Open Access
Add to BibSonomy
Abstract
This paper investigates the self-homodyne detection system with orthogonal polarization between pilot and signal (oSHD) through optical polarization demultiplexing of carrier and signal at receiver. Compared to digital demultiplexing, optical demultiplexing provides around a 0.9dB benefit in receiver sensitivity. The impact of crosstalk due to optical demultiplexing has been investigated in terms of carrier-to-signal power ratio, and delay between signal and carrier. Results show that the oSHD system is very tolerant to crosstalk, with a negligible penalty for crosstalk up to −22dB. Experiments were carried out to evaluate the system performance, including polarization tracking performance, long term stability, bit error rate performance, and tolerance to laser linewidth and delay between signal and pilot. 41Gbaud 64QAM oSHD with 200Gb/s net data rate has been realized by using the scheme with receiver sensitivity better than −4dBm, providing a promising candidate solution for 800Gb/s Ethernet.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
The rapid growth of data centers is driving the speed requirement of Ethernet solutions. There have been huge amounts of research and development activities on optical modulation formats for short reach optical networks [1–3], particularly intra-data centers. For the 400Gb/s Ethernet (400GbE) currently being deployed, 4 level pulse amplitude modulation (PAM4) was adopted. As applications move to the next generation of Ethernet, such as 800GbE or 1.6TbE, the further adoption of PAM4 with the same wavelength number may suffer from complex DSP and poor receiver sensitivity [4] or require high electrical bandwidths, up to 60GHz for transponder components, which is currently commercially challenging, particularly for transimpedance amplifier (TIA). Coherent detection can provide good performance and higher spectral efficiency while relaxing the component bandwidth requirement. However, the cost and DSP power consumption are two major obstacles for coherent detection being used to short-reach applications.
An intermediate solution, which provides higher spectral efficiency than intensity modulation direction detection while having less complexity and lower power consumption than coherent detection, is self-homodyne detection (SHD) [5–8]. SHD is known to be extremely laser linewidth tolerant and has two variants. One is that, the carrier and signal have the same polarization, where a single-sideband (SSB) signal is self-homodyned to a carrier through the square-law detection process. The optical phase information of the transmitted signal is translated directly into electrical phase information and preserved. In order to avoid signal-signal beating induced interference, carrier to signal power ratio (CSPR) higher than 10dB is usually required. A gap had been introduced between the carrier and SSB signal to achieve low CSPR operation [9–12], but this reduces the electrical spectral efficiency. To eliminate the gap in the above scheme, the Kramers–Kronig (KK) receiver has been proposed [13], which can gain high spectral efficiency while still achieving CSPR as low as 6dB. Another method to eliminate the gap is to use nonlinear equalization to suppress the impact of signal-signal beat noise in the digital domain [14–16]. But all of the SSB based schemes mentioned above require receiver bandwidths twice that of the baseband-based signal. For example, an SSB 16QAM system has the same electrical bandwidth requirement as that of PAM4 for the same total data rate.
Another SHD scheme, which uses orthogonal polarization between the pilot carrier and signal, referred to as the oSHD hereafter, has been proposed in [17–18]. In the oSHD system, baseband modulation and detection can be achieved via Stokes vector direct detection (SVDD), where CSPR as low as 0 dB is feasible. A typical SVDD receiver configuration has been well discussed in [17–18], which requires 3 sets of each: balanced photodetectors (BPDs), TIAs, and analog to digital convertors (ADCs). Meanwhile, polarization demultiplexing in the optical domain has been proposed to reduce the DSP power consumption of dual-polarization coherent systems [19–20]. We here propose separating the optical carrier and signal for oSHD system via optical polarization demultiplexing, which only requires two sets of BPDs, TIAs, and ADCs. For a sensitivity-limited system, such as 800Gb/s Ethernet, where no optical amplifier is preferred, a smaller number of BPD + TIA means less power splitting or lower insertion loss for the same receiver input optical power, leading to a higher signal-to-noise ratio. Automatic polarization separation of dual polarization multiplexed channels in optical domain has been demonstrated in [21,22], but only for low order formats. Furthermore, a dual polarization coherent detection system where the signal and local oscillator (LO) were from the same laser source and the LO was remotely delivered to the receiver via a different fiber and its polarization was tracked and stabilized by a polarization tracker at the receiver side, has been proposed in [23]. In this paper, for the first time, we demonstrate automatic endless polarization separation of carrier and signal in the optical domain for an oSHD system with 41Gbaud 64QAM, providing a raw data rate of 246Gb/s per wavelength (net data rate of 200Gb/s with 11% FEC).
The paper is organized as follows. Section 2 introduces the system configuration, where oSHD transmitter and receiver configurations are discussed. Section 3 compares the advantages and disadvantages of polarization separation in digital and optical domain. Section 4 discusses the performance impact of crosstalk due to optical polarization demultiplexing. Section 5 provides experiment verification results, which include control algorithm, tracking stability, bit error rate performance, and tolerance to laser linewidth and delay between signal and pilot. Section 6 summarizes the paper.
2. System scheme
The proposed system consists of an oSHD transmitter, a fiber link, and an oSHD receiver, as shown in Fig. 1. Here the oSHD receiver uses optical demultiplexing to separate pilot and signal, and is referred to as polarization tracking oSHD receiver in the paper. In the oSHD transmitter, a continuous wave (CW) light is divided into two parts by an optical coupler (OC). One part of the CW light is used as the pilot carrier, which is rotated by 90 degrees in polarization via a polarization rotator (PR), while the other part is input into a single polarization IQ modulator (SP-IQM) and is modulated by electrical data. The modulated signal is then combined with the pilot via a polarization beam splitter (PBS). The transmitter output is an oSHD channel consisting of a pilot and a signal with orthogonal polarization. It is well known that the carrier-to-signal power ratio (CSPR), the power ratio of optical carrier to that of optical signal, is an important parameter in determining the required optical signal-to-noise ratio (rOSNR) or receiver sensitivity, which will be discussed in later sections.
Since the Mazh-Zehnder modulators for in-phase and quadrature channel modulation are biased at the transmission null point, the optical signal would suffer from high modulation loss. The modulation loss depends on the modulator Vπ, electrical signal driving swing, pulse shaping roll-off factor, modulation format, and S21 pre-compensation. In general, a lower order of modulation format, lower Vπ and larger driving swing yield lower modulation loss. While for band-limited transmitter, small pulse shaping roll-off factor and stronger S21 pre-compensation lead to higher modulation loss. For a typical silicon-based IQ modulator with a QAM order no lower than 16, the modulation loss (including 3 dB IQ modulation loss) is usually higher than 10 dB. For an optical noise dominated system where optical amplifiers are used, the coupling ratio of the optical coupler inside the oSHD transmitter may be adjusted to achieve 0 dB CSPR for better rOSNR performance. However, for short reach system, where optical amplifiers are avoided to reduce the cost and power consumption, a conventional coupling ratio of 3 dB is desired to maximize the link power budget.
The polarization tracking oSHD receiver consists of an automatic endless polarization separator (AEPS), a 90-degree hybrid mixer, two sets of BPDs, TIAs, and ADCs. The AEPS includes an endless polarization controller with two output ports and an FPGA based control circuit. Small portions of the two outputs of polarization controller are tapped out to the two corresponding monitoring PDs (mPDs). The outputs of the two mPDs are digitalized and fed into the FPGA based controller. The polarization tracking algorithm generates four outputs to control the endless polarization controller and to separate the signal and pilot. The optically demultiplexed pilot and signal are then inputted into the 90-degree hybrid mixer for single polarization self-homodyne detection, where the separated pilot is used as the local oscillator (LO).
The polarization tracking oSHD receiver can be based on various platforms, in the form of discrete components or integration. A discrete endless polarization controller may be based on either 4 rotatable quarter waveplates, i.e. Agilent 11896A, or 4 waveplates with variable retardance, i.e. fiber squeezers, oriented 45° from each other such as in General Photonics PolaRITE III. For a tracker to transform any arbitrary input polarization to any arbitrary output polarization, Stokes parameters may need to be detected. While for our current oSHD system where no accurate phase shift between the two outputs is collected, the detection of two output powers would be sufficient for the separation of signal and carrier.
An integrated version is desired in practical implementation, since it exhibits higher speed, smaller footprint, lower power consumption, and near zero delay between the two output ports of signal and LO. An integrated polarization tracking oSHD receiver based on silicon photonics is shown in Fig. 2, where the endless polarization controller (tracker) with two outputs consists of a polarization splitter rotator (PSR), followed by four pairs of phase shifters and four 2 × 2 50:50 optical couplers placed in alternative order [22,24]. These phase shifters and couplers are equivalent to 4 waveplates with variable retardance (like fiber squeezers) [24] and can realize endless tracking operations with bounded values for control parameters. The PSR separates the light into two parts, one with transverse electric (TE) polarization, and the other with transverse magnetic (TM) polarization. The TM polarization is then converted into the TE polarization via mode conversion, making both two outputs of the PSR have the same TE polarization. The optical coupler can be in the form of a directional coupler or multi-mode interference (MMI) coupler. The phase shifters can be adjusted via thermal heaters, which exhibit low power consumption and is easy to integrate with receiver. Conventional thermal effect has relatively low speed with response bandwidth in tens of kHz. However, with proper waveguide designs like using a thin buried oxide layer to enhance cooling, pre-emphasis in heater driving, and direct doping over the heater, the response time in sub-microseconds has been reported.
3. Comparison with SVDD receiver
As discussed above, oSHD channel can also be demultiplexed in the digital domain by using Stokes vector direct detection. A SVDD receiver configuration based on silicon photonics is shown in Fig. 3. The input signal is separated by the PSR into two light beams, EX and EY. EX is further split into E1 and E2 by an optical coupler (OC1). Similarly, EY is split into E3 and E4 by another optical coupler (OC2). E1 and E4 are sent to a BPD to generate the first Stokes parameter S1. E2 and E3 are input into a 90-degree hybrid mixer to generate 4 outputs, with two outputs sent into a second BPD to generate the second Stokes parameter S2, and the other two outputs sent to a third BPD to generate the third Stokes parameter S3. When ignoring the excess insertion loss of the hybrid, the optimal coupling ratio for the two OCs is 66.7%, which provides equal root mean square values for the 3 Stokes parameters.
From Fig. 2 and Fig. 3, it can be seen that the optical polarization tracking oSHD receiver only needs 2 sets of BPDs, TIAs and ADCs, while the SVDD oSHD receiver requires 3 sets of BPDs, TIAs and ADCs. Larger PD number results in less optical power into each PD for the same total input power into the receiver, leading to reduced signal-to-noise ratio and worse receiver sensitivity. A fair comparison can be carried out on the insertion loss of the two receiver configurations based on the same silicon platform. The optical paths of S2 and S3 from input to corresponding BPD in the SVDD oSHD receiver are straightforwardly equivalent to those of the polarization tracking oSHD receiver. The loss from input fiber coupling, PSR, hybrid mixer and PD responsivity, are the same for the two receiver configurations. The difference is that the polarization tracking oSHD receiver has a polarization tracker, while the SVDD oSHD receiver has a power assignment part as shown in Fig. 3.
The polarization tracker has four 2 × 2 50:50 optical couplers which combine and split input lights, and an optical tap in each output to extract monitor signal. The excess loss per optical coupler is around 0.15 dB, and the splitting loss of the optical tap is 0.22 dB by assuming 5% tapping ratio. This results in about 1 dB insertion loss for the polarization tracker. For the SVDD oSHD receiver, in addition to 0.15 dB excess loss, OC1 and OC2 with 66.7% coupling ratio induces 1.76 dB splitting loss for the optical paths of S2 and S3, respectively. This introduces around 1.91 dB insertion loss for the power assignment part. Therefore, the polarization tracking oSHD receiver provides around 0.9 dB gain in insertion loss to that of the SVDD oSHD receiver, leading to around 0.9 dB benefit in receiver sensitivity. This 0.9 dB gain may be further increased since the 1.76 dB splitting loss for SVDD receiver is fundamental while the tracker loss may be further improved.
4. Impact of crosstalk due to optical demultiplexing
Optical demultiplexing of pilot and signal may induce crosstalk, where some residual pilot may be leaked into the signal and vice versa. This could be caused by a finite polarization extinction ratio of the polarization beam splitter (or PSR in Fig. 2) or required feedback error signal for polarization tracking. The experimental work in the following section shows that crosstalk from optical demultiplexing for oSHD system could be up to −22dB level. This section will look at the impact of crosstalk on the system performance, particularly for higher order QAM which is more susceptible to optical crosstalk. Figure 4 shows a schematic configuration of optical polarization tracking oSHD receiver for crosstalk analysis. The input consists of a pilot carrier and a modulated signal with orthogonal polarization. The corresponding optical fields are represented by EC and ES, respectively. The CSPR is defined as:
The polarization tracker demultiplexes the incoming oSHD channel, and generates the resultant signal (ESig) and carrier (ELO), which can be related to EC and ES via following equations: Here θ is the state of polarization (SOP) angle relative to the TE polarization of the tracker. After the tracking is converged, θ is small and is close to 0, so $\cos \theta \approx 1$. After converting θ to crosstalk ε with $sin \theta = \sqrt \varepsilon $ and including the delay between ELO and ESig, Eqs. (2) and (3) can be written as, Here τ is the delay between the carrier and the signal, which may be introduced in the oSHD receiver. For an integrated oSHD receiver, the delay can be designed to be within fabrication tolerance of sub-millimeter scale.Crosstalk is basically related to the input SOP rotation speed and the speed of tracker. Generally speaking, slow input SOP rotation and a fast tracker can achieve low level of crosstalk.
After the hybrid mixer, the in-phase channel BPD generates two photo currents:
Even though this paper focuses on the delay introduced by the receiver, we acknowledge that the delay between carrier and signal can also be originated from the transmitter. This comes from the mismatch between the 2 optical paths: one going through the IQ modulator and the other going straight before they are polarization orthogonally combined. This kind of delay does impact the performance since there is no carrier recovery in the receiver DSP. However, optical path mismatch can be easily managed in the design phase of the oSDH transmitter, no matter in the platform of silicon, LiNbO3 or indium phosphide.
Simulations were carried out to analyze the impact of the crosstalk. We look at the impact of crosstalk for various CSPRs and delays between signal and carrier. 61.6Gbaud 16QAM and 41Gbaud 64QAM are used in our simulations, which provide 200Gbit/s net data with 11% FEC overhead and 1.6e-2 BER threshold. A polarization tracking oSHD receiver based on silicon platform is assumed. The parameters used in the simulation are detailed in Table 1.
Table 1. Simulation conditions.
The DSP used in the receiver-side has neither phase tracking nor carrier recovery based on phase locked loop. It consists of a matched pulse shaping filter, a one-tap phase recovery block which corrects the phase variation caused by input polarization changes, and a 2 × 2 real value time domain equalizer which compensates residual S21 and adjusts for the timing offset.
Figure 5 shows bit error rate (BER) versus received optical power (ROP) with 0 delay between carrier and signal for 61.6Gbaud 16QAM. The ROP is the total optical power into the receiver including carrier and signal. We look at the impact of various crosstalk levels including no crosstalk, −22 dB, −20 dB, −18 dB and −16 dB crosstalk, where the crosstalk in dB is defined as $10 \times lo{g_{10}}(\epsilon )$. At 10 dB CSPR, there is negligible power penalty for up to −16 dB crosstalk. At 6 dB CSPR, the power penalty is within 0.2 dB at BER=1.6e-2 for up to −16 dB crosstalk. As discussed above, for 0 delay between signal and carrier, carrier-carrier beat noise is a DC and carrier-signal beat noise is a second order term, so the major impact comes from the crosstalk induced signal-signal beat noise, which decreases with the increase of CSPR.
Fig. 5. BER vs. ROP for 61.6Gbaud 16QAM with 0 delay between signal and carrier. (a) CSPR = 10 dB, and (b) CSPR=6 dB.
Figure 6 shows BER versus ROP with 0 delay for 41Gbaud 64QAM. As expected, 64QAM is more sensitive to crosstalk than 16QAM. We again investigate the impact of various crosstalk levels including no crosstalk, −22 dB, −20 dB, −18 dB and −16 dB crosstalk. −22 dB is a number that we can achieve in experiment for 64QAM, as later shown in Fig. 11. With 10 dB CSPR, the penalty at BER=1.6e-2 is negligible for up to −22 dB crosstalk. The penalty is around 0.3 dB at BER=1.6e-2 for up to −16 dB crosstalk. With 6 dB CSPR, the penalty is within 0.2 dB for −22 dB crosstalk, and within 0.8 dB for up to −16 dB crosstalk.
Figure 7 shows BER versus ROP for 41Gbaud 64QAM with a 5ps delay. The performance degrades when there is a delay between carrier and signal. But for a small delay of 5ps between signal and LO, there is no performance degradation due to delay. With 10 dB CSPR, the power penalty at BER=1.6e-2 is still negligible for crosstalk up to −22 dB, and the power penalty is about 0.3 dB for up to −16 dB crosstalk. At low CSPR of 6 dB, the power penalty at BER=1.6e-2 is within 0.2 dB for −22 dB crosstalk and within 0.8 dB for up to −16 dB crosstalk. This delay used in the simulation is very conservative, since it is much larger than the fabrication tolerance of a silicon based oSHD receiver which is within 1ps. For short reach applications without optical amplifier, the CSPR for oSHD is around or above 10 dB. So, the performance degradation for 41GBaud 64QAM is negligible for crosstalk up to −22 dB. These results show that the oSHD system is very tolerant to crosstalk originating from optical polarization demultiplexing.
5. Experiment
Experiments have been carried out to verify the scheme and evaluate the performance of a 200G oSHD system with optical polarization demultiplexing. The experimental setup is shown in Fig. 8. The transmitter was a dual polarization IQ modulator following a DFB laser, where the X polarization was modulated by electrical data, and Y polarization was not modulated and was biased at the transmission peak point to allow the maximum pass of the carrier. The transmitter output was an oSHD channel of 41Gbaud 64QAM with around 10.5 dB CSPR. Its SOP was rotated by an Agilent 11896A polarization scrambler operating at scan rate 5. The 11896A polarization scrambler generates random SOP changes, which can be well described by a Rayleigh distribution. Scan rate 5 generates a mean SOP change rate of 10.5 rad/s and a maximum SOP change rate of 31.5 rad/s. Here the maximum SOP change rate is defined as thrice the mean SOP change rate, where less than 0.1% of the SOP change will be faster than the maximum. Using maximum SOP change rate to evaluate SOP is preferred in engineering since the system performance is dominated by the fastest SOP changes (worst scenario). A General Photonics (GP) polarization controller (PolaRITE III with evaluation board) with 4 fiber squeezers followed by a polarization beam splitter (PBS) was used as the polarization tracker. The evaluation board has a built-in driver for each squeezer which provides a gain of 30, and the PolaRITE III can take voltages between 0 and 150V. In this experiment, we set our evaluation board input to the range of [0-4] V, so the applied voltage to the PolaRITE III was in the range of [0-120] V. The photocurrents from the two mPDs are fed into the control circuit, where they are digitalized and then processed by the control algorithm in FPGA. In FPGA, the monitoring photocurrents are normalized by their sum to eliminate the impact of input power fluctuation. The separation of pilot and signal was realized by a control algorithm via minimizing the output power of signal port. The demultiplexed pilot was used as LO and input into a single polarization integrated coherent receiver (SP-ICR) together with the demultiplexed signal for self-homodyne detection. An optical delay line (ODL) was introduced in either the LO or the signal path to adjust the delay between the signal and carrier. The two outputs of the SP-ICR are collected by a Keysight (KS) digital storage oscilloscope (DSO) with 80G sampling rate. The DSP routine used in experiment was the same as that of simulation.
For the tracker based on the GP polarization controller, there are 4 fiber squeezers, with each corresponding retardance Φk controlled by an evaluation board input voltage Vk (also called control voltage here), with the squeezer index k = 1, …, 4. For each fiber squeezer, Vk was limited to the range of [Vlow, Vupper], where Vlow = 0 V, and Vupper = 4 V in our experiment. Initially, every Vk was set to the middle of the control voltage range, Vmid = (Vlow +Vupper)/2.
The endless tracking control algorithm that was programmed on the FPGA is based on a gradient algorithm and is described as following steps.
For each time step or control loop step with time step index n, there are 4 operation steps, each corresponding to a squeezer.
In each operation step, a control voltage Vk(n) is first applied to the kth squeezer, then the two mPDs detect the photocurrents from the signal port and LO port of the tracker outputs. The two ADCs inside the FPGA A/D convert the two photocurrents, which are then averaged and normalized over their sum.
To get the slope (gradient) of iSig, a small voltage deviation δ is added to Vk(n) to get iSig[Vk(n)+δ], so the slope is calculated as
The sign of K decides whether to increase or decrease the control voltage for the next step. We define a tracking voltage increment ΔV1 and a pulling voltage increment ΔV2: Here A1 and A2 are the scaling factors which are two positive numbers. The new control voltage for next time step is expressed as The operation repeats over the 4 fiber squeezers to form a loop step. Then the procedure is cyclically repeated over loop by loop.ΔV1 in Eq. (12) is adjusted by A1 to make it converge sufficiently fast capped at about 0.0125Vπ,s to avoid any possible instability, where Vπ,s is the required input voltage of the polarization tracker evaluation board to achieve 1π retardance by the squeezer. A2 in Eq. (13) is adjusted to make ΔV2 larger than ΔV1 at the limit of the control voltage, so the voltage is pulled back toward the middle when it is closing the boundary. Essentially, two fiber squeezers with 0° and 45° orientations can support polarization transformation with non-endless control, while four squeezers with 45° alternative orientations will achieve endless tracking. Equation (13) uses linear form of ${V_k}(n )- {V_{mid}},$ a polynomial form can also be used such as
where M is an integer no less than 1. In [22], an erfc function was used to prevent the control voltage to go out of its boundary.In principle, only one mPD is required in the configuration that generates the photocurrent iSignal from signal port of the tracker output. However, the sum of the tracker output powers is a constant for a given input optical power, no matter what the input polarization state is. Therefore, detecting the photocurrents from both signal and LO ports, and normalizing iSignal over their sum can eliminate the impact of input optical power fluctuation.
Figure 9 shows the combined S21 response of the oSHD system, including the responses from DAC, driver, modulator, ICR, and DSO. The S21 bandwidth is mainly limited by the DAC response. The RF response has around 14 dB attenuation at 30 GHz and is sharply rolled-off beyond 33 GHz. The S21 response was pre-compensated in transmitter DSP.
Figure 10(a) shows a typical procedure of optical polarization demultiplexing for carrier and signal, which converged within a few milliseconds. The actual required convergence time depends on the input SOP rotation speed as well as the distance of initial SOP state to the target SOP state. In our setup, the convergence time (tracking speed) is mainly limited by the response time of the fiber squeezers and their drivers and could be improved by using a high-speed tracker. After the convergence of optical demultiplexing, the signal and carrier were well separated, and the tracker output power ratio of signal port to LO port was stabilized and converged to the target value of around −10.5dB: the reciprocal of CSPR of the oSHD channel, with negligible variation. Figure 10(b) shows a typical evolution of the control voltages applied to the four fiber squeezers of the polarization tracker. The voltages were confined within the range of [0∼4]V, which is exactly as designed by our control algorithm to avoid hitting the boundary of control voltages, by pulling it back to middle value if any control voltage indeed close to hits the limit. Figure 10(c) shows the long-term tracker output power ratio of signal port to LO port, showing 100 minutes recording. As seen in the figure, the power ratio of the two output ports is stabilized at the reciprocal of the transmitter CSPR with small variations, which are caused by the CW crosstalk from carrier to signal. For high CSPR, crosstalk from carrier to signal is dominant, while crosstalk from signal to carrier is negligible.
Fig. 10. Control of polarization demultiplexing. (a) Output power ratio of signal port to LO port during convergence; (b) Typical evolution of control voltages applied to polarization tracker; (c) Long term output power ratio of signal port to LO port.
Figure 11 shows the extracted maximum crosstalk for each test case versus the number of tests, where each test recorded 70 thousand numbers of loop step. For polarization scrambling with a mean SOP change rate of 10.5 rad/s (maximum SOP change rate of 31.5 rad/s), the maximum extracted crosstalk varies along the test number, but is below −22 dB. As discussed in the Section 4, crosstalk below −22 dB induces negligible power penalty at BER=1.6e-2 for 41Gbuad 64QAM at or above 10 dB CSPR, making the impact of crosstalk from optical demultiplexing negligible for an oSHD system.
Figure 12 shows the optical spectra measured at different stages. Figure 12(a) is the spectrum of a 41Gbaud 64QAM oSHD channel with polarization multiplexed pilot and signal. Figure 12(b) is for demultiplexed pilot (LO) and Fig. 12(c) is for demultiplexed signal. The figures show that the signal and pilot are well separated after polarization tracking. There is a very low level of residual signal in the demultiplexed pilot, which is negligible compared with the pilot intensity.
Fig. 12. Measured optical spectra. (a), oSHD channel with orthogonal polarization between signal and carrier; (b) Optical carrier after polarization demultiplexing; and (c) Optical signal after polarization demultiplexing.
Bit error rate performance was further evaluated to quantify the optical polarization demultiplexing for 61.6Gbaud 16QAM and 41Gbaud 64QAM, both carrying 246Gb/s raw data rate or 200Gb/s net data rate with 11% FEC. Due to the insertion loss of the optical delay line, the actual power ratio of LO to signal into the SP-ICR will be slightly different from the transmitter CSPR. In the experiment, pulse shaping roll-off factor α of 0.3 was used for 41Gbaud 64QAM, providing a LO-to-signal power ratio of 9.6dB, and α of 0.1 was used for 61.6Gbaud 16QAM, providing a LO-to-signal power ratio of 11dB. Figure 13 shows the measured curves of BER versus ROP and 64QAM constellations with the same mean SOP change rate around 10.5rad/s. At the FEC threshold of 1.6e-2, the receiver sensitivity is around −5.3dBm for 61.6Gbaud 16QAM, and −4.3dBm for 41Gbaud 64QAM, respectively. 16QAM modulation shows better sensitivity and BER floor, but 64QAM achieves higher spectral efficiency while still meeting the performance requirement. For comparison, simulation results with the same α and CSPR as the experiment were also included in Fig. 13. Compared with experiment, simulation results show better BER floor and exhibit around 0.35dB better sensitivity for 41Gbaud 64QAM and around 0.4dB sensitivity for 61.6Gbaud 16QAM. Since the pilot does not experience modulation loss and chip absorption loss from the active area of the oSHD transmitter, it is easier to achieve a typical 6dBm transmitter output power at ∼10dB CSPR. As demonstrated by our experiments, a link budget over 10dB is achievable with 64QAM modulation for a 200G oSHD system via optical demultiplexing.
Fig. 13. Measured BER and constellations: (a) BER vs. ROP for 41Gbaud 64QAM and 61.6Gbaud 16QAM; (b) 64QAM Constellations.
To evaluate the stability and reliability of optical polarization demultiplexing, over 500 BER tests with the same SOP rotation speed have been carried out. Results for 41Gbaud 64QAM are shown in Fig. 14. BER exhibits very small variation around 3e-3 at given ROP of −1.5dBm, and small variation around 1e-2 at given ROP of −3.8dBm. The extremely small variation in BER demonstrates stable and endless polarization demultiplexing of the oSHD system.
Performance tolerance to the delay between signal and LO is an important aspect, as large delay tolerance can lose the requirement on fabrication tolerance. Figure 15 shows BER versus delay between signal and LO for a 200G oSHD system with 64QAM for a given ROP of around 0.5dBm. The delay was introduced and varied by the optical delay line, as shown in Fig. 8. As expected, BER performance degrades with the increase of delay, since there is no carrier recovery in the receiver DSP. While within 1ps delay, the BER degradation is negligible. For a silicon integrated polarization tracking oSHD receiver, the fabrication tolerance for the optical path difference between signal and LO is within 1ps.
Another advantage of SHD is its extremely large tolerance to laser linewidth, so conventional DFB lasers rather than narrow linewidth external cavity lasers (ECL) can be used. Figure 16 compares the BER performance between DFB and ECL for 41Gbaud 64QAM. Here the ECL is an Emcore tunable laser with a typical linewidth less than 100kHz. The delay between the signal and LO was adjusted to be close to zero and the scrambler was also operated at the same SOP rotation rate as above. Figure 16 shows that there is no performance difference between the curves of BER versus ROP using DFB and ECL.
The above results demonstrate that optical demultiplexing of a 64QAM 200G oSHD system exhibits stable polarization tracking, good sensitivity performance with over 10 dB link budget, large tolerance to crosstalk, laser linewidth and delay between signal to LO, and is therefore a promising candidate of next generation 800GbE or 1.6TbE.
6. Summaries
We have investigated the self-homodyne detection system via optical polarization demultiplexing of carrier and signal. Compared to digital demultiplexing, optical polarization demultiplexing provides about 0.9dB sensitivity benefit. Both simulations and lab tests show that the scheme is very tolerant to the crosstalk originating from optical demultiplexing, with negligible performance degradation at crosstalk up to −22dB. Experimental verifications demonstrate the stability of polarization tracking achieved by our control algorithm and circuit. 41Gbaud 64QAM oSHD with 200Gb/s net data has been realized by using the proposed scheme and setup, with receiver sensitivity better than −4dBm. This work has proved that a self-homodyne detection system via optical polarization demultiplexing is a potential candidate solution for 800GbE or 1.6TbE.
Disclosures
The authors declare no conflicts of interest.
References
1. J. Lee, S. Shahramian, N. Kaneda, Y. Baeyens, J. Sinsky, L. Buhl, J. Weiner, U. Koc, A. Konczykowska, J. Dupuy, F. Jorge, R. Aroca, T. Pfau, and Y. Chen, “Demonstration of 112-Gbit/s optical transmission using 56 GBaud PAM-4 driver and clock-and-data recovery ICs,” European Conference on Optical Communication (ECOC) 2015, paper Mo.4.5.4.
2. J. Man, W. Chen, X. Song, and L. Zeng, “A low-cost 100GE optical transceiver module for 2 km SMF interconnect with PAM4 modulation,” in Optical Fiber Communication Conference (OFC) 2014, paper M2E.7.
3. M. I. Olmedo, Z. Tianjian, J. B. Jensen, Z. Qiwen, X. Xiaogeng, and I. T. Monroy, “Towards 400GBASE 4-lane solution using direct detection of Multi-CAP signal in 14 GHz bandwidth per lane,” Optical Fiber Communication Conference/National Fiber Optic Engineers Conference (OFC) 2013, paper PDP5C.10.
4. T. Wettlin, S. Ohlendorf, T. Rahman, J. Wei, S. Calabro, N. Stojanovic, and S. Pachnicke, “Beyond 200 Gb/s PAM4 Transmission using Tomlinson-Harashima Precoding,” European Conference on Optical Communication (ECOC) 2019, Paper Tu2B.6, Sept 2019.
5. T. Miyazaki, “Linewidth-tolerant QPSK homodyne transmission using a polarization-multiplexed pilot carrier,” IEEE Photonics Technol. Lett. 18(2), 388–390 (2006). [CrossRef]
6. P. Johannisson, M. Sjodin, M. Karlsson, E. Tipsuwannakul, and P. Andrekson, “Cancellation of nonlinear phase distortion in self-homodyne coherent systems,” IEEE Photonics Technol. Lett. 22(11), 802–804 (2010). [CrossRef]
7. M. Sjodin, P. Johannisson, M. Karlsson, Z. Tong, and P. A. Andrekson, “OSNR requirements for self-homodyne coherent systems,” IEEE Photonics Technol. Lett. 22(2), 91–93 (2010). [CrossRef]
8. B. J. Puttnam, R. S. Luis, J. M. D. Mendinueta, J. Sakaguchi, W. Klaus, Y. Kamio, M. Nakamura, N. Wada, Y. Awaji, A. Kanno, T. Kawanishi, and T. Miyazaki, “Self-homodyne detection in optical communication systems,” Photonics 1(2), 110–130 (2014). [CrossRef]
9. J. Lowery and J. Armstrong, “Orthogonal-frequency-division multiplexing for dispersion compensation of long-haul optical systems,” Opt. Express 14(6), 2079–2084 (2006). [CrossRef]
10. J. C. Schmidt, A. J. Lowery, and J. Armstrong, “Experimental demonstrations of electronic dispersion compensation for long-haul transmission using direct-detection optical OFDM,” J. Lightwave Technol. 26(1), 196–203 (2008). [CrossRef]
11. Y. J. Wen, X. Shen, and Y. S. Bai, “Method and apparatus for directly detected optical transmission systems based on carrierless amplitude-phase modulation,” US patent: US9374260B2.
12. Y. J . Wen, C. Liao, X. Shen, and Y. S. Bai, “Optical transmitters with unbalanced optical sidebands separated by gaps,” US patent: US9071363 B2.
13. A. Mecozzi, C. Antonelli, and M. Shtaif, “Kramers–Kronig coherent receiver,” Optica 3(11), 1220–1227 (2016). [CrossRef]
14. Z. Li, M. S. Erkılınc, K. Shi, E. Sillekens, L. Galdino, B. C. Thomsen, P. Bayvel, and R. I. Killey, “SSBI Mitigation and the Kramers–Kronig Scheme in Single-Sideband Direct-Detection Transmission With Receiver-Based Electronic Dispersion Compensation,” J. Lightwave Technol. 35(10), 1887–1893 (2017). [CrossRef]
15. S. Randel, D. Pilori, S. Chandrasekhar, G. Raybon, and P. J. Winzer, “100-Gb/s discrete-multitone transmission over 80-km SSMF using single-sideband modulation with novel interference cancellation scheme,” In Proceeding of European Conference of Optical Communications (ECOC) 2015, Valencia, Spain Sept 2015.
16. Z. Li, M. S. Erkılınç, S. Pachnicke, H. Griesser, R. Bouziane, B. C. Thomsen, P. Bayvel, and R. I. Killey, “Signal-signal beat interference cancellation in spectrally-efficient WDM direct-detection Nyquist-pulse-shaped 16-QAM subcarrier modulation,” Opt. Express 23(18), 23694–23709 (2015). [CrossRef]
17. D. Che, A. Li, X. Chen, Q. Hu, Y. Wang, and W. Shieh, “160-Gb/s stokes vector direct detection for short reach optical communication,” OFC2014, Paper Th5C.7, Mar 2014.
18. D. Che, A. Li, X. Chen, Q. Hu, Y. Wang, and W. Shieh, “Stokes vector direct detection for linear complex optical channels,” J. Lightwave Technol. 33(3), 678–684 (2015). [CrossRef]
19. J. Perin, A. Shastri, and J. M. Kahn, “Data center links beyond 100 Gbit/s per wavelength,” Opt. Fiber Technol. 44, 69–85 (2018). [CrossRef]
20. J. Perin, A. Shastri, and J. M. Kahn, “Design of Low-Power DSP-Free Coherent Receivers for Data Center Links,” J. Lightwave Technol. 35(21), 4650–4662 (2017). [CrossRef]
21. C. R. Doerr and L. Chen, “Monolithic PDM-DQPSK receiver in silicon,” the 36th European Conference and Exhibition on Optical Communication (ECOC) 2010, paper PD3.6, Sept 2010.
22. A. Nespola, G. Franco, F. Forghieri, M. Traverso, S. Anderson, M. Webster, and R. Gaudino, “Proof of Concept of Polarization-Multiplexed PAM Using a Compact Si-Ph Device,” IEEE Photonics Technol. Lett. 31(1), 62–65 (2019). [CrossRef]
23. T. Gui, X. Wang, M. Tang, Y. Yu, Y. Lu, and L. Li, “Real-time demonstration of 600 Gb/s DP-64QAM self-homodyne coherent bi-direction transmission with un-cooled DFB laser,” Optical Fiber Communication Conference (OFC) 2020, post-deadline paper Th4C.3.
24. L. Moller, “WDM polarization controller in PLC technology,” IEEE Photonics Technol. Lett. 13(6), 585–587 (2001). [CrossRef]
