Open Access
Add to BibSonomyAbstract
We experimentally demonstrate a Raman-Assisted Fibre Optical Parametric Amplifier (RA-FOPA) with 20dB net gain using wavelength division multiplexed signals. We report amplification of 10x58Gb/s 100GHz-spaced QPSK signals and show that by appropriate tuning of the parametric pump power and frequency, gain improvement of up to 5dB can be achieved for the RA-FOPA compared with combined individual contributions from the parametric and Raman pumps. We compare the RA-FOPA with an equivalent-gain conventional FOPA and find that four-wave mixing crosstalk is substantially reduced by up to 5.8 ± 0.4dB using the RA-FOPA. Worst-case performance penalty of the RA-FOPA is found to be only 1.0 ± 0.2dB over all measured OSNRs, frequencies and input powers, making it an attractive proposal for future communications systems.
© 2015 Optical Society of America
1. Introduction
The continuing increase of fibre optical signal data rates and spectral density is resulting in systems rapidly approaching capacity limits for those windows of the spectrum where low transmission loss and easily-accessible optical gain co-exist (i.e. the ‘C’ and ‘L’ bands) [1]. Commercial optical amplification technology in these bands has been dominated for >30 years by the erbium doped fibre amplifier (EDFA) which limits gain in wavelength division multiplexed (WDM) systems to a combined C/L bandwidth of ~10THz (80nm). In order to help further increase the information-carrying capacity of a single mode optical fibre, research is ongoing into alternative fibre-based optical amplifiers which can provide useable gain in other spectral regions as well as traditional C/L bands.
One research approach is the use of fibre dopants other than erbium to provide the gain – for example thulium has recently been demonstrated within a silica-based host fibre to provide >20dB gain between 1750 and 1950nm [2]. These amplifiers show very promising performance, in particular a significant improvement in operating bandwidth, but lack some potentially useful features such as the ability to tune the operating wavelength region. Another approach is the (discrete or distributed) Raman amplifier which can provide polarisation-independent gain across a bandwidth of ~3THz using a single depolarised pump, and which provides a gain spectrum position defined purely by the pump frequency [3]. The Raman gain spectrum can easily be extended in range by adding additional frequency-offset pump lasers, but at the expense of pump-pump interactions requiring a complex pump control algorithm to obtain flat gain [4]. Other limitations of the Raman amplifier include the comparatively low Raman gain coefficient which restricts the magnitude of the gain to 12-14dB in standard single mode fibre (assuming distributed amplification) [5], and the effect of double Rayleigh backscatter which causes a multi-path interference degradation of signal quality [6]. The Fibre Optical Parametric Amplifier (FOPA) is another actively-researched amplification scheme which employs a phase-matched four-wave mixing process between pump(s) and signal(s) in a highly nonlinear fibre (HNLF) to provide gain more efficiently than via the Raman effect [7]. By careful combination of the HNLF properties (dispersion profile, nonlinear parameter and length) together with pump power and pump frequency tuning, gains as high as 70dB and gain bandwidths of ~200nm have been demonstrated [8, 9]. The FOPA has also been shown operating in polarisation diverse arrangements which has allowed amplification of polarisation-multiplexed signals [10]. However, FOPA performance is also highly dependent on pump quality and can be compromised via instantaneous transfer of pump fluctuations such as relative intensity noise (RIN) to the amplified signals [11], as well as by the generation of detrimental four-wave mixing crosstalk terms when amplifying multiple WDM channels [12].
Recently, the Raman-Assisted (RA)-FOPA has been proposed and studied experimentally, analytically and numerically by a number of groups in order to potentially employ the best properties of both schemes. The RA-FOPA consists of a forward travelling parametric pump within an HNLF which is ‘augmented’ by a backward travelling Raman pump, usually in the same fibre [13]. The Raman pump provides gain not only to the signals but also to the parametric pump and, through careful selection of their respective frequencies, can widen/flatten the gain spectrum as well as provide enhanced gain over and above the linear sum of the individual contributions in certain regions of the spectrum [13,14].
To date, RA-FOPAs have typically been experimentally characterised using a single continuous wave (CW) probe to measure the gain and noise figure in long lengths of HNLF (~1km). A notable exception, where the performance was examined using two 10Gb/s non-return-to-zero (NRZ) data signals, showed a reduced susceptibility to signal gain saturation [15]. It is recognized that the long HNLF length helps maximize the Raman gain, but at the expense of an increased interaction length, resulting in buildup of unwanted four-wave mixing crosstalk products [16].
In this paper we experimentally investigate for the first time the performance of a RA-FOPA when amplifying 10x58Gb/s 100GHz-spaced QPSK signals for an average net gain of ~20dB over the band, split approximately equally between Raman and parametric contributions. We use a short (200m) length of HNLF and compare the RA-FOPA performance directly with a standard FOPA at the same level of gain and find that the RA-FOPA offers significantly reduced inter-channel crosstalk levels and lower performance penalty for the same received optical signal to noise ratio (OSNR).
2. Experimental set-up
The experimental set-up is shown in Fig. 1. The transmitter consisted of ten 100GHz-spaced standard distributed feedback (DFB) lasers ranging from 193.5 to 194.4THz and multiplexed together in a polarisation-maintaining arrayed waveguide grating (PM-AWG). The DFB frequencies were combined with a 100kHz-linewidth laser which was used as the signal under test whereby the relevant DFB laser could be turned off and the 100kHz laser tuned to fill the spectral gap. The ten continuous wave (CW) frequencies were IQ-modulated in a nested Mach-Zehnder modulator using two 29Gb/s decorrelated 231-1 bit streams, producing 10x58Gb/s single polarisation QPSK signals. The signals were transmitted through 5.4km of standard single-mode fibre (SSMF) to provide data decorrelation. The per-signal input power was adjusted using a variable optical attenuator (VOA) to either −20dBm or −12dBm, measured at the output of the VOA (indicated as point A). This was considered to be the input to the black-box RA-FOPA subsystem.
The RA-FOPA subsystem consisted of a parametric pump (PP) from a 100kHz CW laser emitting at 191.57THz (fP) which was spectrally broadened using a lithium niobate phase modulator (Vπ = 5.2V) to provide mitigation against stimulated Brillouin scattering (SBS) [17]. The electrical drive to the phase modulator was a comb of three electrical tones at 100MHZ, 320MHz and 980MHz. The PP was amplified using a high-power EDFA and optically filtered to remove amplified spontaneous emission (ASE) using a circulator and reflective apodised fibre Bragg grating (FBG) centred at fP and with a 3dB bandwidth of 112GHz. The QPSK signals were combined with the PP by transmitting the signals through the FBG, which had <0.3dB insertion loss at the signal frequencies. The PP and signals were then launched into 200m of HNLF after first being passed through a fused fibre 1455/1550nm WDM combiner (WDM1) to filter out backward travelling Raman pump light. Polarisation controllers (PC1 and PC2) were used to align the pump and signal polarisations and the input PP power and SBS back-reflected power were monitored using 1% tap couplers and optical power meters (OPM) calibrated for loss. The HNLF was commercially-sourced and was designed to have low (<1nm) variation of the zero dispersion wavelength (ZDW) with length. The HNLF insertion loss was 0.9dB, ZDW 1564nm, dispersion slope 0.084ps/(nm2.km) and nonlinear parameter 8.2(W.km)−1.
At the HNLF output, a second 1455/1550nm fused WDM combiner (WDM2) was used to combine a backward-travelling (with respect to the signals) Raman pump laser at 1455nm (fR = 206THz). The Raman pump (RP) source was a depolarised fibre laser with maximum output power of 5.6W (37.5dBm) and 3dB linewidth of ~225GHz (1.6nm).
Following WDM2, the PP and L-band idlers were removed using an optical high pass filter (HPF), and the 10x58Gb/s signals sent to a receiver subsystem. Total passive insertion loss of the RA-FOPA was only 3.5dB (including HNLF), measured between points A and B in Fig. 1. At the receiver, the signal under test was optically demultiplexed using a 100GHz 3dB bandwidth bandpass filter (BPF) and optically amplified to a fixed power of 10dBm using an EDFA in constant power mode. For performance comparison between different signal frequencies and amplification schemes, the received optical signal to noise ratio (OSNR) was adjusted by combining the signal under test with unpolarised broadband ASE and the spectrum captured using an optical spectrum analyser (OSA). The OSNR was subsequently calculated relative to a noise bandwidth of 0.1nm (12.5GHz).
For signal quality and bit error rate (BER) counting, the signal under test was mixed with a 12dBm, 100kHz linewidth local oscillator (LO) in a standard coherent receiver with balanced detection. The electrical traces were captured using an 80GSample/s, 36GHz oscilloscope. Data was processed off-line using standard digital signal processing (DSP), including a Viterbi 4th-power phase recovery scheme with block length set to 21 symbols [18]. Performance was characterised from the recovered constellation using the error vector magnitude (EVM) quality-factor [19], averaged over 57kSymbols. At low OSNR (<15dB), full error counting was also performed over 570kSymbols to provide a consistency check with a BER derived from the EVM (see Section 5).
3. RA-FOPA WDM gain spectra
For initial optimisation of the RA-FOPA, output spectra were recorded with the PP turned off and the RP input power to the HNLF varied. The on-off Raman gain was then calculated for each of the signal peaks by subtracting the relevant reference power with both pumps turned off. This is plotted in Fig. 2(a) for a per-signal input power of −20dBm. It can be seen that the signal gain increases with RP power as expected, with a maximum average gain over the ten signals achieved of 11.5dB at 37dBm (5W) pump power. The Raman gain per unit pump power is found to be constant at ~2.3dB/W, with the gain value at maximum pump power limited by the 200m length of HNLF. The gain ripple/gain slope values of the spectra increased from 0.2dB/0.1dB at ~1dB average gain to 0.65dB/-0.5dB at ~5dB average gain and 1.6dB/-1.4dB at maximum gain. Gain ripple is defined here as Gainmax-Gainmin, and gain slope as the vertical deviation of a linear regression trendline plotted through all ten signal gains.
Fig. 2 (a) Graph showing the 10x58Gb/s signal-peak on-off gain variation against frequency measured at the RA-FOPA output (OP) for different indicated RP powers (PP off). The per-signal input (IP) power was −20dBm. (b) OP spectra of the RA-FOPA at 12.5GHz resolution bandwidth for a per-signal IP power of −20dBm. In order of OP power, the spectra are: both pumps off (green trace); RP only at 36dBm IP power to HNLF (blue dashed trace); PP only at 33dBm IP power to HNLF (red trace); both pumps on simultaneously at 36dBm and 33dBm respectively (black trace). (c) Graph showing signal-peak on-off gain variation against frequency for per-signal input power of −20dBm and same pump powers for: RP only (blue squares); PP only (red triangles); the sum of the individual RP only and PP only gains (purple crosses); both pumps on (black diamonds). (d) As for (c) but with per-signal input power increased to −12dBm.
For the subsequent work described in this paper, the RP was set at 36dBm, providing an average Raman-only gain of 9.5dB with gain ripple of 1.2dB and slope of −1.1dB (blue trace in Fig. 2(a)).
The individual and combined spectral effects of the PP and RP are illustrated in Fig. 2(b) at the output of the RA-FOPA for the −20dBm per-signal input power case. The RP power was maintained at 36dBm and the PP fine-tuned in terms of input power and frequency in order to obtain as flat gain as possible whilst achieving an average (over all signals) net-gain of 20dB (23.5dB on-off gain). A PP power of 33dBm (2W) and frequency of 191.57THz (1564.92nm) was found to be optimum for this (black trace). With the RP on and PP off (blue dashed trace), the unabsorbed backward RP power (measured at point C in Fig. 1) was found to be 35dBm. This dropped significantly to 26.7dBm when the PP was turned on which is attributed to considerable Raman gain of the PP within the HNLF [15]. It can additionally be seen from Fig. 2(b) that with the PP on and the RP off (red trace), the OSNRs of all the signals were degraded significantly compared with the RP-only case. This is attributed to non-optimised matching of the parametric gain spectrum with the signal frequencies under test (rather than for the combined RA-FOPA which was matched) – indeed the output OSNR can be seen to improve considerably when both pumps were turned on in RA-FOPA mode.
Figures 2(c) and 2(d) show the signal-peak on-off gain for the same pump conditions as above and for per-signal input powers of −20dBm and −12dBm respectively. It can clearly be seen that the resultant gain profiles are extremely similar for the two signal powers, with very little gain saturation seen in the high power case (<0.5dB) even though each signal is amplified to an output power of + 8dBm (much higher than signal power levels typically employed in long-haul transmission systems). These plots also show that the overall gain spectrum of the RA-FOPA is considerably more than the “sum of its parts”. In Fig. 2(c) for example, the PP-only plot shows a considerable drop in gain from ~14dB to ~10dB for signals higher than 194.0THz. The crosses show the sum of the individual PP-only and RP-only gains, producing a gain ripple of 5.7dB and a gain slope of −4.9dB over the signal band. With both pumps switched on, the gain ripple is actually found to be 2.1dB with gain slope of + 1.5dB – a significant reduction, and purely due to the amplification of the PP along the HNLF length via Raman amplification which modifies the parametric (and hence combined) gain spectrum. This effect has been reported previously in the literature [13,14], but to the best of our knowledge has not been demonstrated before in a WDM context.
4. RA-FOPA vs standard FOPA – gain flatness and crosstalk
It is instructive to compare the performance of the 20dB net-gain RA-FOPA described above with that of a conventional FOPA under the same input power and gain conditions. Note again that it was not possible to produce 20dB net-gain using pure Raman-only amplification with this length of HNLF for direct comparison. For the 20dB FOPA-only case, the RP was turned off and the PP input power increased to 34.2dBm (2.6W) and the PP pump frequency tuned slightly to 191.55THz (1565.12nm). Figure 3(a) shows the signal-peak gain variation produced for the two amplifiers at a per-signal input power of −12dBm.
Fig. 3 (a) Graph showing the variation in signal-peak on-off gain against frequency for the RA-FOPA (RP = 36dBm, PP = 33dBm, black squares) and the FOPA-only (PP = 34.2dBm, red triangles) amplifiers. Per-signal input power was −12dBm and the average net gain 20dB for both, with ten signals amplified. (b) Output spectra of the RA-FOPA (black traces) and the FOPA-only (red dashed traces) at 12.5GHz resolution bandwidth with −12dBm per-signal input power and 20dB average net gain. Three traces are plotted for each amplifier with one of the signals at 193.5THz, 194.0THz or 194.4THz removed to show the crosstalk level. The crosstalk displayed is therefore that resulting from only nine signals.
It can be seen that the signal gain profile is very similar for the two amplifier types, which is surprising given the considerable differences between the two pumping mechanisms and the difference in PP power evolution. The gain ripple is seen to be marginally smaller in the RA-FOPA case (1.9 vs 2.3dB) but measurement error of this is estimated at ± 0.2dB.
The spectral traces of Fig. 3(b) however clearly show a significant difference between the two amplifier-types in terms of the level of nonlinear crosstalk. The spectra were produced by sequentially turning off representative signals 193.5THz, 194.0THz and 194.4THz (i.e. leaving nine signals in total) and recording the spectrum each time for both the RA-FOPA and the FOPA-only. In the FOPA-only case (red dashed traces), four-wave mixing (FWM) products of high magnitude are conspicuously produced at the high frequency side of the spectrum, resulting from phase-matched interactions between different combinations of PP and signal photons. This crosstalk is a significant barrier to the overall performance and potential usefulness of FOPAs in WDM systems. The RA-FOPA (solid black traces) shows a considerable reduction of these products and hence improvement in actual OSNR and effective noise figure of the amplifier. The comparison is summarised in Table 1. It should also be noted that for an input power of −20dBm per signal, FWM products were not visible above the amplifier noise floor for either FOPA-only or RA-FOPA.
Table 1. Reduction of unwanted FWM crosstalk for RA-FOPA vs FOPA-only at −12dBm per signal input power and 20dB net-gain. Peak to crosstalk delta measured with ten signals for signal peak and nine signals for crosstalk peak.
It can be seen from Table 1 that the RA-FOPA offers reduced crosstalk for all three of the signals tested across the amplification band and most significantly at 194.4THz where the reduction is 5.8 ± 0.4dB. The improvement is most likely even larger in actual operation as the crosstalk peak level measurement only includes the contribution from nine neighbouring signals in order to take the measurement. One other point to notice from Fig. 3(b) is the variation in the crosstalk level shown at 193.5THz as the three signals are sequentially removed. The max-min variation is larger for the FOPA-only case at 2.6dB compared with 1dB using the RA-FOPA.
These results suggest that due to the smaller proportion of parametric gain contributing to the total gain, the RA-FOPA provides a reduced opportunity for unwanted FWM to both grow and to be influenced in magnitude by neighbouring channels (a possible byproduct of the restricted growth). For a FOPA operating with exponential gain, the approximate signal-signal FWM crosstalk level for N symmetrically distributed signals can be estimated as [7]:
where NFWM is (3/4)N(N-1) + 1 and Psigout is the output signal power. With the Raman contribution ignored, Eq. (1) predicts a reduction in crosstalk of ~18dB when comparing the 33dBm PP power used in the RA-FOPA over the PP power of 34.2dBm used in the FOPA-only case (assuming 10dB higher signal output power in the latter). If the extra gain contribution of the Raman in the RA-FOPA is also included, this 18dB reduction would be further lowered towards the experimentally observed value due to amplification of the parametric pump from Raman gain causing an increased amount of parametric mixing and crosstalk. Equation (1) is a simple model and an exact link with the experimentally observed behaviour requires a complete and broadband model of the hybrid amplifier.5. RA-FOPA vs standard FOPA – signal quality performance comparison
In this section we present results comparing the bit error rate (BER) and error vector magnitude (EVM) performance of the RA-FOPA and the standard FOPA, thus allowing us to assess the impact of physical time-domain degradations such as pump-signal intensity fluctuation transfer [20,21] which static CW measurement techniques (e.g. standard EDFA noise figure measurements [22]) cannot.
The amplified QPSK signals were coherently detected and real-time oscilloscope traces captured and processed offline for analysis. In the interests of processing time, this limited the minimum BER that could be calculated via error counting to ~1x10−4 which equates to EVM = 0.269 or EVMdB = −11.4dB when calculated as 20log10(EVM). To assess performance at higher OSNR when no errors were produced, the quality of the QPSK constellation was characterised using an EVM value derived from the recovered symbol IQ vectors [19]. To gain some confidence in this as a figure of merit, we compared performance at low OSNR between the BER derived from this EVM (assuming Gaussian noise) and the actual BER (calculated after counting at least 100 errors). Examining the signal at 194.0THz after the 20dB FOPA-only amplifier, and with a measured received OSNR of 14.5dB, we recorded 113 errors out of 287844 bits giving an actual BER of 3.9e-4 (EVMBER = −10.6dB). From the same data, the estimated EVMEST from the sample vectors was found to be −10.5dB giving an estimated BEREST of 4.3e-4 which is excellent agreement. In practice, the dominating noise at high OSNRs after amplification may not be completely Gaussian (e.g. if it is due to PP phase modulation) but as a relative measurement of penalty at a particular OSNR, we believe EVM is appropriate.
In Figs. 4(a) and 4(b), the EVMdB is plotted against received OSNR for the RA-FOPA (black traces) and FOPA-only (red dashed traces). Back-to-back (B2B) performance was found to be identical across the frequency band and is also plotted as a reference (green trace). Figures 4(a) and 4(b) also indicate the equivalent log10(BER) on the secondary axis. This is calculated from the linear EVMlin using Eq. (2) and is included as an indication of the estimated BER for those unfamiliar with EVM and is not a direct measurement. Note that this secondary axis is non-linear.
where erfc is the complementary error function and sqrt is the square root. Results are presented for the central and two edge signals (193.5, 194.0 and 194.4THz) and thus show representative performance across the amplified band. Figure 4(a) is for a per-signal input power of −20dBm and Fig. 4(b) is for −12dBm. At low received OSNR, the EVM of all amplified signals can be seen to converge with that of the back-to-back (B2B) transmitter-receiver indicating that the EVM here is limited by the added ASE noise, rather than signal degradation. At higher OSNR, the signal EVM diverges from the B2B by different amounts for the different amplifier types and for different signal frequencies. It can be seen in both Figs. 4(a) and 4(b) that the FOPA-only amplifier sees increased divergence compared with the RA-FOPA, especially at the high frequency end of the spectrum. This divergence is important because optical amplifiers must function well over a wide range of output OSNR conditions if used in transmission systems. Figures 4(a) and 4(b) also include inset constellation plots for the 194.4THz QPSK signals received after the two amplifier types at 17dB OSNR and 27dB OSNR respectively. It can be seen that as expected there is no visible difference between the constellations at 17dB OSNR, but at 27dB OSNR there is a noticeable increase in phase-noise visible for the FOPA-only constellation in comparison to the RA-FOPA, thus explaining the relative penalty seen.Fig. 4 (a) Graph showing EVM (dB) against OSNR (dB) for the RA-FOPA (black symbols, solid trendline) and FOPA-only (red symbols, dashed trendline) operating with 20dB net-gain and −20dBm per signal input power. Curves are shown for three representative signals across the amplified band at 193.5THz (squares), 194.0THz (triangles) and 194.4THz (circles). Back to back performance is shown for comparison (green diamonds). Inset constellations are for 17dB OSNR and 194.4THz for RA-FOPA and FOPA-only. Equivalent estimated BER is shown on the secondary axis. (b) As for (a) but with an input power of −12dBm per signal. Inset constellations are for 27dB OSNR and 194.4THz for RA-FOPA and FOPA-only.
Figures 5(a) and 5(b) summarise the EVM penalty against frequency for the RA-FOPA (dashed green) and FOPA-only (solid blue) at −20dBm and −12dBm per signal respectively. Penalties are shown for 27dB, 19dB and 13dB OSNR and the experimental error in the measurement was estimated to be ~ ± 0.2dB. In the −20dBm case of Fig. 5(a), performance between the two amplifiers is therefore judged to be the same at 13dB and 19dB OSNR. At 27dB OSNR, performance of the FOPA-only at 194.4THz is markedly worse than the RA-FOPA. At the higher input power of −12dBm as shown in Fig. 5(b), at 27dB OSNR the performance deteriorates further for the FOPA-only at both 194.0THz and 194.4THz. Performance is also worse here for the FOPA-only at a reduced OSNR of 19dB OSNR as can be seen for the signal at 194.4THz. Overall, the performance penalty of the RA-FOPA is seen to be remarkably constant with signal frequency and input/output power.
Fig. 5 (a) Graph showing EVM penalty (dB) against frequency (THz) for the RA-FOPA (solid green symbols, dashed trendline) and FOPA-only (blue/white symbols, solid blue trendline) operating with 20dB net-gain and −20dBm per signal input power. Penalty curves are shown at 27dB OSNR (squares), 19dB OSNR (diamonds) and 13dB OSNR (triangles). (b) As for (a) but with an input power of −12dBm per signal.
The worst-case penalty over the three representative signals reduced from 1.8 ± 0.2dB in the FOPA-only to 1.0 ± 0.2dB in the RA-FOPA over both input powers. It is tempting to correlate the superior RA-FOPA performance with the lower observed FWM crosstalk levels shown in Fig. 3(b) and tabulated in Table 1. However this may not be a direct cause as no mixing products were seen above the system noise floor at the lower power for either amplifier and yet performance was still better in the RA-FOPA. This does not however rule out that the EVM improvement of the RA-FOPA may be related to the reduction in crosstalk in the sense that they could both derive from a ‘parent’ physical effect. Alternatively, the two may be unrelated and the EVM improvement might derive from some other mechanism – for example, it is possible that under the lower parametric gain conditions of the RA-FOPA there may be smaller time-domain gain variations imposed on the signals from pump phase and intensity variation (“dither and/or RIN transfer”). This could explain the increased phase noise seen in the 27dB OSNR FOPA-only constellation of Fig. 4(b). Further investigation is needed to verify the exact cause of the RA-FOPA performance improvement in what is a highly complex set of wave interactions.
6. Summary
An RA-FOPA employing forward parametric pumping and backward Raman-pumping in the same HNLF has been experimentally demonstrated for the first time in a WDM optical communications context. The RA-FOPA has been characterised at a net gain of 20dB in terms of its gain spectra, four-wave mixing crosstalk level and signal quality penalty when amplifying 10x58Gb/s 100GHz-spaced QPSK signals at two different input powers (−20dBm and −12dBm per signal). The performance has also been directly compared with a standard FOPA operating under the same conditions. It was found that the RA-FOPA offered equal or better performance to the standard FOPA in all measurements undertaken:
a) Gain ripple at 20dB net-gain was marginally lower for the RA-FOPA at 1.9dB vs 2.3dB, albeit with measurement error estimated at ± 0.2dB. It is expected that this difference will widen for larger gain bandwidths.
b) Detrimental FWM crosstalk was significantly reduced for the RA-FOPA. At 194.4THz (the high frequency edge of the signal band), the RA-FOPA showed a reduced crosstalk level of 5.8 ± 0.4dB (relative to the signal peak) compared with the equivalent gain FOPA. We believe that this improvement arises from the reduced parametric pump (PP) input power of the RA-FOPA (33dBm) compared with the standard FOPA (34.2dBm). The PP receives Raman gain as it propagates and therefore increases in power through the length of the HNLF [15]. This can result in proportionally less high power pump-signal parametric interaction compared with the standard FOPA as long as the PP does not grow excessively. A trade-off therefore exists between the HNLF properties (length, loss, dispersion and nonlinearity) and the parametric and Raman pump powers launched into it in determining the amount of signal gain vs FWM crosstalk levels. This is of much interest and will be examined in future work.
c) The worst-case EVM penalty seen over a 14dB output OSNR range was 1.0 ± 0.2dB for the RA-FOPA compared with 1.8 ± 0.2dB for the FOPA. For the RA-FOPA, the penalty at a particular OSNR remained constant with both signal frequency and input power whereas it degraded in the FOPA as these parameters were increased. For low OSNR operation, the performance of both amplifiers converged with that of the ASE-loaded back-to-back system. The improved EVM performance of the RA-FOPA may be related to the same mechanisms which cause the reduced FWM crosstalk, but could also be explained by the approximately exponential dependence of parametric gain on pump power close to the gain peak [7] - the lower RA-FOPA pump power would then likely offer a reduced susceptibility to pump modulation transfer for the same pump modulation-index [20,21].
Acknowledgment
This work was funded by the UK EPSRC under grant EP/J009709/2.
References and links
1. R. Essiambre, G. Kramer, P. J. Winzer, G. J. Foschini, and B. Goebel, “Capacity limits of optical fiber networks,” J. Lightwave Technol. 28(4), 662–701 (2010). [CrossRef]
2. Z. Li, S. U. Alam, J. M. O. Daniel, P. C. Shardlow, D. Jain, N. Simakov, A. M. Heidt, Y. Jung, J. K. Sahu, W. A. Clarkson, and D. J. Richardson, “90 nm gain extension towards 1.7 μm for diode-pumped silica-based thulium-doped fiber amplifiers,“ in European Conference and Exhibition on Optical Communication (ECOC 2014), paper Tu.3.4.2.
3. R. H. Stolen and E. P. Ippen, “Raman gain in glass optical waveguides,” Appl. Phys. Lett. 22(6), 276–278 (1973). [CrossRef]
4. V. E. Perlin and H. G. Winful, “Optimal design of flat-gain wide-band fiber Raman amplifiers,” J. Lightwave Technol. 20(2), 250–254 (2002). [CrossRef]
5. M. N. Islam, Raman amplifiers for telecommunications, 1: physical principles (Springer, 2004).
6. P. B. Hansen, L. Eskildsen, A. J. Stentz, T. A. Strasser, J. Judkins, J. J. DeMarco, R. Pedrazzani, and D. J. DiGiovanni, “Rayleigh scattering limitations in distributed Raman pre-amplifiers,” IEEE Photon. Technol. Lett. 10(1), 159–161 (1998). [CrossRef]
7. M. E. Marhic, Fiber Optical Parametric Amplifiers, Oscillators and Related Devices (Cambridge, 2008).
8. T. Torounidis, P. A. Andrekson, and B. E. Olsson, “Fibre-optical parametric amplifier with 70-dB gain,” IEEE Photon. Technol. Lett. 18(10), 1194–1196 (2006). [CrossRef]
9. M. E. Marhic, K. Y. K. Wong, and L. G. Kazovsky, “Wideband tuning of the gain spectra of one-pump fiber optical parametric amplifiers,” IEEE J. Sel. Top. Quantum Electron. 10(5), 1133–1141 (2004). [CrossRef]
10. I. Sackey, F. Da Ros, T. Richter, R. Elschner, M. Jazayerifar, C. Meuer, C. Peucheret, K. Petermann, and C. Schubert, “Design and performance evaluation of an OPC device using a dual-pump polarization-independent FOPA,” in European Conference and Exhibition on Optical Communication (ECOC 2014), paper Tu.1.4.4. [CrossRef]
11. J. M. Chavez Boggio, A. Guimarães, F. A. Callegari, J. D. Marconi, and H. L. Fragnito, “Q penalties due to pump phase modulation and pump RIN in fiber optic parametric amplifiers with non-uniform dispersion,” Opt. Commun. 249(4), 451–472 (2005). [CrossRef]
12. Á. Szabó, B. J. Puttnam, D. Mazroa, A. Albuquerque, S. Shinada, and N. Wada, “Numerical comparison of WDM interchannel crosstalk in FOPA and PPLN-based PSAs,” IEEE Photon. Technol. Lett. 26(15), 1503–1506 (2014). [CrossRef]
13. C. J. S. de Matos, D. A. Chestnut, P. C. Reeves-Hall, and J. R. Taylor, “Continuous-wave-pumped Raman-assisted fiber optical parametric amplifier and wavelength converter in conventional dispersion-shifted fiber,” Opt. Lett. 26(20), 1583–1585 (2001). [CrossRef] [PubMed]
14. S. H. Wang, L. Xu, P. K. A. Wai, and H. Y. Tam, “6.4-dB enhancement of the gain of a Raman-assisted fiber optical parametric amplifier over the sum of the gains of individual amplifiers,” in Optical Fiber Communication Conference (OFC 2008), paper JThA13. [CrossRef]
15. X. Guo, X. Fu, and C. Shu, “Gain saturation in a Raman-assisted fiber optical parametric amplifier,” Opt. Lett. 38(21), 4405–4408 (2013). [CrossRef] [PubMed]
16. M. Jamshidifar, A. Vedadi, and M. E. Marhic, “Reduction of Four-Wave-Mixing Crosstalk in a Short Fiber-Optical Parametric Amplifier,” IEEE Photon. Technol. Lett. 21(17), 1244–1246 (2009). [CrossRef]
17. S. K. Korotky, P. B. Hansen, L. Eskildsen, and J. J. Veselka, “Efficient phase modulation scheme for suppressing stimulated Brillouin scattering,”in Tech. Dig. Int. Conf. Integrated Optics and Optical Fiber Communications2, 110–111, Paper WD2–1 (1995).
18. A. J. Viterbi and A. M. Viterbi, “Nonlinear estimation of PSK-modulated carrier phase with application to burst digital transmission,” IEEE Trans. Inf. Theory 29(4), 543–551 (1983). [CrossRef]
19. R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photon. Technol. Lett. 24(1), 61–63 (2012). [CrossRef]
20. F. Yaman, Q. Lin, S. Radic, and G. P. Agrawal, “Impact of pump-phase modulation on dual-pump fiber-optic parametric amplifiers and wavelength converters,” IEEE Photon. Technol. Lett. 17(10), 2053–2055 (2005). [CrossRef]
21. A. Mussot, A. Durécu-Legrand, E. Lantz, C. Simonneau, D. Bayart, H. Maillotte, and T. Sylvestre, “Impact of pump phase modulation on the gain of fiber optical parametric amplifier,” IEEE Photon. Technol. Lett. 16(5), 1289–1291 (2004). [CrossRef]
22. D. Derickson, Fiber Optic Test and Measurement (Prentice Hall, 1998).
