1

Clock Synchronization for Wireless Sensor Networks:

A Survey

Bharath Sundararaman, Ugo Buy, and Ajay D. Kshemkalyani

Department of Computer ScienceUniversity of Illinois at Chicago

2

Outline

I. Introduction II. Synchronization protocols III. Comparison IV. Conclusion & comments

3

I. Introduction

4

Why synchronization?

The time of the day at which an event happened.

The time interval between two events. The relative ordering of events.

5

Requirements

Cope with unreliable network transmission and unbounded message latencies.

Able to estimate the local time on the other node’s clock.

Time must never run backward. Should not degrade system performance.

6

Several issues in synchronization (1/5)

Master-slave v.s. peer-to-peer synchronization Master-slave. The slave node consider the cloc

k reading of the master as the reference time and attempt to synchronize with the master.

Peer-to-peer. Any node can communicate directly with every node in the network. They are more flexible but are also more difficult to control.

7

Several issues in synchronization (2/5)

Clock correction versus untethered clocks Clock correction. Correcting the local clock in e

ach node to run on par with a global time scale or an atomic clock.

Untethered clock. Build a table of parameters that relate the local clock of each node to the local clock of every other node in the network. When timestamps are exchanged between nodes, they are transformed to the local clock values of the receiving node.

8

Several issues in synchronization (3/5)

Internal synchronization v.s. external synchronization Internal synchronization. The goal is to minimiz

e the maximum difference between the readings of local clocks of the sensors.

External synchronization. A standard source of time is provided.

9

Several issues in synchronization (4/5)

Probabilistic v.s. deterministic synchronization Probabilistic synchronization. Provide a probabi

listic guarantee on the maximum clock offset with a failure probability that can be bounded or determined.

Deterministic synchronization. Guarantee an upper bound on the clock offset with certainty.

10

Several issues in synchronization (5/5)

Sender-to-receiver v.s. receiver-to-receiver synchronization Sender-to-receiver synchronization. The receiver synch

ronizes with the sender using the time stamps received. Message delay is calculated by measuring the round-trip delay.

Receiver-to-receiver synchronization. If any two receivers receives the same message in single-hop transmission, they receive it at approximately the same time. The receivers exchange the time at which they received the same message and compute their offset based on the difference in reception time.

11

Terminology (1/3)

Time: The time of a clock in a machine p is given by the function Cp(t), where Cp(t) = t for a perfect clock.

Frequency : Frequency is the rate at which a clock progresses. The frequency at time t of clock Ca is C’a(t).

Offset: Clock offset is the difference between the time reported by a clock and the real time.

12

Terminology (2/3)

Skew: The skew of a clock is the difference in the frequencies of the clock and the perfect clock. The skew of a clock Ca relative to clock Cb at time t is (C’a(t) − C’b(t)).

Drift (rate): The drift of clock Ca is the second derivative of the clock value with respect to time, namely C’’a (t). The drift of clock Ca relative to clock Cb at time t is (C’’a (t) − C’’b (t)).

13

Terminology (3/3)

A timer is said to be working within its specification if

where constant ρ is the maximum skew rate specified by the manufacturer.

14

II. Synchronization protocols

15

Remote clock reading method

The client then sets its time to Stime (accurate time from the server) + (T1-T0)/2 (time required to transmit the message).

The time for any message to be sent is highly variable due to network traffic and message routing.

16

Time transmission method (1/2)

M is the source node and S is the target node.

M sends a series of synchronization messages to S. The ith message is sent at time Ti of M’s clock and received at time Ri of S’s clock.

M STi Ri

17

Time transmission method (2/2)δ: the offset between clock S and M

d: message delay

18

Set-valued estimation method (1/3)

We assume that the local times ti and tj on processors Pi and Pj ,respectively, can be related by the linear equation: ti = aijtj + bij

where aij and bij represent the relative skew and offset between the two hardware clocks.

Time-stamped triples.

19

Set-valued estimation method (2/3)

20

Set-valued estimation method (3/3)

21

Reference broadcast synchronization (1/5)

RBS seeks to reduce nondeterministic latency using receiver-to-receiver synchronization and to conserve energy via post-facto synchronization.

22

Reference broadcast synchronization (2/5)

Time critical path: nondeterministic delay

23

Reference broadcast synchronization (3/5)

Receiver j will compute its offset relative to any other receiver i as the average of clock differences for each packet received by nodes i and j.

24

Reference broadcast synchronization (4/5)

25

Reference broadcast synchronization (5/5)

The largest sources of error are removed. Require O(n2) message exchanges for a

network of n nodes.

26

Romer’s protocol (1/4)

Uses innovative time transformation algorithm for achieving clock synchronization

Assumptions: There is a maximum skew ρ of computer clocks Whenever a message is exchanged between

two nodes, the connection remains long enough for the two nodes to exchange one additional message.

27

Romer’s protocol (2/4)

Real time difference Δt Computer clock difference ΔC1, ΔC2

Skew upper bound for node 1 and node 2 are ρ1 and ρ2,, respectively.

tCt 222 11

11t

C

tCt 111 11

1

1

1

1

11

C

tC

11

221

1

2

1

1

1

1CCC

28

Romer’s protocol (3/4)

The message delay between two node is estimated by bounding it within interval

[0, rtt]

29

Romer’s protocol (4/4)

Require low resource and message overhead. The synchronization error increases with the

number of hops along the path of the message containing the timestamp.

30

Timing-sync Protocol for sensor networks (1/3)

A self-configuring hierarchical structure. A node in this structure can simultaneous

act as a synchronization server to a number of client nodes and as a synchronization client to another node.

31

Timing-sync Protocol for sensor networks (2/3)

Two phase. Level discovery phase. It is based on constrained flooding. The

root node is assigned level 0; The receiver assign themselves a level that is one greater than the level in the packet received.

Synchronization phase. T2=T1+δ+d and δ represents the clock offset between two nodes and d represents the propagation delay.

dTT 12

dTT 34

2

;2

34123412 TTTTd

TTTT

32

Timing-sync Protocol for sensor networks (3/3)

The protocol is scalable and the accuracy does not degrade significantly as the size of the network is increased.

The protocol requires a hierarchical infrastructure which makes it unsuitable for highly mobile nodes.

33

III. Comparison

34

Quantitatively evaluation (1/2) Synchronization precision Piggybacking

Reduction of message traffic

Computational complexity Run time and memory requirements Number of messages exchanged

Convergence time Total time required to synchronize a network

Network size Compatibility with sleep mode

Synchronize and active only when application demands it.

35

Quantitatively evaluation (2/2)Protocols Precision Piggybacking Complexity Convergence

timeNetworksize

Sleep mode

RBS [19] 1.85 ± 1.28 μs

N/A High N/A 2-20 Nodes Yes

Romer [56] 3 ms Yes Low N/A Unknown Yes

TSPN [25] 16.9 μs No Low Unknown 150-300 Nodes

Yes

36

Qualitatively evaluation (1/2)

Energy efficiency Accuracy

How well the time maintained within the network is true to the standard time

Scalability Overall complexity Fault tolerance

Poor reliability of message delivery

37

Qualitatively evaluation (2/2)

Protocols Accuracy EnergyEfficiency

OverallComplexity

Scalability Fault Tolerance

RBS[19] High High High Good No

Romer [56] Low High Low Poor No

TSPN[25] High Average Low Good Yes

38

IV. Conclusion & comments

39

Conclusion

The design considerations presented will help designers in building successful synchronization scheme, best tailored to his application.

40

Comments

Pros It is a good start to begin studying synchronization. Design tradeoffs are discussed and these help us in

designing synchronization protocols.

Cons Some protocols’ description are too rough to be useful. The authors didn’t conduct any experiment to verify the

claimed results.