Investigation of Nonlinear DPSK Fading Due to Cross-Polarization Modulation Marcus Winter Klaus...

Post on 22-Dec-2015

214 views 1 download

Tags:

Transcript of Investigation of Nonlinear DPSK Fading Due to Cross-Polarization Modulation Marcus Winter Klaus...

Investigation of Nonlinear DPSK Fading Due to Cross-Polarization Modulation

Marcus WinterKlaus Petermann

Hochfrequenztechnik-Photonik

TECHNISCHEUNIVERSITÄTBERLIN

Dario Setti

http://www.marcuswinter.de/publications/ofc2009

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

what are we talking about?

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

a typical system

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

selective upgrade of an existing 10 Gbps NRZ infrastructure with DPSK channels (10 / 40 Gbps)

worst case for interchannel nonlinearities

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

cross-polarization modulation(XPolM)

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

XPolM is very similar to XPM

nonlinear variation of the

birefringence refractive index

proportional to sum of interfering channel

Stokes vectors powers

results in the modulation of signal

polarization phase

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

nonlinear polarization effects only

XPolM demonstration / quantification setup

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

polarization states (SOPs) of the CW probe at the transmitter

500 × 256 bits

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

polarization states (SOPs) of the CW probe at a receiver

500 × 256 bits

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

DPSK fading

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

the balanced detector output current depends on the relative polarization between the interfering bits

I cos(Δθ/2)

(Δθ is the angle between Stokes vectors of the two symbols)

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

exemplary SOP evolution over 100 bits @ 10 Gbps

adjacent bits are not completely uncorrelated

the angle Δθ remains relatively small

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

how can we quantify fading?

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

the autocorrelation function (ACF) of thetime series of SOPs

ACF(T) = E[Ŝ(t) · Ŝ(t-T)] = E[cos Δθ(T)]

is an average function of the angle Δθ between SOPs with time interval T

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

every iteration of a fixed system (with random initial parameters) will yield a different set of output SOPs

these result in individual sample ACF(T)

the ensemble of all possible initial parametersis described by ACF(T)

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

average signal fading: ~1.5%

BUT: probability for 15% fading is approximately 10-4

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

systems with high symbol rates are less affectedthan low-rate systems

• T is smaller • ACF(T) is monotonously decreasing

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

systems with residual dispersion per span (RDPS)have higher ACFs than those without (at equal DOP)

• RDPS correlates the distortions in neighboring bits

typical 10G systems have significant RDPSto suppress XPM

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

SOP evolution over 100 bits @ 10 Gbps (systems with equal average DOP)

no RDPS 25% RDPS

there is less “motion” with RDPS

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

what is the (relative) impact on systems?

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

CW probe + 10 × 10 Gbps NRZ interferers

determine the ROSNR of the CW probe as if it were a10 Gbps DPSK signal (all zeros)

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

by using an all-zero DPSK data sequence, we minimize any influence of GVD, PMD, SPM on the result

by correlating the sample ACF(T) and the sample ROSNR penalty for each iteration, we can extrapolate the contribution of XPolM to the interchannel penalty

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

ROSNR penalty (dispersion map 1)

25% RDPS, 0.5 ps/km1/2, Pch = 4 mW

no statistical correlation penalty (distribution) is XPM-related

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

NLT for XPM penalty ~ 1dB is only 1 mWalso no statistical correlation

ROSNR penalty (dispersion map 2)

no RDPS, 0.1 ps/km1/2, Pch = 1 mW

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

XPolM should not be ignored completely

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

ROSNR penalty in a PolDM subchannel

Winter et al., LEOS Annual Meeting 2008, WH3

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

summary

TECHNISCHEUNIVERSITÄTBERLIN

Fachgebiet Hochfrequenztechnik

XPolM induces polarization changesbetween adjacent bits

such polarization misalignment leads to fadingof the detected DPSK signal

the effect is very small compared to XPM distortions

scalar simulations are sufficient(regarding nonlinear DPSK fading)