Abstract
We report the gain properties of the multi-wavelength pulsed Raman pump. Group velocity difference between pump pulses imposes an upper limit to the pulse repetition rate beyond which pump-to-pump interaction occurs. Pump-to-pump interaction effect can be severe especially when pump instantaneous power is high due to small duty cycle. Nevertheless, bigger duty cycle imposes much lower limit to the pulse repetition rate. High repetition rate and bigger duty cycle are required to improve the amplifier noise figure. Therefore, optimized level of duty cycle and pulse repetition rate is required. In addition, shorter pump wavelength separation helps to increase the upper limit of pulse repetition rate. Ultimately this will improve the noise performance of TDM Raman amplifier.
©2006 Optical Society of America
1. Introduction
Time Division Multiplexed (TDM) Raman pumping scheme is an attractive solution to avoid pump-to-pump nonlinear Four Wave Mixing (FWM) and pump-to-pump interaction by temporally separating the pump forming train of pump pulses [1]. Pump wavelengths are time division multiplexed in counter-propagation direction so that pump-signal RIN transfer can be minimized at relatively lower modulation frequency. In order to reduce the receiver penalty due to the temporal gain variation, pulse repetition rate higher than 100 kHz is required [2]. Higher instantaneous gain in backward direction, especially at small pulse duty cycle imposes higher backward ASE noise, which ultimately increases the forward ASE through Rayleigh backscattering. Noisy pump also affects the amplifier noise figure significantly especially when there is group velocity matching between backward ASE and pump [3]. All these effects can be reduced by having higher pulse repetition rate and/or bigger pulse duty cycle. Pulse repetition up to 1GHz was recommended to reduce the backward ASE enhancement. Nevertheless, high repetition rate may cause pump pulses start to overlap. This interaction causes pump-to-pump interaction and possibly noise due to four-wave mixing effect. Therefore in this letter, we investigate the upper limit of pulse repetition rate and the impact of pump-to-pump interaction when the repetition rate is beyond this limit. In section two, we present the analytical model of the TDM gain property and employ this model into multi-wavelength case. We define the upper limit of the pulse repetition rate and discuss its impact to the flat Raman gain in section three followed by conclusion.
2. Analytical model
Consider Raman pump pulses propagate in backward direction with repetition rate f, and duty cycle ρ. Signals and pumps propagate with group velocities Vs(n) and Vp(m) respectively, where n and m denote the number of signal and pump wavelengths. The interaction between counter-propagating pump and signal is governed by:
where P(z), S(z), αp, αs, υp, υs and gR denote pump power, signal power, fiber attenuation at pump and signal wavelength, pump and signal wave-number and Raman gain coefficient (W-1km-1) respectively.
The solution of (1) and (2) can simply be
where Δz denotes segmented interaction length.
Raman gain is generated by pump pulses interacting with the signal pulse within period of
Assuming initial number of pump pulses exist in the amplification medium at time t=0 is , thus total number of pump pulses interacting with signal pulse is
Interaction time between two counter-propagating pulses is given by
where Tp and Ts are the active period of pump and signal pulses respectively. Assume that signal bit rate is much higher than f such that Ts→0 for very small segmentation of signal pulses. Interaction length between two thus can be defined as
The interaction between counter-propagating pulses occurs at regular interval with total number of interactions equals to Ntot. The interval distance Ln can thus be approximated as:
Under no pump depletion condition, the pump power distribution is given by Pi(zi,0)=PL.exp{-αp(L-(i-1)Ln}, where Pi(z,0) denotes the pump power at location points zi=i.Ln and at instantaneous time t=0. PL denotes the pump power at z=L. Temporal gain fluctuation is obtained by temporally shifting the pump power distribution with a regular interval Δt, i.e. Pij(z, j. t).
From (4), (8) and (9), on-off Raman gain is defined as:
Inserting the pump power distribution into (10), the on-off gain becomes
At any amplification length much longer than the effective length Leff(L≫Leff) and pulse repetition rate is sufficiently high, τ approaches a constant value as shown in Fig. 1 below.
Raman on-off gain can thus be approximated as:
3. Results and discussion
We plot flat Raman gains by using multi-wavelength Raman pump on 100 km Non Zero Dispersion Shifted Fiber (NZDF) with fiber attenuation coefficient αp=0.22 dB/km, 2 ps/nm-km dispersion at 1550 nm, zero dispersion slope S0=0.078 ps/nm2-km and peak Raman gain coefficient gR=0.598W-1km-1. Figure 2 shows the on-off Raman gain in C-band for two different pump wavelength combinations. Raman pumps were RIN-free and square modulated with 500 kHz repetition rate, 50% duty cycle, and infinite extinction ratio. Pump power of each pump source is within a range of 0.8W–1.2W to generate 12 dB average on-off gains. From our observation, when pump wavelengths are far away separated, i.e pump configuration in (A), one or two pump source will dominate (i.e. higher power) the others in order to maintain similar flat gain within similar band. Peak pump power as high as 1.5W is required by the pump source with wavelength λp=1444nm in (A). Closer pump wavelength separation makes almost-uniform pump power levels possible.
High instantaneous gain in backward direction enhances the backward ASE, which ultimately enhances the forward ASE. This affects the amplifier noise figure severely especially when pump noise is high. To reduce this detrimental effect, higher duty cycle and higher repetition rate are required. Pulse repetition rate up to 1GHz is required to introduce sufficient walk-off that averages the backward ASE [3]. However, multi-wavelength pumps set the upper limit of the repetition rate before two neighboring pump pulses start to interact. The maximum pulse repetition rate required before two pump pulses with group velocity Vp1 and Vp2 interact is defined as:
where ḺC̱ denotes the location where two neighboring pump pulses start to collide and after which interact with each other. Based on the linear dispersion profile, can be estimated. Obviously, two neighboring pump pulses having wider wavelength separation has lower if zero-dispersion wavelength is not located in between the two pump wavelengths. Table 1 shows the various fmax for two pump wavelength combinations as used in Fig. 2 above and for various LC. We set zero-dispersion wavelength λ 0=1400nm and 50% pulse duty cycle. In our simulation, maximum pulse repetition rates fmax are measured based on the worst condition, which is when two neighboring pump pulses are having the biggest group-velocity difference, i.e. pump pulses with λp=1421nm and 1479nm are adjacent. When pump wavelength separations are smaller, as shown in column 2, the biggest group velocity difference is smaller, which then increases fmax∸
Figure 3(a) above shows that under group velocity matching condition, i.e. zero-dispersion wavelength is located in the mid of two pump wavelengths, fmax is dramatically increased, i.e. when pump wavelengths are λp=1421nm, 1442nm, 1453nm,1479nm and λ0=1450nm, fmax can be higher than 10 GHz. Under this condition, slight change of pulse duty cycle will significantly change the allowable fmax as shown in Fig. 3(b).
When pulse repetition rate is set higher than fmax, pump interaction starts to occur. We quantify the impact of pump-to-pump interaction to the flatness of Raman gain for the two pump wavelength combinations in Fig. 4 below based on similar condition as in Fig. 2. Pulse repetition rate as high as 92 MHz does not significantly affect the flatness of the gain in Fig. 4(a) as compared to Fig. 4(b). This is because the location of collision LC in Fig. 4(a) is further away from the one in Fig. 4(b), such that smaller pump powers are involved in the power transfer process. A 92 MHz pulse repetition rate causes gain ripple degradation about ±0.2dB in Fig. 4(a) and ±0.7dB in Fig. 4(b). Furthermore, pulse repetition rate as high as 122.5 MHz can cause gain ripple degradation up to ±3dB in Fig. 4(a) and ±5.5dB in Fig. 4(b). Higher instantaneous gain at smaller duty cycle may induce much more pump power transfer, which makes the gain profile more sensitive to the pump-to-pump interaction effect. Further work needs to be done to investigate the noises generated by four-wave mixing during the pump-interaction process.
4. Conclusions
Multi-wavelength TDM Raman pump requires more stringent requirement to the pulse repetition rate and duty cycle to avoid the pump-to-pump interaction effect. Smaller pulse duty cycle induces more severe pump-to-pump interaction effect due to high instantaneous pump power. Having zero-dispersion wavelength in the mid of pump wavelengths and/or having shorter pump wavelength separation help to increase the maximum repetition rate. Further investigation needs to be done to observe the significance of maximum pulse repetition rate to the noise performance of TDM multi-wavelength Raman amplifier.
References and links
1. L. F. Mollenauer, A. R. Grant, and P. V. Mamyshev, “Time-division multiplexing of pump wavelengths to achieve ultrabroadband flat, backward-pumped Raman gain,” Opt. Lett. 27, 592–594 (2002). [CrossRef]
2. P. J. Winzer, J. Bromage, R. T. Kane, P. A. Sammer, and C. Headley, “Repetition rate requirements for time-division multiplexed Raman pumping,” IEEE J. Lightwave Technol. 22, 401–408 (2004). [CrossRef]
3. J. Bromage, P. J. Winzer, L. E Nelson, M. D. Mermelstein, C. Horn, and C. H. Headley, “Amplified spontaneous emission in pulse-pumped Raman amplifiers,” IEEE Photon. Technol. Lett. 15, 667–669 (2003). [CrossRef]
4. G. Bolognini, S. Sugliani, and F. D. Pasquale, “Double Rayleigh scattering noise in Raman amplifiers using pump time-division multiplexing schemes,” IEEE Photon Technol. Lett. 16, 1286–1288 (2004). [CrossRef]