1. Introduction
Unrepeatered transmission systems provide a low-complexity and cost-effective solution for point-to-point connections on the order of a few hundred kilometers. The main objective for these systems is to achieve a single-span system between the points of interest without any in-line active elements between the terminals. Recently, high-order modulation formats and coherent detection using digital signal processing (DSP) have been introduced to increase the transport capacity. In these systems, the distributed Raman fiber amplifier (DFRA) is an attractive alternative to the conventional lumped erbium-doped fiber amplifier (EDFA). The DRFA ensures a relatively constant power distribution of the optical signals along the link and thereby improves the optical signal-to-noise ratio (OSNR) and the fiber nonlinearity tolerance [
1,
2,
3,
4,
5,
6,
7]. Using DRFA with a large effective area fiber, unrepeatered transmission of a 30 Gbaud dual-polarization quadrature phase shift keying (DP-QPSK) over 444 km [
2] and a 28 Gbaud DP-16 quadrature amplitude modulation (DP-16QAM) over 240 km [
3,
4] have been demonstrated. Unrepeatered transmission of a 40 × 256 Gb/s DP-16QAM signal over an ultra-low loss fiber with a very large effective area was demonstrated by using a first-order backward-pumped DRFA combined with a discrete EDFA at the receiver [
5]. Transmission performance analysis was performed for differential phase shift keying (DPSK) transmission using standard fiber and second-order Raman amplification in [
6], showcasing the different balance between impairments for different transmission lengths and higher-order distributed amplification. Recently, an unrepeatered Nyquist PDM-16QAM transmission over 364 km was demonstrated using Raman amplification and multi-channel digital back-propagation [
7].
In DRFAs, the intensity fluctuations of the pump will induce both intensity and phase noise of the received signal. This is known as relative intensity noise (RIN) transfer which deteriorates the system performance and limits the system reach. The effect of RIN transfer on the intensity noise of the received signal has been intensively investigated in intensity modulation direct-detection systems [
8,
9,
10]. The results show that the characteristics of the RIN transferred to the signal are strongly affected by different aspects, for example fiber dispersion, pumping configuration, and polarization-dependent gain [
11,
12,
13,
14,
15]. The RIN transfer to both the amplitude and phase noise of multilevel coherent optical communication systems using DRFA has been studied using numerical analysis and system simulations, and evaluated with both unrepeatered and repeatered transmission experiments [
16,
17,
18,
19,
20,
21]. The stochastic intensity fluctuations of the Raman pump laser induce extra phase noise of the signal due to the pump signal cross-phase modulation. This is known as relative phase noise (RPN) and it degrades the system performance of phase-sensitive detection schemes. For example, the analytical model for the statistical property of RPN in first- and higher-order distributed Raman amplifiers [
15] indicates that the impairments dependent on the Raman pumping order and RIN tolerance are different for each configuration. The study also shows that for the multi-level modulation format, in addition to RIN and RPN, Raman pump RIN will also induce polarization crosstalk through the cross-polarization modulation effect which has about the same impact on the performance as the RPN. A comprehensive theoretical and simulation study of the RPN was presented for M-ary phase shift keying (PSK) and 16QAM modulation formats in Reference [
18]. The RPN-induced impairment was evaluated and compared for QPSK, 8PSK, 16PSK, and 16QAM signals. We have previously experimentally characterized the pump noise–induced impairments of an unrepeatered single polarization of a 28 Gbaud 16QAM link using backward-pumped Raman amplification [
22], and for a dual-polarization 28 Gbaud 16QAM link [
23] using bidirectionally pumped Raman amplification.
In this paper we report on a systematic experimental characterization of both the amplitude and phase noise of the signal due to RIN transfer from the Raman pump and evaluate the transmission impairments of a coherent DP-16QAM link using first-order pumped Raman amplification in a standard single-mode fiber (SSMF). We adopt a frequency domain representation approach to reconstruct the noise enhancement due to the Raman pump RIN transfer in both amplitude and phase. Regarding the transmission, we first evaluate the maximum reach, determined by the 7%-overhead hard-decision forward error correction (HD-FEC) threshold, of a backward pumping scheme. We subsequently investigate the transmission improvement one can achieve with the aid of additional forward pumping. For each pumping scheme, the phase noise and amplitude noise are measured as a function of the Raman pump power. Three different carrier phase recovery (CPR) algorithms are tested and compared for each transmission scenario: decision-directed phase-lock-loop (DDPLL) [
24], two-stage feed-forward based on QPSK partitioning (QPSKP) with a sliding window average [
25] and feed-forward based on blind phase search (BPS) [
26,
27].
This paper is organized as follows. In
Section 2, we present the experimental setup to evaluate DFRA-induced impairments. Then we show and discuss the obtained results in
Section 3. The conclusions are finally drawn in
Section 4.
2. Experimental Setup
The experimental setup for signal transmission and noise characterization is shown in
Figure 1. For transmission, a pseudo-random bit sequence with word length of 2
15 − 1 and its negated pattern are generated by a pulse pattern generator (PPG) operating at 28 Gbaud. Decorrelation is performed by delaying the negated pattern by 84 bits. A two-bit digital to analog converter (DAC) with 19 GHz analog bandwidth is employed to convert the binary sequences into four-level pulse amplitude modulation signals, where an additional 77-symbol delay is introduced between the two outputs to drive an IQ modulator with 25 GHz 3 dB bandwidth to generate an optical 28 Gbaud 16QAM. The transmitter external cavity laser (ECL) at 1550 nm has a linewidth of ~40 kHz. Polarization multiplexing (PolMux) emulation is performed after the IQ modulator to achieve 224 Gbit/s DP-16QAM. An optical booster, an optical bandpass filter (OBPF) with bandwidth of 1.2 nm and a variable optical attenuator (VOA) are placed before the fiber link for OSNR adjustment.
The transmission link is comprised of 200 km SSMF. First order pump at 1455 nm (PYL-1-1455-R, IPG Photonics, Oxford, MA, USA) is used to Raman amplify the signal which is divided into two parts 95% (backward) and 5% (forward) for distributed Raman amplification. An automatic gain controlled pre-amplifier EDFA with fixed output power is placed after the link. A coherent receiver with a ~40 kHz linewidth local oscillator (LO) and an 80 GSa/s digital storage oscilloscope (DSO) are used for detection. Signal demodulation is performed off-line with a DSP chain consisting of low pass filtering; chromatic dispersion (CD) compensation, timing recovery, polarization de-multiplexing and three different CPR methods before a bit-error-rate counter.
For a better understanding of the signal impairments induced by the distributed Raman amplification, we also characterized the amplitude and frequency noise of the received signal with the use of the digital coherent receiver based on the method described in [
28]. In this measurement the transmitter laser is unmodulated and transmitted over the Raman amplified link with different pump configurations.
3. Results
The OSNR improvement using distributed Raman amplification was first evaluated.
Figure 2 illustrates the measured OSNR versus the power of the first-order Raman pump for two pumping schemes, first using backward pumping with 95% of the pump power and second using bidirectional pumping, using the remaining 5% pump power for co-pumping. The experiments were done at two launched powers, 5 and 2 dBm, into the 200 km SSMF link. From
Figure 2 it is observed that with 95% backward pumping, the received signal OSNR increased with the pump power compared to the reference case when the Raman pump was off. In addition, with the extra 5% of the power in the forward pumping, the OSNR enhancement was higher, especially at high power. It is important to note that the maximum forward pump used was only 40 mW.
In coherent optical communication systems using multi-level modulation formats, a high OSNR only is not enough to ensure a good transmission quality since other impairments such as the phase noise have a large impact on the system performance [
18]. Therefore, we characterized the received signal amplitude and phase noise induced by the Raman pump RIN transfer.
Figure 3 shows the measured relative frequency noise and amplitude noise power spectrum density (PSD) of the received signal after bidirectional Raman amplification for different pump powers. All selected cases are with a 200 km SSMF length and a 5 dBm launched power. In this measurement, the transmitter laser was unmodulated and transmitted over the Raman amplified link with different pump configurations, as shown with the green dots in
Figure 1. It is noted that this method characterizes the combined noise properties from both the received signal and the narrow linewidth LO. From the frequency noise PSD shown in
Figure 3a, one can see that with a low Raman pump power, the noise shows a property that can be closely approximated as white phase noise. However, at a high pump power, the frequency noise level in the low frequency regime rises, while at the high frequency side the level decreases, which is due to the improvement of the signal OSNR (as shown in
Figure 2). It is noted that there are spurs rising at around 11 GHz at the high pump power, which we attribute to the stimulated Brillouin scattering.
Similarly, from
Figure 3b it is observed that as the pump power increases, the amplitude noise level at the high frequency side decreases, which is also due to the improvement of the signal OSNR. However, at the lower frequency regime a rise of the noise level at a higher pump power is observed. This noise enhancement can be attributed to the RIN transfer from the Raman pump. It is noted that the frequency regime of the enhanced noise is important for a potential effective mitigation by optical or digital means, as it defines the mitigation bandwidth of the optical or digital filter.
The Q factor (calculated from the measured bit error rate) is presented in
Figure 4 as a function of the Raman pump power. The results for 200 km at a launched signal power of 5 and 2 dBm with the two different pumping schemes were taken after demodulation by the DDPLL, QPSKP and BPS algorithms, respectively. From the results it can be seen that with the increase of the pump power, the performance firstly improves due to the enhanced OSNR as shown in
Figure 2. The optimum was found at around 550 mW in all cases.
Figure 5 shows the constellations of DP-16QAM with the bidirectional pumping scheme for different pump powers. It is noted that in the constellations with a high pump power, nonlinear distortions are not clearly observed. The signal suffers from the induced amplitude and phase noise as shown in
Figure 3, indicating that the induced noise dominates the performance despite the high OSNR values at the higher pump power. Consequently, unrepeatered coherent transmission should be optimized by balancing the signal OSNR and the Raman pump–induced impairments. By comparing the backward and bidirectional pumping schemes, we could observe a clear performance improvement around the optimum pump power, enabling the bidirectional pumping scheme to achieve a Q factor above the 8.5 dB HD-FEC limit after 200 km SSMF unrepeatered transmission. Moreover, it was observed that the BPS slightly outperformed DDPLL and QPSKP in mitigating the Raman-induced impairments, thanks to the accurate phase tracking capability at the high OSNR [
27]. One should note that due to the experimental constraints, the power ratio between the forward and backward pump was fixed during this work; however, it is worth further investigation as an important aspect to draw a more generic conclusion. Additionally, a longer fiber span, higher modulation orders, as well as higher-order pumping schemes are to be investigated in our further studies.
4. Conclusions
We have experimentally characterized the induced amplitude and phase noise due to the RIN of the Raman pump and studied their impact on the system performance of an unrepeatered 224 Gbit/s DP-16QAM link. The results show that a Q factor above the 8.5 dB HD-FEC limit after 200 km SSMF can be achieved with a bidirectional pumping scheme, using an optimized pump power and a small fraction (5%) of the pump in the forward direction. At low pump powers, additive noise due to a low OSNR is the dominant noise source, while at high pump powers, the induced noise from the Raman pump dominates. Additionally, the noise enhancement in both the amplitude and phase was observed to concentrate on the low-frequency regime, while the high-frequency regime was not as influenced. This indicates that the system performance may be further improved at high pump powers by suppressing the RIN transfer or by using advanced higher-order pumping schemes, or with certain digital mitigation methods focusing on low-frequency noise suppression. Further experimental studies for various types of impairments and their balance is required for the optimal design of the advanced multi-level modulation format signal–based unrepeatered transmission systems using higher-order DRFA. On the other hand, RPN impairment mitigation with more powerful DSP algorithms is worth investigating.