) signal over 20 km in [7] and limited 20-Mbaud 64 and 128QAM over 525 km in [8]. The high-order modulation formats, such as PDM-16QAM and PMD-64QAM, taking the advantage of high bandwidth occupancy, tend to be effective to realize commercial 100G or beyond adopting heterodyne detection. For the 100G or beyond coherent system with required transmission distance smaller than 1000 km, the inferiority of the SNR sensitivity in heterodyne detection is not so obvious. Conversely, less number of ADCs and easy implementation of the DSP for IF downconversion make heterodyne detection a potential candidate for100G or beyond transmission system.

In this letter, we propose and experimentally demonstrate a simplified coherent receiver based on heterodyne detection with only two balanced PDs and two ADCs. Compared to the conventional hybrid for homodyne detection, the polarization-diversity hybrid here is also simplified. The downconversion of the detected IF signals and I/Q separation are realized in digital frequency domain after ADCs. Using this scheme, we successfully demonstrated 8 × 112-Gb/s wavelength-division-multiplexing (WDM) PDM-16QAM over 720-km single-mode fiber (SMF)-28 with heterodyne detection and erbium-doped fiber amplifier (EDFA)-only amplification. To our knowledge, it is the first report on a 100G transmission system adopting heterodyne coherent detection based on DSP.

2. Principle of simplified heterodyne detection based on DSP

Figure 1(a) shows the principle of proposed coherent receiver with simplified heterodyne detection. The proposed receiver consists of two polarization beam splitters (PBSs) for polarization-diversity splitting between PDM signal and local oscillator (LO), two optical couplers (OCs) and two balanced PDs. We can see that only two balanced PDs and two ADCs are needed. Compared to the conventional hybrid for homodyne detection, the polarization-diversity hybrid here is also simplified.

 

Fig. 1 (a) The principle of simplified coherent receiver with heterodyne detection. (b) The principle of downconversion and I/Q separation.

Download Full Size | PDF

Figure 1(b) shows the signal process for downconversion and I/Q separation by simple digital frequency domain format. Here, we suppose the bandwidth of the ADC is large enough. After heterodyne detection at the balanced PDs with the LO, the generated electrical signal consists of both the baseband and IF components carrying the entire I and Q components after fast Fourier transform (FFT). The LO has a large frequency offset ω_IF = ω_s – ω_LO, where ω_s and ω_LO are the frequency of the received signal and LO, respectively. The IF components are extracted by multiplying a transfer function that of a rectangular bandpass filtering profile. By re-choosing the frequency point with maximum power as the new zero-frequency point and reverse two parts separated by the this new zero-point frequency point, the IF signal can be down-converted to baseband by doing this frequency shift. Then the down-converted signal is transformed to digital time domain by inverse fast Fourier transform (IFFT). The I and Q components of the signal after IFFT are the corresponding I and Q components of the PDM signal. Compared to the external IF downconversion based on frequency beating with radio-frequency (RF) signal by using electrical mixer, the DSP for signal downconversion in digital frequency domain can be simple implemented. Although the required analog bandwidth and sampling speed of the PDs and ADCs are significantly increased (at least doubled), the basic operation of DSP for heterodyne detection is much hardware-efficient on the premise of the reduced PD and ADC number.

3. System setup

3.1 The generation and transmission of an 8 × 112-Gb/s WDM PDM-16QAM with simplified heterodyne detection

Figure 2 shows the experimental setup for the 8 × 112-Gb/s WDM PDM-16QAM on a 25-GHz grid with simplified heterodyne coherent detection. Eight external cavity lasers (ECLs), with the line-width less than 100 kHz and the output power of 14.5 dBm, are divided into two groups and used as the continuous wavelength (CW) light source for the odd and even channels, respectively. Each group of ECLs has 50-GHz neighboring frequency spacing. The odd and even groups of ECLs are respectively combined by two polarization-maintaining optical couplers (PM-OCs), and then modulated by two I/Q modulators (I/Q MODs). The two I/Q MODs are respectively used to modulate the two groups of optical carriers with I and Q components of the 14-Gbaud electrical four-level signals, which are generated by two digital-to-analog converter (DAC) with sample rate of 28 GSa/s (2 samples per symbol). For the generation of optical 16QAM signal, the two parallel Mach-Zehnder modulators (MZMs) in each I/Q MOD are both biased at the null point and driven at the full swing to achieve zero-chirp 0- and π-phase modulation. The phase difference between the upper and the lower branches of each I/Q MOD is controlled at π/2. Before I/Q modulation, two middle levels of the 14-Gbaud four-level signal are pre-compensated by multiplying 0.89 to overcome the nonlinear characteristics of the two parallel MZMs in each I/Q MOD. Inset (a) in Fig. 2 shows the eye diagram of the generated optical 16QAM signal. Polarization multiplexing for each path is realized by the polarization multiplexer, comprising a PM-OC to halve the signal, an optical delay line (DL) to provide a delay of 150 symbols, and a polarization beam combiner (PBC) to recombine the signal. The two paths are combined using a 1 × 2 optical coupler. The WDM signal is launched into the straight line of 9 × 80-km SMF-28. Each span has the average loss of 18 dB and the chromatic dispersion (CD) coefficient of 17ps/km/nm at 1550 nm in the absence of optical dispersion compensation. EDFA is used to compensate the loss of each span. The total launched power (after EDFA) into each span is 8 dBm, corresponding to ~-1dBm per channel at 112 Gb/s. At the receiver, a tunable optical filter (TOF) with the 3-dB bandwidth of 0.2 nm is used to choose the desired channel. An ECL with a line-width less than 100 kHz is used as the LO. Two PBSs and PBCs are used to realize the polarization and phase-diversity coherent detection of the LO and the received optical signal before balanced detection. The analog-to-digital conversion is realized in the digital scope with the sampling rate of 120 GSa/s and the electrical bandwidth of 45 GHz. The optical spectra before and after optical hybrid are respectively shown as inset (b) and (c) in Fig. 2, where the optical power difference of the LO and signal light wave is larger than 20 dB.

 

Fig. 2 Experimental setup. (a): eye diagram of the 14-Gbaud optical 16QAM, (b) and (c): optical spectra before and after heterodyne receiver (0.1 nm-resolution), (d)-(e): electrical spectra of x-polarization in three different cases. PM-OC: polarization-maintaining optical coupler, DAC: digital-to-analog converter, PBC: polarization beam combiner, DL: delay line, ATT: optical attenuator, TOF: tunable optical filter, ADC: analog-to-digital converter.

Download Full Size | PDF

3.2 Digital signal processing

The process of digital IF downconversion is described in Section II. Figure 2(d) shows the electrical spectrum after ADC. It can be seen that the centralized frequency of IF is 12.5 GHz. An obvious frequency dip appears at the frequency of 45 GHz, where is the end of ADC bandwidth. After multiplying a rectangular transfer function of passband from 0~40 GHz, the useful IF containing I and Q components is extracted as shown in Fig. 2(e). By re-choosing the zero-frequency point and doing the reverse process mentioned in Section 2, the electrical spectrum of the down-converted signal is shown as Fig. 2(f). For the equalization based on DSP, a T/2-spaced time-domain finite impulse response (FIR) filter is used for the compensation of chromatic dispersion (CD). The two complex-valued, 13-tap, T/2-spaced adaptive FIR filters are based on the classic constant modulus algorithm (CMA) followed by three-stage CMA, to realize multi-modulus recovery and polarization de-multiplexing [3]. The carrier recovery is performed in the subsequent step, where the 4-th power is used to estimate the frequency offset between the LO and the down-converted signal. The phase recovery is obtained by feed-forward and followed Least-Mean-Square (LMS) algorithms. Finally, differential decoding is used for bit-error-ratio (BER) calculating after decision.

4. Experimental results

Figure 3 shows the back-to-back (B2B) BER versus optical signal to noise ratio (OSNR) for the 14-Gbaud WDM PDM-16QAM channel at 1549.34 nm in the case of different offsets between LO and signal light wave. Here, the frequency offset is set at 12.5 GHz in order to realize the detection of two channels with one LO light wave. For the 12.5-GHz offset case, the required OSNR is 23 dB at the BER of 1 × 10⁻³. By removing the LO to increase the offset to 15 GHz, the improvement is very small. In this heterodyne detection, the optical power difference between LO and the signal light wave is larger than 20 dB. Compared to the PDM-16QAM components in the baseband, the useful frequency components on the IF play a dominant role in the whole spectrum. The crosstalk from the baseband electrical signal is not evident. There is about 3-dB OSNR penalty at the BER of 1 × 10⁻³ for simplified heterodyne detection compared to the conventional homodyne detection.

 

Fig. 3 The measured B2B BER versus OSNR for the 14-Gbaud WDM PDM-16QAM channel at 1549.34 nm.

Download Full Size | PDF

Figure 4(a) gives the BER/OSNR versus transmission distance for the 14-Gbaud WDM PDM-16QAM channel at 1549.34 nm. It can be seen that there is 3-dB OSNR decrease for the WDM signal when the transmission distance increases from 400 km to 800 km. Meanwhile, the corresponding BER increases from 4 × 10⁻⁴ to 9.2 × 10⁻³. It’s because the enhanced nonlinear effect becomes dominant and worsen the BER performance. Figure 4(b) gives the varying BER performance of the WDM PDM-16QAM channel at 1549.34 nm after 400-km SMF-28 transmission by changing the launched power into each span. The total input power of 8 dBm (corresponding to ~-1 dBm per channel) gives the best BER performance.

 

Fig. 4 (a) The BER/OSNR versus transmission distance, (b) The BER versus input power after 400-km SMF-28 transmission.

Download Full Size | PDF

Figure 5(a) shows the optical spectra before and after 720-km SMF-28 transmission. The insets (I) and (II) are the obtained constellations for at the 1549.34-nm WDM channel before (28-dB OSNR case with attenuation) and after (25.6-dB OSNR) transmission, respectively. The average OSNR of the signal after 720-km SMF-28 transmission is 25.6 dB. The measured BER of all channels after transmission over 720-km SMF-28 is shown in Fig. 5(b). The BER for each of the 8 channels with optimum launch power is smaller than the FEC limit of 3.8 × 10⁻³ [9].

 

Fig. 5 (a) The optical spectra (0.1-nm resolution) before and after transmission over 720-km SMF-28. (b) The BER of all channels over 720-km SMF-28.

Download Full Size | PDF

5. Conclusion and discussion

We first experimentally demonstrate a coherent receiver based on simplified heterodyne detection for 112-Gb/s PDM signal. The 8 × 112-Gb/s WDM PDM-16QAM on a 25-GHz grid is generated and transmitted over 720-km SMF-28. The number of the output ports of the optical hybrid can be reduced into two. The benefits from the simplified heterodyne receiver architecture and the effective DSP in digital frequency domain for entire frequency-domain IF downconversion and I/Q separation are experimentally demonstrated.