Analytical Model for 100 Gb/s Discrete Multi-Tone...

Analytical Model for 100 Gb/s

Discrete Multi-Tone Modulation

40Gb/s and 100Gb/s Fiber Optic Task Force

IEEE 802.3bm Interim Session

Phoenix, Arizona

23-24 January 2013

Ilya Lyubomirsky

Finisar Corp.

23-24 January 2013 Fiber Optic Task Force2

Outline

■ Objectives

■ DMT Analytical Model

■ Impact of Thermal and Shot Noise

■ Impact of Clipping Nonlinearity

■ Impact of RIN

■ Comparison of 100Gb/s DMT with PAM-M

■ Conclusion

23-24 January 2013 Fiber Optic Task Force3

Objectives

■ Continue higher order modulation analysis development in

nicholl_01b_0312, ghiasi_01a_0912, and

lyubomirsky_01a_1112_optx

■ Gain physical insights from analytical models

■ Compare performance of 802.3bm SMF PMD alternatives

23-24 January 2013 Fiber Optic Task Force4

100 Gb/s DMT System Architecture

Source: J. Wei, et. al., “Performance Studies of 100 Gigabit Ethernet Enabled by

Advanced Modulation Formats,” IEEE 802.3, Next Gen. 40Gb/s and 100Gb/s

Opt. Ethernet Study Group, ingham_01_0512_optx, May, 2012

23-24 January 2013 Fiber Optic Task Force5

DMT Effective SNR Model

23-24 January 2013 Fiber Optic Task Force6

DMT Effective SNR Model (continued)

Clipping

Qc(x)Laser

23-24 January 2013 Fiber Optic Task Force7

DMT Effective SNR Model (continued)

PD

23-24 January 2013 Fiber Optic Task Force8

DMT Effective SNR Model (continued)

Clipping

Qc(x)

Bussgang Theorem: Clipping noise ncl(t)

uncorrelated with x(t)

Closed form formulas derived in E. Vanin, “Performance evaluation of intensity

modulated optical OFDM system with digital baseband distortion,” Opt. Exp.,

Vol. 19, No. 5, pp. 4280-4293, 2011

23-24 January 2013 Fiber Optic Task Force9

Monte-Carlo Simulation Parameters

Parameter Value

Sampling rate, Fs 60 Gs/s

FFT size, N 128

Number of nonzero subcarriers, Nsc 55

High freq. subcarriers padded to zero 8

DC subcarriers padded to zero 1

Cyclic prefix, CP 4

Clipping ratio, Rcl 8 dB

QAM modulation order, M 16

Noise bandwidth, f 25.8 GHz

Thermal noise density, Sth 16 pA/sqrt(Hz)

Photodiode responsivity, 0.8 A/W

23-24 January 2013 Fiber Optic Task Force10

100 Gb/s DMT Analytical Model Results

1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

1.E-01

-14.5 -13.5 -12.5 -11.5 -10.5 -9.5

BER

Average Optical Power (dBm)

Analytical Model, Thermal Noise

Monte-Carlo Simulation, Thermal Noise

Analytical Model, Thermal+Shot Noise

Monte-Carlo Simulation, Thermal+Shot Noise

23-24 January 2013 Fiber Optic Task Force11

Optimization of Clipping Ratio

1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

1.E-01

0 2 4 6 8 10 12 14 16

BER

Clipping Ratio (dB)

P = -12 dBm

P = -11 dBm

P = -10 dBm

Clipping

Noise

Dominates Thermal

Noise

Dominates

23-24 January 2013 Fiber Optic Task Force12

Impact of RIN

1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

1.E-01

-14 -13.5 -13 -12.5 -12 -11.5 -11 -10.5 -10 -9.5 -9

BER

Average Optical Power (dBm)

No RIN

RIN = -143 dB/Hz

RIN = -140 dB/Hz

RIN = -137 dB/Hz

Rcl = 8 dB

23-24 January 2013 Fiber Optic Task Force13

Analytical Model for PAM-M

For optimum receiver thresholds, the symbol error probability is determined

by an average over all the PAM eye Q-factors

2

0

21022

1

1

1

)(1

2

)()(

f)102(

1

1

1

2

2-M0,...,k, Q(k)

M

k

s

k

RIN

kthk

avkk

kk

kk

kPM

P

kQerfckP

IqIS

ER

ER

M

RPII

II

-0.5 0 0.5-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

Time (UI)

Am

plit

ude

PAM4 @50G with BT4 25G BW

0

1

2

3

I0

I1

I2

I3

23-24 January 2013 Fiber Optic Task Force14

100 Gb/s PAM-M Analytical Model Results

1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

-16 -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5

BER

Average Optical Power (dBm)

PAM-8, no RIN

PAM-8, RIN = -143 dB/Hz

PAM-8, RIN = -140 dB/Hz

PAM-4, no RIN

PAM-4, RIN= -140 dB/Hz

PAM-4, RIN = -137 dB/Hz

ER= 10 dB

23-24 January 2013 Fiber Optic Task Force15

100 Gb/s DMT versus PAM-4 and PAM-8

1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

-15 -14 -13 -12 -11 -10 -9 -8 -7

BE

R

Average Optical Power (dBm)

DMT, No RINDMT, RIN = -143 dB/HzPAM-8, No RINPAM-8, RIN = -143 dB/HzPAM-4, No RINPAM-4, RIN = -143 dB/Hz

23-24 January 2013 Fiber Optic Task Force16

Conclusions

■ We presented an analytical model for DMT modulation

based on an effective SNR approach

■ The model provides a useful baseline for performance, as

well as physical insight on the interplay of DMT system

parameters