1. Optimization of DSP-based Nyquist-WDM PM-16QAM Transmitter
V. Curri(1), A. Carena(1), G. Bosco(1),
P. Poggiolini(1), F. Forghieri(2)
(1) DET, Politecnico di Torino, Torino, Italy.
(2) Cisco Photonics Italy srl, Monza, Italy.
ECOC 2012 Paper Tu.4.A.5

2. Motivation and outline
Assuming to use a DSP, what is the best Tx setup for PM-16QAM in Nyquist WDM links?
Details of Tx structure based on DAC
Evaluation of the optimal CP and BWDAC for each DAC resolution (Nbit) with and without compensation of electro-optical nonlinearity of MZM
Optimal bandwidth (BWDAC) and clipping (CP)
Sensitivity vs. Df, ER and MD
Analysis vs. Tx loss
Comments and conclusions
ECOC Paper Tu.4.A.5

3. Transmitter layout +1 Vp Vin -1 DSP Nested MZM b MZM DAC Nbit RDAC
LSBIx
MSBIx
DSP
DAC
LSBQx
MSBQx
I/Q
Nested MZM
LSBIy
MSBIy
LSBQy
MSBQy
Laser
PBC
Pol
Rotator
Etx(t)
Ptx
VIx
VQx
VIy
VQy
Ideal
4-PAM
Nyquist
filter
ArcSin
ON/OFF
Pre-emphasis
DSP G2 G1 b
Vin Vp -1 +1
MZM
Vin
Vout
DVDAC
5-Bessel
LPF
Vin,DAC
DAC
Vout,DAC G3
RDAC
Nbit
BWDAC
ECOC Paper Tu.4.A.5

4. ArcSin operation vout -1 +1 vin VDAC,in k·Vnyquist
Vnyquist is the max swing of the signal at the output of the Nyquist filter
k properly scales the signal amplitude, depending on MD and CP, in order to achieve the exact compensation of the MZ nonlinearity
k=f(MD,CP)
Exact MZ nonlinearity compensation is possibile only if MD>(100-CP)
ECOC Paper Tu.4.A.5 4 4 4

5. Analysis: general parameters
PM-16QAM
Symbol rate: Rs=32 Gsymbol/s (Rb=256 Gb/s)
Simulation with error counting on 128 Ksymbols (1 Mbit)
5 channels
BER target: 10-3
Back-to-back setup with noise loading
Rol-off: b = 0.05
DAC rate: RDAC = 2 SpS
ECOC Paper Tu.4.A.5

6. Optimization process Ranges for parameters Optimization process Tx ASE
51-taps
LMS Rx
ASE
noise +
Prx
OSNR Df
Nbit CP MD ER
Tx spectrum – Df=33 GHz
Ranges for parameters
Df = 32 -> 50 GHz
MD = 30 -> 70%
CP = 0 -> 40%
BWDAC = 8 -> 32 GHz
Nbit = 4 -> 8
Optimization process
Evaluation of the optimal CP and BWDAC for each Nbit with and without ArcSin operation and ER = 20 dB
Sensitivity vs. Df, ER and MD with CP and BWDAC fixed
Analysis vs. transmitter loss with respect to ideal PM-16QAM
ECOC Paper Tu.4.A.5

7. Results vs. BWDAC
For a given OSNR, for each Df, Nbit we obtained a grid of BER vs.(BWDAC, MD,CP)
with ArcSin ON/OFF
An example
Optimal BWDAC resulted to be 0.5·RS, i.e. 16 GHz, independently of channel spacing (Df) and DAC resolution (Nbit)
Therefore BWDAC = 16 GHz is used in the following analyses
log10(BERmin)
ECOC Paper Tu.4.A.5

8. Results vs. CP An example
For a given OSNR and the optimal BWDAC, for each Df, Nbit we obtained a grid of BER vs. (CP, MD)
with ArcSin ON/OFF
An example
log10(BERmin)
ECOC Paper Tu.4.A.5

9. Optimal clipping percentage
Nbit=8
Nbit=6
Nbit=5
Nbit=4
ArcSin OFF
100-CP [%]
ArcSin ON
100-CP [%]
Nbit 8 6 5 4 CP
12%
14%
20%
29%
Nbit 8 6 5 4 CP
20%
26%
36%
ECOC Paper Tu.4.A.5

10. Analysis vs. modulator ER
For optimal BWDAC and CP, for each Df, Nbit we carried on an analysis of OSNR penalty varying the modulator ER with ArcSin ON/OFF
OSNR vs. MD: Df = GHz, Nbit=5
ArcSin OFF
ArcSin ON
ECOC Paper Tu.4.A.5 10 10

11. From previous results we can evaluate the
Optimal MD vs. ER
From previous results we can evaluate the
optimal MD vs. ER
Optimal MD vs. ER, Df = GHz
ArcSin OFF
ArcSin ON
ECOC Paper Tu.4.A.5 11 11

12. Summary: OSNR@BER=10-3 vs. ER20 30 40 22 24 26 28
ER [dB]
OSNR [dB] N
bit
=8 - DSP no arcsin
=8 - DSP arcsin
=6 - DSP no arcsin
=6 - DSP arcsin
=5 - DSP no arcsin
=5 - DSP arcsin
=4 - DSP no arcsin
=4 - DSP arcsin
Quantum Limit
Df = GHz
ER = 20 dB
limits penalty
to 0.5 dB
ECOC Paper Tu.4.A.5 12 12

13. Summary: OSNR@BER=10-3 vs. Df
ER = 20 dB 35 40 45 50 20 22 24 26 28 30 D
f [GHz]
OSNR [dB]
Df=1.1·Rs
limits penalty
to 0.5 dB N
bit
=8 - DSP no arcsin
=8 - DSP arcsin
=6 - DSP no arcsin
=6 - DSP arcsin
=5 - DSP no arcsin
=5 - DSP arcsin
=4 - DSP no arcsin
=4 - DSP arcsin
Quantum Limit
Optical Nyquist
ECOC Paper Tu.4.A.5 13 13

14. Power Tx evaluator Analysis vs. Loss Loss (Nbit,CP), MD, ER
ArcSin ON/OFF
Power
evaluator
Loss
Reference: CW (Loss = 0 dB)
Ideal rectangular PM-16QAM:
Loss =-10·log10{1/2·[(1)2+(1/3)2]} = 2.55 dB
We evaluate Loss varying MD and ER for each Nbit, given the optimal CP previously evaluated
Given ER, we have a MD  Loss correspondence
Results vs. MD can be re-plotted vs. Loss
ECOC Paper Tu.4.A.5 14 14

15. OSNR@BERtarget vs. Loss
ER = 20 dB, Df = 33 GHz
Solid lines: ArcSin ON
Dashed lines: ArcSin OFF
Zoom: Nbit = 5
~3 dB
~0.2 dB
IDEAL PM-16QAM
Similar results for all scenarios
ECOC Paper Tu.4.A.5 15

16. Comments and conclusions
With b=0.05 in the shaping filter and ER=20 dB we can pack channels down to 1.1·Rs with penalty below 0.5 dB.
The requirement in the DAC resolution is Nbit≥5 with a clipping percentage of 20%.
The use of ArcSin
gives very limited sensitivity advantages that we can roughly estimate below 0.5 dB
permits to reduce the transmission Loss of 2 to 3 dB, i.e., using the same laser source we can get 2 to 3 dB more power at the modulator output with respect to the absence of ArcSin
ECOC Paper Tu.4.A.5

17. This work was supported by CISCO Systems within a SRA contract
Acknoledgemets
This work was supported by CISCO Systems within a SRA contract
The simulator OptSimTM was supplied by RSoft Design Group
(now part of Synopsys)
ECOC Paper Tu.4.A.5

18. OSNRpenalty vs. BWDAC/Rs given RDAC
Df = GHz – b = 0.05
OSNRpenalty vs. BWDAC/Rs given RDAC
0.3
0.4
0.5
0.6
0.7
0.8
0.9 1
1.1
1.2
1.5 2
2.5 3
3.5 4 BW
DAC /R S
OSNR
penalty
[dB]
1.27 SpS
1.40 SpS
1.56 SpS
1.75 SpS
2.00 SpS
2.33 SpS
2.80 SpS
ECOC Paper Tu.4.A.5