Distributed Frequency Synchronization for

OFDMA-based Wireless Mesh Networks

Jung-Hyun Kim, jh.kim06@yonsei.ac.kr, Hong-Yeop Song

Yonsei Univ. Seoul, KOREA

Jihyung Kim, Kwangjae Lim, and Dong Seung Kwon

Electronics and Telecommunications Research Institute (ETRI), Daejeon, KOREA

IWSDA’13, Oct. 27 - Nov. 1, 2013

Contents

Background Synchronization Issue in Wireless Mesh Network

Previous frequency synchronization schemes

Sponsor-based frequency synchronization (SFS)

Consensus-based frequency synchronization (CFS)

Distributed frequency synchronization (DFS)

Modified DFS (mDFS) Bias Compensation

Simulation Results

Conclusions

2

Background

Synchronization Issue in Wireless Mesh Network In wireless mesh network, each distributed node locally

synchronizes with its neighbor nodes to communicate with them.

3

Locally synchronized

Background

Synchronization Issue in Wireless Mesh Network Distributed nodes carry out resource allocation and scheduling

to avoid multiple pair interference.

4

Locally synchronized

Locally synchronized

Background

Synchronization Issue in Wireless Mesh Network Even if perfect resource allocation and scheduling scenario is

assumed, multiple pair interference exists due to synchronization problem.

5

Locally synchronized

Locally synchronized Locally synchronized

Strong interference

Background

Synchronization Issue in Wireless Mesh Network Even if perfect resource allocation and scheduling scenario is

assumed, multiple pair interference exists due to synchronization problem.

6

Globally synchronized

Global synchronization is required

Background

Synchronization Issue in Wireless Mesh Network In OFDM systems, there exist three different problems related to

synchronization.

7

Freq. Sync.

Clock Sync.

Time Sync.

SFS

CFS

DFS

Background

Sponsor-based Frequency Synchronization (SFS) Each node synchronizes with only its sponsor node [1].

8

[1] I. Akyildiz and X. Wang, Wireless Mesh Networks. John Wiley Publishing Company, 2009.

5

7 3

6

7

4

Sponsor node

Background

Consensus-based Frequency Synchronization (CFS) Each node adjusts its carrier frequency to the average value

[2],[3].

9

[2] M. K. Maggs, S. G. O’Keefe, and D. V. Thiel, “Consensus clock synchronization for wireless sensor networks,” IEEE Sensors Journal, vol. 12, no. 6, pp. 2269-2277, June, 2012. [3] Jung-Hyun Kim, Jihyung Kim, Kwangjae Lim, and Dong Seung Kwon, "Distributed Frequency Synchronization for Global Synchronization in Wireless Mesh Networks," ICNWC, Lucerne, Switzerland, Oct. 15-16, 2012.

5

7 3

6

5 5 5

4

5

Background

Distributed Frequency Synchronization (DFS) [3] A global frequency synchronization method for wireless mesh

network

Modified Distributed Frequency Synchronization (mDFS) Modified version of DFS dealing with some practical issues and

bias problem

10

[3] Jung-Hyun Kim, Jihyung Kim, Kwangjae Lim, and Dong Seung Kwon, "Distributed Frequency Synchronization for Global Synchronization in Wireless Mesh Networks," ICNWC, Lucerne, Switzerland, Oct. 15-16, 2012.

DFS

Core algorithms of DFS Repetitive estimation, averaging offsets, sharing estimates

11

DFS

Sharing Estimates

Repetitive Estimation

Averaging Offsets

DFS

Repetitive Estimation Each node receives a preamble signal several times from a

neighbor node to obtain diversity gain.

12

Preamble signals

Fading, Noise, Interference,

etc.

t4

t3

t1

t2 t5

DFS

Sharing Estimates To increase estimation accuracy more, we use sharing message.

13

5 1 -4+e1 +4+e2

5 1

5 1

-4+(e1-e2)/2

-4+e1

-4-e2

! Estimated value

Sharing message (offset value estimated and transmitted by neighbor node)

Additional diversity

DFS

Averaging Offsets Each node estimates frequency offsets with neighbor nodes and

then calculates the average value of these offsets.

14

+7

-6 -9

+3

( ) ( )

7 6 3 9 01

5

Modified DFS

15

Bias Compensation

mDFS

Sharing Estimates

Repetitive Estimation

Averaging Offsets

Core algorithms of mDFS Repetitive estimation, averaging offsets, sharing estimates

+ Bias compensation

Modified DFS

With and without bias compensation

16

20 40 60 80 100 120 140 160 180 200-10

-5

0

5

10

15

Number of updates

Norm

aliz

ed c

arr

ier

frequency

Initial average value

Convergent value With bias compensation

=

Not synchronized Synchronized

— w\o bias compensation — w\ bias compensation

Hexagon network

Convergent value Without bias compensation

bias

Modified DFS

Why we need the bias compensation? In group separation and merge scenario, strong interference

can occur temporally.

17

Strong interference

Group separation

Group merge

Modified DFS

How to compensate the bias? Each node compensates its carrier frequency using both of own

variation and neighbor’s variation.

18

[4] Y. Chen, R. Tron, A. Terzis, and R. Vidal, “Corrective consensus: Converging to the exact average,” in Proc. IEEE Conf. Decision Control, Dec. 2010, pp. 1221–1228.

aFbF

a bF F 5

b aF F 3

a bF F 5

2a bF F 3

2

a bF F 1

2a bF F 1

2

b a a bF F F F

3 5 3 5

2 2 2 2 22 2

a b a bF F F F

3 12

2 2

Averaging offsets Bias compensation

Each node subtracts average of variations from its carrier frequency

Average of variations

Bias elimination

Estimation error

Simulation Parameters

19

[5] IEEE 802.16m, Part 16: Air Interface for Broadband Wireless Access Systems: Advanced Air Interface, May, 2011. [6] Recommendation ITU-R M.1225, "Guidelines for the evaluation of radio transmission technologies for IMT-2000," 1997.

OFDM parameters IEEE 802.16m [5]

Channel model RMa, LoS model [6]

Network topology 21 randomly distributed nodes in 40km × 40km simulation plain One hop distance is 10km (SNR = 8.47dB)

velocity 120km/h

Synchronization

threshold (εth)

1% subcarrier spacing

Number of repetitive estimation

3

Period of bias compensation

every 5 DFS updates

ε1 ε2

εN

max( ) ( , , )i th i N , 1

εi : frequency offset between a target node

and its ith neighbor node

Synchronization success criteria

…

Simulation Results

20

[1] I. Akyildiz and X. Wang, Wireless Mesh Networks. John Wiley Publishing Company, 2009. [2] M. K. Maggs, S. G. O’Keefe, and D. V. Thiel, “Consensus clock synchronization for wireless sensor networks,” IEEE Sensors Journal, vol. 12, no. 6, pp. 2269-2277, June, 2012. [3] Jung-Hyun Kim, Jihyung Kim, Kwangjae Lim, and Dong Seung Kwon, "Distributed Frequency Synchronization for Global Synchronization in Wireless Mesh Networks," ICNWC, Lucerne, Switzerland, Oct. 15-16, 2012.

0 10 20 30 40 50 6065

70

75

80

85

90

95

100

Number of updates

Synchro

niz

ation s

uccess p

robabili

ty(%

)

Number of total nodes : 21, Average number of neighbor nodes : 3

mDFS

CFS

SFS

0 10 20 30 40 50 60 70 8070

75

80

85

90

95

100

Number of updates

Synchro

niz

ation s

uccess p

robabili

ty(%

)

Number of total nodes : 21, Average number of neighbor nodes : 3

mDFS

CFS

SFS

DFS update is aligned [3] – impractical

DFS update is not aligned

mDFS : proposed scheme CFS : consensus-based freq. sync. in [2],[3] SFS : sponsor-based freq. sync. in [1]

Conclusions

The proposed mDFS

Reduces memory size by replacing estimated average offset instead of storing all repetitive estimates

Provides offset calculation method for reception failure of preamble signal and/or sharing message

Defines update period and provide an example of it without loss of both convergence speed and estimation performance

Compensates the bias of convergence value of carrier frequency using corrective consensus algorithm

21