Direct Detection Optical OFDM (DD-OOFDM): Theory and First

Direct Detection Optical OFDM (DD-OOFDM):Theory and First Experiments

Abdulamir Ali, Jochen Leibrich, Werner Rosenkranz(aal@tf uni-kiel de)([email protected])

Christian-Albrechts-Universität zu Kiel

Berlin, 7. July 2009

L N T

Motivation

• In DD-OOFDM, modulator bias was shown by simulation to

be a key parameter to achieve high sensitivity– be a key parameter to achieve high sensitivity

– allow for balancing spectral efficiency versus sensitivity

• Aim of this contribution:

– to show the improvement of sensitivity due to optimized biasing by a

theoretical model

– to experimentally verify the impact of modulator bias on sensitivity

– to show several experiments aiming at optimization of sensitivity

-2-L N T

Chairfor Communications

Lehrstuhl fürNachrichten- und Übertragungstechnik

J. Leibrich and A. Ali, Workshop on Optical Communications, July 2009 , Berlin

Outline

1. Introduction

2. Theoretical model for impact of MZM biasing

3. Experimental results

first experiments (proof of concept)– first experiments (proof of concept)

– experiments aiming at improved sensitivity

4. Conclusion

-3-L N T




Introduction: OOFDM Principles

• Direct detection optical-OFDM Simple (low effort)

Needs carrier (low sensitivity) and frequency gap (spectrally inefficient )Needs carrier (low sensitivity) and frequency gap (spectrally inefficient )Laser

OFDM

• Coherent detection optical-OFDM

MZMOpticalchannel

De-mod

OFDMmod.

Coherent detection optical-OFDM High sensitivity (no carrier) and high spectral efficiency (no gap)

Complex receiver

OpticalOFDMMod

MZM

Re

Im

OFDMDe-

Re900H

y

LO1

Opticalchannel

Mod.

MZM 900

Im

mod.

Im

brid

LO2

-4-L N T




LO2

Introduction: Modulator (MZM) bias

• Transfer characteristic of MZM

Pout, Eout

+1r

Optical

power

eld

and

pow

er

Bias at quadrature

Bias at

1 0 8 0 5 0 1vmod/V

E-field

-1

Rel

ativ

e fi Bias at

|Vbias/V|=0.8

• Bias point strong carrier power l li i t

-1 -0.8 -0.5 0 1

sign

al

Electrical reduced carrier power li i t low nonlinear impact

|Vbias/V|=0.5 (quadrature point)

Driv

ing

s

|Vbias/V|=0.8 more nonlinear impact

-5-L N T




Theoretical Model: Setup for Receiver

• Optically preamplified receiver

( )( )EDFA

Bo Be

E ta( ) i tp( ) i t( )E tp( )

• Back-to-back transmission: receiver input signal Ea(t) is equal to MZM output signal

• Modeling in bandpass domain:

cos 2 cos 2biasMZM a c

s t VE t E t P f t

2MZM a cfV with:

s(t) : OFDM-signal driving the MZMV : bias voltage of MZMVbias : bias voltage of MZMV : switching voltage of MZMP : optical output power of CW-laserf : optical carrier frequency

-6-L N T




fc : optical carrier frequency

Theoretical Model: Computation

• After EDFA and optical filter:

2 2biass t VE P fG

cos 2 co

cos 2 sin

s 22

2i c q c

biasp cE t

n t f t n t f

P f tV

t

G

i c q cf f

with:ni(t), nq(t) : in-phase and quadrature component of EDFA noise

• After photodiode with responsivity R:

G : EDFA gain

After photodiode with responsivity R:

2cos2

biasp

s t Vi t R GP

V

2 22 co

2

.2

1s2

biasi i qn t

s t Vn t GP n

Vt

V

-7-L N T




2 2V

Theoretical Model: Expression for BER

• First-order approximation results in electrical signal-to-noise ratio:22 'S B V 2tan

2 2o s bias

e

S B VOSNRN B V V

with:OSNR : optical signal-to-noise ratio

• In case of QPSK-modulation:

Bo’ : optical reference bandwidth

s : standard deviation of OFDM-signal

In case of QPSK modulation:

22 '21 1 1 1erfc erfc tano s biasS B VBER OSNR 2erfc erfc tan

2 2 2 2 2 2o s bias

e

BER OSNRN B V V

-8-L N T




Theoretical Model: Results

• 10Gb/s DD-OOFDM b2b with QPSKsubcarrier modulationOSNR(0 1 ) f BER 10 3

22 '21 1erfc tano s biasB VBER OSNR • OSNR(0.1nm) req. for BER=10-3 vs.

modulator biaserfc tan

2 2 2 2e

BER OSNRB V V

15 GHz

• Solid line: theoretical model– neglects nonlinearity of

50

45

4010-3 s=0.01V– neglects nonlinearity of

• modulator• photo diode

40

35

30

B]@

BER=

1

s=0.05V

0 1• Dashed line: numerical results– include nonlinearity– best sensitivity ≈13dB OSNR

25

20O

SNR

[dB s=0.1V

s=0.2V

– best sensitivity ≈13dB OSNR 15

10-1 -0.9 -0.8 -0.7 -0.6 -0.5

V Vbias/

-9-L N T




bias

Outline

1. Introduction

2. Theoretical model for impact of MZM biasing

3. Experimental results

first experiments (proof of concept)– first experiments (proof of concept)

– experiments aiming at improved sensitivity

4. Conclusion

-10-L N T




Experimental setup (Transmitter)

Offline (MATLAB)

Real

S/P

QPS

K

F F

T

P/S

Datain

DAC

Realsignal

(t)

AWG Amplifier

26dBBiasLAN

Q I F s(t)

MZM

DFB

• Complex conjugate extension real signal• OFDM signal is generated offline and transferred to the AWG via LAN• AWG (arbitrary waveform generator): 20GS/s 5 8GHz interleaving mode• AWG (arbitrary waveform generator): 20GS/s, 5.8GHz, interleaving mode• Amplifier: two-stage amplification to use the full characteristic of MZM

-11-L N T




Experimental setup (Receiver)

ADC LANOpticalAtt 2

OpticalAtt 1

EDFA ScopeBe=12GHzReceivedopticalsignal Att.2Att.1signal

S/P

F F

T

QPS

K-1

P/SSynch.2.5 BER

Offline (MATLAB)

F Q

Referencesignal

• Optical Att.1: OSNR tuning• Optical Att.2: controls input power to photodoide

S ( l ti li ill ) 50GS/ BW 18GH

Offline (MATLAB)

• Scope (real-time sampling oscilloscope): 50GS/s, BW=18GHz• Synchronization: first 4 OFDM symbols (reference signal) to find maximum

correlation (start of time signal)

-12-L N T




System parameters

• Back-to-back transmission• QPSK modulation• 1024 OFDM symbols• NFFT=2048 • Data ratesData rates

10Gb/s: N=511, no frequency gap Wg=0 5 Gb/s : N=256, with frequency gap Wg=N

10Gb/s 5Gb/s

5 GHz 2.5 GHzAWG output AWG output

-13-L N T




Experimental results

• DSB transmission• Direct output from AWG

Synchronization Received constellation1

0.5

Rx signal after synch.Tx. signal

0

Am

plitu

de

-0.5

due to noise from AWG

2750 2800 2850 2900 2950 3000-1

Sample

-14-L N T





10Gb/s: comparison between theoretical and measurement results

• DSB transmission

• Vbi =0.5V (quad. point), σ =0.05V

1

theoreticalmeasurement

----Vbias 0.5V (quad. point), σs 0.05V

deviation from theoretical because f i t f IMD d th l i

2

BER

) 2.8dB

of impacts of IMD and thermal noise

BER floor @ BER≈10-5

3

4

-log 10

(B

3.4dB

5.1dB@

20 22 24 26 28 30 32 34 36 38

4

5

6

OSNR[dB]

-15-L N T





10Gb/s: influence of modulation depth σs


• Vbi =0.5V (quad. point)

1

0 05s/VVbias 0.5V (quad. point)

Measured improvement is less thanth ti l i t

2

BER

) 4.1dB

0.050.10.150.2

theoretical improvement

BER floor for small and large 3

4

-log 10

(B

signals

BER floor for σs=0.2V starts earlier b f hi h MZM li it 20 22 24 26 28 30 32 34 36 38

4

5

6

because of high MZM nonlinearity OSNR[dB]

-16-L N T





5Gb/s: benefit of frequency gap


• Vbi =0.5V (quad. point), σ =0.05V

1

without gapwith gapVbias 0.5V (quad. point), σs 0.05V

Improvement of ~6.5dB @ BER=10-3 2

BER

)

3dB for half data rate 3.5dB benefit of the gap 3

4

-log 10

(B

~ 6.5dB

With gap no BER floor

20 22 24 26 28 30 32 34 36 38

4

5

6

OSNR[dB]

-17-L N T





5Gb/s: influence of modulation depth σs


• Vbi =0.5V (quad. point)

1

s/VVbias 0.5V (quad. point)

Modulation depth should be b t 0 1V d 0 2V

2

BER

)

0.050.10.150.2

between 0.1V and 0.2V

σs= 0.2V, MZM nonlinearity is high 3

4

-log 10

(B

s y g BER floor at BER≈10-5

15 17 19 21 23 25 27 29 30

4

5

6

OSNR[dB]

-18-L N T





1

V /V

5Gb/s: influence of bias point for σs=0.2V

• DSB transmission0.50.60.70 8

Vbias/V

20(B

ER)

0.80.9

3-log 10

4

5

6

• Asymmetrical clipping performs better than symmetrical clipping

15 17 19 21 23 25 27 29 306

OSNR[dB]

-19-L N T




Asymmetrical clipping performs better than symmetrical clipping

Discussion

• Stability: A DC voltage applied to the bias port may induce drift in the operating

point over time adjustment is requiredp j q

• improvement:O ti l filt ft EDFA f ASE i d tiOptical filter after EDFA for ASE noise reductionElectrical filter after PD for thermal noise reduction

• SSB transmission (with optical channel)Avoid power fading due to chromatic dispersion

-20-L N T




SSB transmission

OFDMsignal Bias

EDFA FBGDAC EDFA

AWG

MZM SSBfilter

EDFA(booster)

ASEfilter

FBG0.2nmtunable

DACBe=12GHz

ADC DSP

EDFA(preamp.) Scope

Transfer function:3dB BW≈0 4nm -20dB transition BW≈10GHzTransfer function:3dB BW≈0.4nm 20dB transition BW≈10GHz

20dBm 23dBm

0.4nm( 50GH ) 0.085nm

(~10GHz)(~50GHz)

-21-L N T




SSB Transmission

• Optimization: trade off between sideband suppression and bias point Low sideband suppression: biasing near null power point High sideband suppression: biasing near quadrature point

SSB filter

f [GHz]-5 -2.5 0 2.5 5

2

2.5

3og10

(BER

)

3

3.5

4

-Lo

-22-L N T




0.75 0.8 0.85 0.9 0.954

Vbias/V


1

5Gb/s: B2B transmission

DSBSSB

2

0(BER

)

3

4

-Log

1

~2dB

4

565 6 7 8 9 10 11 12 13 14 15

6

OSNR[dB]

-23-L N T





• SSB transmission• Received at OSNR=10dB and BER ~ 1e-3

Time signal and spectrum after photodiode Constellation

Time signal

spectrum

Intermodulation distortion

-24-L N T




Conclusions

• Analytical:The increase in sensitivity when optimizing MZM bias voltage is treated

analytically for DD-OOFDManalytically for DD-OOFDMGood agreement of theoretical analysis and numerical simulation over a wide

range of values for the bias voltage

• Experimental: sources of sensitivity degradationMZM nonlinearity driving signal of appropriate amplitude σs=0.1V

….0.2V

Ph di dPhotodiode • Intermodulation distortion IMD needs frequency gap• Thermal noise needs filter

• Further aspects (with optical channel) Synchronization: training symbol Synchronization: training symbol Channel estimation: number of OFDM symbol required

-25-L N T




Thank youThank you

-26-L N T




-27-L N T




Documents

Direct Detection Optical OFDM (DD-OOFDM): Theory and First