1. HFA Optimization for Nyquist WDM Transmission
V. Curri, A. Carena
OptCom
DET, Politecnico di Torino, Torino, Italy.
Paper W4E.4

2. Motivations
Need for larger capacity using the installed infrastructures
Larger capacity means larger OSNR
Some Raman pumping as a complement of EDFA – Hybrid Fiber Amplifier (HFA) - is a possible solution but some questions need answers
How much pump power do I need?
Co- or counter-propagating pump or both?
What is the Raman merit on different fiber types?
Supposing uniform and uncompensated links operated with NyWDM channel combs based on multilevel modulation formats and coherent Rx, the GN-model may help in giving proper answers
Paper W4E.4

3. Outline The analyzed system scenarios
System Figure of Merit including ASE noise and NLI
Raman merit vs. EDFA
GN-model and Raman amplification
HFA optimization based on GN-model
Results for span loss As = 20 dB
Moderate pumping regime: a simplified scenario
Summary results for different span losses
Preliminary specific simulative validations
Conclusions
Paper W4E.4

4. …… NyWDM channel comb GTx(f) Bopt Rs Df GWDM f
NyWDM channel comb Df = Rs
PM-m-ary-PAM
Flat PSD: GWDM W/Hz
Pch= GWDM Rs
Use on the entire C-band Bopt= Nch·Rs
For calculations we consider
Bopt = 4 THz
Rs= 32 Gbaud
Nch = 125
Paper W4E.4

5. Uniform and uncompesated link
xNs
Ls, adB, D, Aeff, n2 gR, RCF
NyWDM Tx
Polarization
Diversity Rx
DSP
GEDFA
FEDFA=4.5 dB
Ppump,co
Ac,in
DOP=0
Ppump,counter
Ac,out
DOP=0
NZDSF
adB = dB/km
D = 3.8 ps/nm/km
Aeff = 70 mm2
SMF
adB = dB/km
D = 16.7 ps/nm/km
Aeff = 80 mm2
PSCF
adB = dB/km
D = 21 ps/nm/km
Aeff = 135 mm2
ap,dB = adB+0.05 dB/km - n2 = 2.5 x m2/W - Flat gR = 0.33 x m/W - RCF = -27 dB
Paper W4E.4

6. Equivalent span: loss, ASE noise and NLI
ASE noise PSD
Energy
per photon
Total span loss
GHFA +
ASE
noise
Ac,in
As=adB·Ls
Ac,out
HFA +
NLI
disturbance
NLI PSD independent of other spans
Transparency condition
Note: so far no specific model for NLI evaluation
Paper W4E.4

7. Co- and counter-pumping and percentage of RA
Small-signal Raman gain
Overall pump power
Percentage of Raman gain in HFA
Percentage of co-prop pump
From full-EDFA to full Raman
Co- and counter-prop RA
Paper W4E.4

8. independent of format and BER
System Figure of Merit
Max OSNR (Min BER)
Max reach
Nph= GWDM /Eph is the average number of photons per symbol
In general, Feq and GNLI depend on Nph, percRA and percco
System FoM
independent of format and BER
Optimal power
Optimal FoM
Paper W4E.4

9. Raman merit Raman merit
We define as Raman Merit DRA the increasing of optimal FoM when applying some Raman pumping, with respect to the full-EDFA setup, given the link and HFA configuration
Raman merit
Note that Raman merit besides depending on the link structure varies with (percRA, percco)
Paper W4E.4

10. GN-Model and Raman amplification
Can we use the GN-model to evaluate performance with RA?
Several papers also at OFC15, e.g.:
Invited paper M2I.6 by R. Pastorelli (Cisco Systems)
Tutorial paper W1C.1 by W. S. Pelouch (Xtera Communications)
Paper W4E.4

11. Using GN-model to evaluate DRA
We need to evaluate Feq and GNLI vs. (Nph, percRA, percco) z 1
HFA Ls
EDFA
GRA
Normalized NLI efficiency from GN-model
Feq from accurate model of RA
Paper W4E.4

12. HFA optimization
For each system scenario, due to Raman pump depletion, G(z) depends on Nph, given (percRA, percco)
Consequently, both Feq and GNLI depend on Nph
MPI induced by Rayleigh backscattering degrades Feq for high pump power, so it must be considered
In general, the HFA optimization needs an accurate characterization of Feq and GNLI vs. Nph
Then, Fsys can be computed vs. Nph for every (percRA, percco)
Fsys,opt is the maximum Fsys for every (percRA, percco)
Paper W4E.4

13. The analysis Implemented analytical tools Considered system scenarios
Accurate Raman amplification model giving G(z) and Feq for every pumping scheme and including MPI
GN-model evaluating GNLI from the output of the RA model
Considered system scenarios
Links made of uniform fiber types: NZDSF, SMF and PSCF previously described
Fiber span loss As = 20 ->30 dB, step 2 dB (Ls=As/adB)
percRA = 0% (full EDFA) -> 100% (full RA), step 11%
percco = 0% (counter-propagating pump), -> 100% (all co-propagating pump) step 25%
Displayed results
Results f or As = 20 dB and SMF
Results comparing fibers for As = 20 dB
Summary results vs. As
Paper W4E.4

14. SMF – As=20 dB (Ls=100 km) - GRA vs. Pch
½ Raman
½ EDFA
Full
Raman
Paper W4E.4

15. SMF – As=20 dB (Ls=100 km) - GNLI vs. Pch
Full
EDFA
½ Raman
½ EDFA
Full
Raman
Paper W4E.4

16. SMF – As=20 dB (Ls=100 km) - Feq vs. Pch
Full
EDFA
½ Raman
½ EDFA
Full
Raman
Paper W4E.4

17. SMF – As=20 dB (Ls=100 km) - Fsys vs. Pch
Full
EDFA
Fsys reference value for Raman merit vs. EDFA
½ Raman
½ EDFA
Full
RAman
Paper W4E.4

18. All fibers – As=20 dB - Pch,opt vs. percRA
NZDSF
SMF
PSCF
Paper W4E.4

19. All fibers – As=20 dB - DRA vs. percRA
NZDSF
SMF
PSCF
Paper W4E.4

20. All fibers – As=20 dB – Pch,opt vs. percRA
Full
Counter-prop
½ Counter-prop
½ Co-prop
Full
Co-prop
Paper W4E.4

21. All fibers – As=20 dB – DRA vs. percRA
Full
Counter-prop
½ Counter-prop
½ Co-prop
Full
Co-prop
Paper W4E.4

22. Comments (I) Full counter-prop HFAs (percco=0%)
Experience low pump-depletion effect
For moderate pumping-regime (roughly, percRA <66%) it is practically negligible
Low pump-depletion means no major transients and enables the use of percco=0% HFAs in reconfigurable networks
Are affected by negligible RA-induced NLI enhancement
Are the best choice for moderate pumping-regime
If co-propagating pump is used, the effect of pump depletion is more and more evident approaching percco=0%, as well as RA-induced NLI
Paper W4E.4

23. Comments (II)
The best choice in terms of DRA seems to be close to Full-RA (percRA =100%) for all fibers but the incremental benefit (∂DRA/∂Ppump) is larger for low percRA (<50%)
Approaching percRA =100%, a minor percentage of co-propagating pump may improve DRA only of fraction of dB but reduces the optimal power per channel (e.g., 5 dB for SMF)
Possible reduction of power consumption
Raman merit vs. percRA is very similar for all fibers, in particular for high dispersive fibers like SMF and PSCF
For full counter-prop HFAs, Raman merit vs. percRA is practically the same for all fibers
Paper W4E.4

24. Moderate pumping regime
If RA is a supplement to the EDFA gain, we can suppose the moderate pumping regime roughly corresponding to percRA <66%
Supposing also full conter-prop pumping, pump depletion is negligible as well as MPI , consequently Feq and GNLI are ~ constant
Raman merit and optimal power assume analytical close forms as for full-EDFA amplification
DFeq=Feq,EDFA,dB-Feq,HFA,dB, always positive, is the RA noise reduction
DGNLI=10log10(GNLI,HFA/GNLI,EDFA) always positive as well, is the RA NLI enhancement due to distributed gain
If DGNLI can be neglected
Paper W4E.4

25. All fibers - DRA vs. fiber-span loss
NZDSF
SMF
PSCF
Paper W4E.4

26. All fibers - Maximum DRA vs. fiber-span loss
Maximum attainable
(best) Raman merit
Percentage of co-propagating pump giving the best Raman merit
Percentage of RA giving the best Raman merit
Paper W4E.4

27. Comments Increasing the fiber-span loss As,
benefit of RA increases
DRA gap between percco=25% and percco=0%, that is negligible for As = 20 dB, grows with As clearly showing that the best choice is to split the pump in 25% co and 75% counter
DRA gap between the fiber types grows from 0.2 dB for As = 20 dB to 0.5 dB for As = 30 dB
For all fibers and losses the best HFA is approaching the full Raman as percRA,opt is always  90 %
Paper W4E.4

28. Preliminary simulative validation
What about specific validations on the analyzed scenarios?
Full max reach simulations are extremely time-consuming but we have preliminary results for NZDSF
Paper W4E.4

29. Conclusions We presented a method for HFA optimization
Results show that Raman pumping up to full-RA improves OSNR for all scenarios but incremental benefit (∂DRA/∂Ppump) is larger for low percRA (<50%)
Counter-prop pumping always allows to reach system improvements close to the optimum
Some co-pumping
gives increasing extra DRA with span enalarging
always reduces Pch,opt
Prevalent co-pumping (percco >50%) never gives advantages
Raman merit is similar for all fibers and practically equal in case of counter-prop pumping
Preliminarily simulative validations confirm the proposed method
Paper W4E.4

30. Acknowledgements Thanks to Mr. Carlos Gomez
Universidad de Valladolid (Spain)
Prof. Gabriella Bosco and Prof. Pierluigi Poggiolini
OptCom, DET, Politecnico di Torino (Italy)
Dr. Fabrizio Forghieri
Cisco Photonics Italy
Mr. Enrico Ghillino
Synopsys (USA)
Paper W4E.4