exploring the benefits of cdma in optical networksgoldbe/papers/sgoldberg_prism.pdfexploring the...

13
Lightwave Communications Research Laboratory Princeton University Exploring the Benefits of CDMA in Optical Networks Sharon Goldberg PRISM Workshop on Optical and Wireless Communications Technologies February 3, 2006

Upload: dangngoc

Post on 13-May-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

Lightwave Communications Research LaboratoryPrinceton University

Exploring the Benefits of CDMA

in Optical Networks

Sharon Goldberg

PRISM Workshop on Optical and Wireless Communications Technologies

February 3, 2006

Lightwave Communications Research LaboratoryPrinceton University

OCDMA Project Team

PrincetonPaul Prucnal, Ivan Glesk, Evgenii Narimanov

Camille Bres, Varghese Baby, Sharon Goldberg, Yue-Kai Huang, Bin Li

Hofstra Wing Kwong

TelcordiaRobert Runser, Tom Banwell

Kambrook Tom Curtis

DARPAJag Shah, Henryk Temkin

Lightwave Communications Research LaboratoryPrinceton University

Wireless CDMA

Cellular Systems

• Used in: (for example)– Wireless Cellular Systems – Unlicensed Spectrum Systems (2.4 GHz, 5.8 GHz)

• Allows:– Asynchronous multiple access– Frequency reuse

• Provides:– Immunity to interference– Variable QoS, Variable data rates– Soft blocking

Optical CDMA• Apply concept of wireless CDMA

to optical domain• Incoherent Coding

– Time-Amplitude – Spectral-Amplitude – Wavelength-Time

• Coherent Coding– Temporal-Phase– Spectral-Phase

Code Division Multiple Access (CDMA)

Lightwave Communications Research LaboratoryPrinceton University

t0 t1 t2 t3 t4 tN

Incoherent (Wavelength-Time) Optical CDMAEach user is assigned one or more unique code sequences.

OCDMAEncoder1 0 1 1 1 0 1 1

Input data Encoded data

λ0λ1λ2λ3

Example of Code Sequence

OCDMA Encoder 1

OCDMADomain

OCDMADomain

OCDMADecoder K

OCDMAEncoder K

Data

Data

Receiver Electronics

OCDMA Tx

OCDMADecoder 1

Receiver Electronics

OCDMA RxOCDMA

Encoder 3

DataOCDMA Tx

OCDMA TxOCDMA Rx

Lightwave Communications Research LaboratoryPrinceton University

crosscorrelation

Wavelength-Time Optical CDMA Decoding

Recover data using optical correlator; a peak indicates code match.• Amplitude of peak increases with number of wavelengths.• For clock & data recovery, codes are designed with minimal autocorrelation side lobes.

autocorrelationCodes are designed to have bounded cross-correlation (≤ 1) for any timeshift between codewords

⇒ users can transmit asynchronously⇒ low multiple access interference

Lightwave Communications Research LaboratoryPrinceton University

Bandwidth on demand: Trading BER for Capacity

• Additional users can be accommodated at the cost of increased BER– For traditional schemes (WDM, TDM) once limit is reached, no other users can be added

on the network

0 20 40 60 80 10010

-12

10-10

10-8

10-6

Number of Simultaneous Transmissions M

Bit

Err

or P

roba

bilit

yP

e

10-9 13 users at 10-9 BER

22 users at 10-8 BER

“Soft blocking”

Incoherent OCDMA using 4-wavelength 101-timeslot carrier hopping prime code without FEC

Lightwave Communications Research LaboratoryPrinceton University

Capacity Analysis*: OCDMA vs Wavelength Routed NetworkWe compare broadcast-and-select networks with K = 32 subscribers where• Calls connected on a circuit-by-circuit basis. • Each circuit is active (carries data) with probability p.

Tx 1

Tx K

K:1Tx 2

1:K

Rx 1

Rx 2

Rx K

OCDMA:Define max BER threshold of 10-9

⇒ System admits M = 13 simultaneous transmissions⇒ When M > 13, BER degrades causing an outage.⇒ Ensure that outages occur with probability < 10-3

0 20 40 60 80 10010-12

10-10

10-8

10-6

Number of Simultaneous Transmissions M

Bit

Err

or P

roba

bilit

yP

e 4-wavelength101-timeslot prime code

10-9 13 users for 10-9 BER

WRN:Assume a WRN with 13 wavelengths ⇒ When 13 circuits are connected, new calls are blocked.⇒ Ensure that blocking occurs with probability < 10-3. * Goldberg, Prucnal, “On the teletraffic

capacity of optical CDMA”, in submission.

Lightwave Communications Research LaboratoryPrinceton University

Capacity Analysis*: OCDMA vs Wavelength Routed NetworkWe compare broadcast-and-select networks with K = 32 subscribers where• Calls connected on a circuit-by-circuit basis. • Each circuit is active (carries data) with probability p.

Tx 1

Tx K

K:1Tx 2

1:K

Rx 1

Rx 2

Rx K

OCDMA:Define max BER threshold of 10-9

⇒ System admits M = 13 simultaneous transmissions⇒ When M > 13, BER degrades causing an outage.⇒ Ensure that outages occur with probability < 10-3

Capacity of OCDMA and WRN:Average number of connected calls

WRN:Assume a WRN with 13 wavelengths ⇒ When 13 circuits are connected, new calls are blocked.⇒ Ensure that blocking occurs with probability < 10-3. 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.90

5

10

15

20

25

30

35

Circuit Activity, p

Ave

rage

num

ber o

f con

nect

ed c

alls OCDMA Capacity

WRN Capacity

* Goldberg, Prucnal, “On the teletraffic capacity of optical CDMA”, in submission.

Lightwave Communications Research LaboratoryPrinceton University

Differentiated Services with OCDMA

1ns(1.25GHz)

200ps(5GHz)

Variable Rate• Exploit the variable length property (cross-correlation ≤ 1 for any rate)• Support users transmitting at different rates at the physical level

(5,11) codewordLower BER

(3,11) codewordHigher BER

Variable QoS• Exploit the variable weight property (cross-correlation ≤ 1 for any weight)• Higher code weight (more λ’s) improves BER but uses more resources• Priority channels can occupy more λ’s – variable QoS at the physical level

Lightwave Communications Research LaboratoryPrinceton University

OCDMA Selective Speedup for Packet NetworksRegular packets sent one at a time with weight w codewords and high reliability (low BER)

Speedup packets sent three at a time with weight w/3 codewords and lower reliability (higher BER)

λ0λ1λ2λ3λ4λ5

t0 t1 t2 t3 t4 tN

λ0λ1

λ2λ3

λ4λ5

t0 t1 t2 t3 t4 tN

Selective speedup is a tradeoff between performance and delay:

–Send packets at higher rate

–Without using more network resources

–Without affecting performance of other users on the media

–With a performance penalty on sped-up packets – i.e. higher BER

Lightwave Communications Research LaboratoryPrinceton University

Applications of OCDMA Selective Speedup

Packet dropped!

Active Queue ManagementInstead of dropping packets during congestion, send some with selective speedup!

Regular packets sent one at a time with weight w codewords and high reliability (low BER)

Speedup packets sent three at a time with weight w/3 codewords and lower reliability (higher BER)

λ0λ1λ2λ3λ4λ5

t0 t1 t2 t3 t4 tN

λ0λ1

λ2λ3

λ4λ5

t0 t1 t2 t3 t4 tN

Selective speedup decreases packet dropping probability, queue length and delay.

Lightwave Communications Research LaboratoryPrinceton University

Applications of OCDMA Selective Speedup

Instead of dropping packets during congestion, send some with selective speedup!Selective speedup decreases packet dropping probability, queue length and delay.

Active Queue Management

Class Based Queuing (CBQ)

Gold

Silver

Bronze

SchedulerPackets

Traditionally, CBQ schedules packets according to packet latency and link utilization priority level.

Selective speed-up adds a new degree of freedom –reliability (BER) vs latency

Lightwave Communications Research LaboratoryPrinceton University

Leverage the Tradeoffs with Optical CDMA

• Bandwidth on demand Tradeoff: BER vs Capacity– No hard limit on the number of active users

• Differentiated servicesTradeoff: BER vs Capacity– Variable QoS - enabled by variable weight codes– Variable rate - enabled by variable length codes

• Reduced congestion in packet-switched networksTradeoff: BER vs Delay– The selective speedup– Applications to:

• Active Queue Management• Class Based Queuing