You are on page 1of 6

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts

for publication in the WCNC 2009 proceedings.

Packet Size Optimization for Goodput and Energy Efficiency Enhancement in Slotted IEEE 802.15.4 Networks
Yan Zhang and Feng Shu IMEC-NL/Holst Centre High Tech Campus, Eindhoven, The Netherlands y.zhang@imec-nl.nl, feng.shu@imec-nl.nl
AbstractTo address system goodput and energy efficiency enhancement, this paper studies packet size optimization for IEEE 802.15.4 networks. Taking into account of the CSMA-CA contention, protocol overhead, and channel condition, new analytical models are proposed to calculate the goodput and the energy consumption. Optimal packet sizes are derived from the new models for different network scenarios. The analytical results are validated by the simulation results. The results from our work can be used to facilitate efficient packet segmentation. Index Termspacket size, optimization, 802.15.4, goodput, energy efficiency

I. INTRODUCTION IMING to achieve ubiquitous wireless networks, extensive works are carried out on providing short distance wireless communication solutions. Recently, the IEEE 802.15.4/ZigBee standards have attracted a lot of attention. The IEEE 802.15.4 standard [1] is released to regulate low-rate and low-cost short distance wireless personal area networks (LR-WPANs). Physical layer (PHY) and medium access control (MAC) layer are specified in this standard, based on which the ZigBee Alliance defines the upper layer protocols [2]. In the IEEE 802.15.4 standard, the formats of PHY and MAC packet head are well-defined, whereas the payload of a data packet is variable with the constraint of maximum PHY service data unit (PSDU) size. Therefore, packet size optimization is open to be designed to facilitate effective packet segmentation at the transport layer. Since only the MAC payload is the effective data transmission, it might be desired to make the payload part as long as possible to enhance the radio resource efficiency. However, on the other hand, the chance that a certain data packet is impaired by the error-prone wireless channel also increases with the packet size. Moreover, due to the nature of the contention-based single channel random access, the cost of packet collision also increases with the increase of data packet size. Hence, the packet size should be determined by trading off all the related factors. The problem of packet size optimization has been studied previously for different systems with different targets, such as throughput, energy efficiency and latency [3]-[9]. Wang et al.

[7] propose the idea to improve the energy efficiency for a wireless local area network (WLAN) by optimizing packet size and contention window size. Making use of link adaptation, Wu et al. [8] formulate the packet size optimization problem in multi-rate wireless networks to achieve goodput enhancement. By taking routing, channel condition and error control schemes all into consideration, Vuran and Akyildiz [6] investigate cross-layer packet size optimization for sensor networks in different transmission environments. However, to the best of the authors knowledge, there is no previous literature studying packet size optimization for IEEE 802.15.4 networks, which initiates the discussions developed in this paper. By considering both MAC and PHY layer constraints, we study the problem of how to optimize the packet size in terms of maximizing resource efficiency and energy efficiency. The system under study is assumed to be in saturation mode, which means that every node in the system always has a packet to be transmitted at any moment. This simplified assumption is made because of the following reason. As a contention-based single medium system, IEEE 802.15.4 network confronts significant performance deterioration with the increasing user number and traffic load. However, with the growing demands on short range wireless communication and the requirements on improving user experience, high network activity is an inevitable problem that the system designers have to tackle in the near future. In this paper, we aim to alleviate this bottleneck of an IEEE 802.15.4 network by conducting cross-layer packet size optimization. The remainder of this paper is organized as follows. First Section II gives a brief overview of the IEEE 802.15.4 standard and the Markov chain model of the slotted CSMA-CA mechanism. The objective functions of packet size optimization are derived in Section III. In Section IV we compare analytical predictions and simulation results, and point out the achieved performance enhancement. Finally, this paper is concluded in Section V. II. INTRODUCTION OF IEEE 802.15.4 Before going into details of the discussion in this paper, we first give a brief overview of the IEEE 802.15.4 standard and the Markov chain model of slotted CSMA-CA mechanism.

978-1-4244-2948-6/09/$25.00 2009 IEEE

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2009 proceedings.

A. Standard overview An IEEE 802.15.4 network can operate in either beacon-enabled or nonbeacon-enabled mode. In our work, only the beacon-enabled mode is considered. The beacons are used to synchronize all the devices in the network, and to bound the superframes. A superframe has an active and an inactive portion. The inactive portion is deployed for the low rate and low power consideration. In the active part, the superframe is partitioned into three parts: a beacon, a contention access period (CAP), and a contention free period (CFP). In the CAP part, the nodes employ the CSMA-CA mechanism to access the channel. The CFP part is divided into guaranteed time slots (GTSs) on a reservation based approach. A node can first send a GTS request on the CAP part using CSMA-CA, and then the network coordination will decide the GTS allocation. The superframe structure is shown in Fig. 1.
Beacon CAP CFP GTS GTS Inactive

Markov chain. Here we develop our analysis based on the results concluded in these papers. The per user 2-D Markov chain for unacknowledged data transmission is shown in Fig. 2, where each state indicates one backoff delay period. The state is defined as {s(t ), c(t )} .

s (t ) {0, 1,.., m} is the stochastic process representing the


backoff stage, and s(t ) = 1 is the transmission stage. c(t ) is the stochastic process representing the state of the backoff counter at slot t . The probabilities that the node will detect channel busy in the first CCA backoff period (CCA1) and second CCA backoff period (CCA2) are and respectively. The packet to be transmitted takes L backoff periods.
1

-1,L-1

-1,0 1/W0 1 0,-2 1 0,-1


1

1/W0 0,1
1

0,2 1/W1

0,3

0,W0 -2

0,W0 -1

1/W1
1
1 1 1

IEEE 802.15.4 frame structure

1 1,-2

1,-1

1,1

1,2

1,3

1,W1 -2

1,W1 -1

Fig. 1. IEEE 802.15.4 super frame structure.

In the slotted CSMA-CA, the basic time slot unit is called one backoff period, which is aligned with the superframe boundary. The access contention procedure is determined by three variables: NB , CW and BE . NB stands for the maximum backoff times allowed in one transmission attempt. CW is the contention window length, which is the number of backoff periods the channel should be clear of activity before a transmission can commence. The value of CW is defined to be 2 in the standard. Backoff exponent BE determines the maximum backoff periods a node should wait before attempting to access the channel. For simplicity, we use m on behalf of NB , use i to indicate a certain backoff, and use Wi to represent the maximum delay in backoff stage i . The slotted CSMA-CA procedure works as follows. First i is initialized to be 0, and Wi is initialized to be 2 MinBE . The node randomly selects a number within the interval of [0, Wi 1] to initialize the backoff delay counter. On the boundary of each slot, the backoff counter is decremented by one. When the counter reaches zero, the node performs clear channel assessment (CCA). If the carrier is detected to be idle, the node begins to transmit in the following slot. Otherwise, i is increased by 1, while Wi is doubled. And then another backoff begins. With Wi upper bounded by 2 MaxBE , the transmission attempt repeats until a successful transmission takes place or m is reached by i .

1/Wm 1 m,-2

1/Wm
1

m,-1

m,1

m,2

m,3

m,Wm-2

m,Wm-1

Pf

Fig. 2. 2-D Markov chain model of IEEE 802.15.4 CAP [11].

The state transition probabilities can be calculated by [11]: (1) P{i, k | i, k + 1} = 1, k 1 (2) P{i, k | i 1,1} = ( + (1 )) / Wi , i m, k Wi 1 (3) P{0, k | i,1} = (1 )(1 ) / W0 , i < m (4) P{0, k | m,1} = 1 / W0 Equation (1) indicates the decrement of backoff counter and the first CCA slot which happens with probability 1. Equation (2) represents the probability that the node selects a random backoff delay in the next backoff stage when there is a failure in either CCA procedure. Equation (3) states it is possible to enter the first delay line from a backoff stage other than the last backoff stage only when both CCA procedures are successful. The probability that the transition from the last backoff stage to any delay state in the first backoff stage is given in (4). Assume the steady-state probabilities of the above Markov chain is bi , k = P{( s (t ), c(t )) = (i, k )} . Note that when the
backoff delay is zero, the node enters the CCA procedure immediately. Although the zero backoff case is not one of the states in the Markov chain, we would like to use the transitional state bi ,0 for the ease of description, which has the same value as bi ,1 . From (1) and (2), we get bi ,0 = [( + (1 ) )]i b0,0 , 0 < i m Hence the chance that a node attempts carrier sensing is (5)

B. Markov chain analysis of slotted CSMA-CA mechanism As discussed in previous studies [10]-[12], the slotted CSMA-CA mechanism in an IEEE 802.15.4 system with saturation traffic can be modeled as a two-dimensional (2-D)

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2009 proceedings.

= bi ,0
i =0

(6)

Assume there are N nodes in the system. By considering the interaction between users on the single medium, the probability of sensing channel busy in the first CCA is = L(1 (1 ) N 1 )(1 )(1 ) (7) Similarly, the probability that the node senses channel busy in the second CCA is

Fig. 3. IEEE 802.15.4 data frame format.

(8) 2 (1 ) N Since the probabilities of all the states should sum to 1, the steady-state probabilities , and b0,0 can be calculated numerically using the fixed point approach. III. PACKET SIZE OPTIMIZATION In the IEEE 802.15.4 standard, the PHY and MAC layer packet head formats are well-defined, which introduce fixed protocol overhead to the data packet transmission. The key contribution of this paper is that we deduce the optimized packet size in terms of resource and energy efficiency maximization. To accurately depict the system performance, the objective functions are set up on account of protocol overhead, channel condition and CSMA-CA contention. A. Protocol overhead Here we will give a brief explanation of the data frame format defined in the standard. As shown in Fig. 3, the MAC data service enables the transmission and reception of MAC protocol data units (MPDUs) across the PHY data service. An MPDU consists of MAC header (MHR), MAC service data unit (MSDU), and MAC footer (MFR). The MSDU field is the variable data payload part. The length of MHR part depends on the address field size and whether auxiliary security is enabled or not. If the address field is fixed to be 4 bytes and the security field is disabled, the length of MHR and MFR together is 9 bytes. The MPDU is passed to the PHY as the PHY service data unit (PSDU), which becomes the PHY payload. The PHY payload is prefixed with a synchronization header (SHR), containing the Preamble Sequence and Start-of-Frame Delimiter (SFD) fields, and a PHY header (PHR) containing the length of the PHY payload in octets. The preamble sequence and the data SFD enable the receiver to achieve symbol synchronization. The SHR, PHR, and PSDU together form the PHY packet. The length of SHR depends on the PHY parameters, and the length of PHR is fixed to one bytes. The maximum PSDU size is 127 bytes as defined in the standard. Note that only the MAC payload part MSDU of a data packet carries useful information bits, and the rest fields are protocol overhead.

1 (1 ) N 1

B. Channel condition At the PHY layer, BER is used as the parameter to indicate the wireless channel condition, which can be detected by the node locally. Given a certain BER value, the packet error rate can be deduced by PER = 1 (1 BER)l (9) where l is the data packet size in bit. Comparing with the packet size represented by the number of backoff periods L in the Markov chain, l can be calculated by l = Ltb R (10) where tb is the time interval of one backoff period, and R is the physical layer bit rate.
C. CSMA-CA contention The CSMA-CA contention behavior can be approached by the Markov chain model as explained in Section II. B. Here we make use of the steady-state probabilities derived from the Markov chain model to capture the influence of multiple access contention on system performance, and develop a new equation to describe the transmission collision rate more accurately. According to the Markov chain model shown in Fig. 2, the probability to transmission ptr is given by the probabilities that the node attempts carrier sensing when the channel is idle [11]. ptr = (1 )(1 ) (11) Because the CSMA-CA mechanism in IEEE 802.15.4 standard employs only passive channel assessment without a channel access announcement, such as RTS-CTS handshake in IEEE 802.11, the data packets are prone to collision as long as more than one node takes the CCA step at the same time when the channel is idle. In [11], the collision probability is represented as

1 (1 ) N 1 (Equation (27) in [11]). However, this


formulation neglects that there is also a chance that no users are carrying out channel sensing at a certain moment. Besides, if the channel is busy, multiple users start CCA at the same moment will not lead to packet collision. By taking care of these possibilities, here we derive a new expression for packet collision probability by (12) Equation (12) denotes that packet collision happens when there is more than one user transmitting at the same moment.
D. Resource efficiency The resource efficiency of a wireless network can be gauged by the maximum goodput to be acquired with given radio

pcol = 1 (1 ptr ) N Nptr (1 ptr ) N 1

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2009 proceedings.

resource. Goodput is the application level throughput, which denotes the number of information bits, excluding protocol overhead and data retransmission, forwarded successfully from the source to the targeted destination per unit of time. Given the three constraints discussed in the previous sections, the goodput can be modeled as
G= (l l H ) ptr (1 pcol )(1 PER ) Dtr + Dbk

frequency band. As illustrated in Fig. 3, the minimum MAC layer overhead is 9 bytes. Hence, together with PHY head, the minimum protocol overhead of a data packet is fixed to be 15 bytes. The packet size can be optimized with a fixed packet head length and the maximum physical layer payload constraints.
TABLE I. SIMULATION PARAMETERS 3 CSMA-CA MinBE related 5 MaxBE parameters 5 NB 0.32 ms Backoff period tb Preamble field length SFD Maximum PSDU size Packet size L User number N BER 4 bytes 1 bytes 127 bytes [3,13] 5, 10, 15, 20, 25, 30, 35, 40, 45, 50 1e-3, 1e-4, 1e-5, 1e-6

(13)

where l H is the fixed protocol overhead, Dtr denotes the expectation of time used for data transmission, and Dbk denotes the expectation of delay due to random backoff and channel sensing. Referring to the Markov chain model described in Section III, we get Dtr = lptr m i (14) W 1 Dbk = bi ,0 ( k + 1 + (1 )) 2 i =0 k =0
Substitute (9)-(12) and (14) into (13), the goodput can be predicted from the steady-state probability. E. Energy efficiency Mean node power consumption is widely used to evaluate the energy efficiency. However, since the energy consumption is closely related to the node behaviors, the values of mean node power consumption are not comparable without specifying the traffic profile and throughput performance. Here we use energy consumption per bit of goodput as the measurement to discuss the energy efficiency. Thus the objective function can be modeled as
E= LPTX ptr + Dbk PSL + LCCA ( PRX PSL ) (l l H ) ptr (1 pcol )(1 PER )

(15)

where P , PRX and PSL are the energy consumptions per TX backoff period in transmission, receiving and sleeping state respectively. LCCA , denoted by the number of backoff periods, is the interval of channel sensing in both CCA1 and CCA2, and can be calculated as LCCA = bi ,0 (1 + (1 ))
i =0 m

In the slotted IEEE 802.15.4 system, the smallest time granularity in resource allocation is one backoff period. Only when the packet size is aligned with the slot boundary of the backoff period, the radio resource is used most efficiently. Therefore, we set the data packet size as integer times of the backoff period. With the given physical layer data rate, the maximum packet size is about 13 backoff periods, while the packet head equals 1.5 backoff periods. In our discussions, the duration of one data packet increases from three backoff periods to 13 backoff periods. Given the PHY layer data rate in the chosen carrier frequency, the packet size represented by the number of backoff periods can be easily converted to the packet size in bit with a scaling factor of 80 bits calculated by 250 kbps 0.32 ms . To investigate the energy efficiency, we adopt the TI Chipcon 2420 transceiver [13] as the energy model. The power consumption parameters are shown in Table II.
TABLE II. POWER CONSUMPTION MODEL 2.1 V 19.7 mA 17.4 mA 0.426 mA

Supply voltage

(16)

Current consumption (RX) Current consumption (TX) Current consumption (idle)

8 , which indicates that the The effective factor equals 20 channel sensing lasts the first eight symbols out of the 20 symbols in one backoff period.
IV. NUMERICAL RESULTS FROM ANALYSIS MODEL AND
SIMULATION VALIDATION

A. Simulation Parameters In order to verify the analytical deduction on the packet size optimization, system level simulations are conducted with a set of typical parameters as listed in Table I. In the simulations, we assume the network operates in the 2.4 GHz frequency band with physical layer data rate of 250 kbps. According to [1], the length of preamble field and SFD are fixed with the specified

B. Simulation Results and Analysis 1) Steady-state probabilities in Markov model By substituting (1)-(8) into (9), the steady-state probabilities of the 2-D Markov chain model can be acquired through numerical calculation. On the other hand, we can also deduce the counterparts from simulation results. Here we provide one example to show the difference between the results generated from the two methods. Fig. 4 depicts the probability of a packet to be transmitted in one CSMA-CA procedure (including several backoff stages) with the parameters shown in Table I. Here the transmission probability is given by ptr / b0,0 , which assumes that the node

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2009 proceedings.

starts from the first backoff stage. Fig. 4 illustrates the agreement between numerical results and simulation results. There is some small difference between the two sets of results, which comes from the granularity of the numerical calculation.
1 0.9

15 14 13

Optimal Packet Size

12 11 10 9 8 7 6

Probability of transmission with CCA

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 5 packet size = 3, sim packet size = 3, ana packet size = 7, sim packet size = 7, ana packet size = 13, sim packet size = 13, ana 10 15 20 25

BER = 1e-3, sim BER = 1e-3, ana BER = 1e-4, sim BER = 1e-4, ana BER = 1e-5, sim BER = 1e-5, ana BER = 1e-6, sim BER = 1e-6, ana

5 5

10

15

20

25

30

35

40

45

50

Number of users

Fig. 5. Optimal packet sizes in terms of goodput maximization.


30 35 40 45 50

15 14

Number of users

Fig. 4. Probability of transmission.

2) Optimal packet size Here we specify a certain application scenario by the number of users in the system, and specify a certain network scenario by the application scenario in a certain channel condition. For a certain application scenario, if the CSMA-CA related parameters are fixed to the values shown in Table I, the steady-state probabilities in the Markov chain model are uniquely determined by the packet size. Hence, the goodput and energy consumption can be predicted with (13) and (15) in case of different packet sizes and channel conditions. By identifying the maximum goodput or minimum energy consumption, we can derive the optimal packet size accordingly for a certain network scenario. The analytical results are compared with the simulation results in Fig. 5 and Fig. 6. There are some small discrepancies between the two sets of results, which comes from the accumulated errors of numerical calculation of the steady-state probabilities. In the worst channel condition with BER of
10 3 , the optimal packet size is around six or seven backoff periods. When the channel condition is improved, it is preferred to use the maximum packet size of 13 backoff periods. With Fig. 5 and Fig 6, it can be seen that the results of optimal packet size roughly converge to the same values with the two different objective functions. Hence, the transport layer can achieve the goodput and energy efficiency maximization concurrently by using an optimized packet size for a certain network scenario. The results also disclose that the optimal packet size is sensitive to the channel condition but the number of users. For a certain node, the channel condition can be detected locally, which facilitates the results from our work to be implemented into a real system.

13

Optimal Packet Size

12 11 10 9 8 7 6 5 5 10 15 20 25 30 35

BER = 1e-3, sim BER = 1e-3, ana BER = 1e-4, sim BER = 1e-4, ana BER = 1e-5, sim BER = 1e-5, ana BER = 1e-6, sim BER = 1e-6, ana

40

45

50

Number of users

Fig. 6. Optimal packet sizes in terms of energy efficiency maximization.

3) Performance enhancement With the derived optimal packet sizes in different network scenarios, here we will further discuss the performance gain achieved from the packet size optimization. Without considering the channel condition and the influence of packet size on CSMA-CA contention, the data transmission efficiency can be defined as the ratio of data payload to the whole packet size. ( L LH ) = (17) L In our discussion, packet size L [3,13] and packet head LH = 1.5 , which are both denoted in the number of backoff periods. Hence, the transmission efficiency of the largest packet is 1.7692 times of the efficiency of the smallest packet. However, simulation results show that the goodput enhancement due to cross-layer packet size optimization can be much higher than the result from the simplified transmission efficiency model (17). Fig. 7 shows the ratio of the highest goodput over lowest goodput resulted from different packet sizes. Only when the channel is in the worst condition

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2009 proceedings.

( BER = 10 3 ) and the user number is below 25, the goodput enhancement is lower than the maximum efficiency gain derived from the simplified model. With the improvement of channel condition, the goodput in case of the optimal packet size is more than tripled in contrast to the worst case, which goes far beyond the maximum gain to be achieved from merely enlarging the payload part. Similarly, we can study the energy consumption difference due to the different packet sizes. Fig. 8 shows the energy efficiency improvement. Here the energy efficiency enhancement is represented by (E E MIN ) E = MAX (18) E MAX where EMIN and EMAX are the minimum and maximum energy consumption used to transmit one bit goodput throughput due to different packet sizes. As depicted in Fig. 8, the energy efficiency enhancement increases with the user number. This validates our initial design target of alleviating the bottleneck scenario in IEEE 802.15.4 networks. V. CONCLUSION Targeting to maximize resource efficiency and energy efficiency, the problem of packet size optimization is formulated in this paper. We make use of the per user Markov chain model to capture the nature of CSMA-CA contention. Then the key performance measurements, system goodput and energy consumption, are re-modeled by taking protocol constraints, channel conditions and access contention into consideration. Based on the new models, we derive the optimized packet sizes with different network scenarios. The analytical results are verified by system level simulation. The results from this paper can be used to bring significant performance enhancement to IEEE 802.15.4 networks by segmenting the data stream in a more efficient way.
5 4.5 4 3.5 BER = 1e-3 BER = 1e-4 BER = 1e-5 BER = 1e-6

80

70

60

E (%)

50

40 BER = 1e-3 BER = 1e-4 BER = 1e-5 BER = 1e-6

30

20 5

10

15

20

25

30

35

40

45

50

Number of users

Fig. 8. Energy efficiency enhancement.

REFERENCES
[1] [2] [3] [4] IEEE 802.15.4, Wireless medium access control (MAC) and physical layer (PHY) specifications for low rate wireless personal area networks (LR-WPANS), Standard, IEEE, 2006. http://www.zigbee.org/en/spec_download/zigbee_downloads.asp P. Lettieri, M. B. Srivastava, Adaptive frame length control for improving wireless link throughput, range, and energy efficiency, IEEE INFOCOM 1998, vol. 2, pp. 564 -571, April 1998. F. Poppe, D. De Vleeschauwer, and G. H. Petit, Choosing the UMTS airinterface parameters, the voice packet size and the dejittering delay for a voice-over-IP call between a UMTS and a PSTN party, IEEE INFOCOM 2001, Apr. 2001, vol. 2, pp. 805-814. Y. Sankarasubramaniam, I. F. Akyildiz, and S.W. McLaughlin, Energy efficiency based packet size optimization in wireless sensor networks, IEEE Internal Workshop on Sensor Network Protocols and Applications, 2003, pp. 1 -8. M.C. Vuran and I.F. Akyildiz, Cross-Layer Packet Size Optimization for Wireless Terrestrial, Underwater, and Underground Sensor Networks, IEEE INFOCOM 2008, Apr. 2008, pp 780-788. X. Wang, J. Yin and D.P. Agrawal, Effects of contention window and packet size on the energy efficiency of wireless local area network, IEEE WCNC 2005, Mar. 2005, vol. 1, pp. 94-99. D. Wu, S. Ci, H. Sharif, and Y. Yang, Packet Size Optimization for Goodput Enhancement of Multi-Rate Wireless Networks, IEEE WCNC 2007, Mar. 2007, pp. 3575-3580. S. Stoa, I. Balasingham, T.A. Ramstad, Data Throughput Optimization in the IEEE 802.15.4 Medical Sensor Networks, ISCAS 2007, May 2007, pp. 1361-1364. T. R. Park, T. H. Kim, J. Y. Choi, S. Choi, and W. H. Kwon, Throughput and energy consumption analysis of IEEE 802.15.4 slotted CSMA/CA, Electronics Letters, vol. 41, no. 18, pp. 10171019, 2005. S. Pollin, M. Ergen, S. C. Ergen, B. Bougard, L. Van der Perre, F. Catthoor, I. Moerman, A. Bahai, and P. Varaiya, Performance Analysis of Slotted Carrier Sense IEEE 802.15.4 Medium Access Layer, IEEE GLOBECOM, 2006, Nov. 2006. R. K. Patro, M. Raina, V. Ganapathy, M. Shamaiah, and C. Thejaswi, Analysis and improvement of contention access protocol in IEEE 802.15.4 star network, IEEE Int.l Conference on Mobile Adhoc and Sensor Systems (MASS07), Oct. 2007, pp. 1-8. http://focus.ti.com/docs/prod/folders/print/cc2420.html

[5]

[6] [7] [8] [9] [10] [11]

Ghighest/Glowest

3 2.5 2

[12]
1.5

1 5

10

15

20

25

30

35

40

45

50

[13]

Number of users

Fig. 7. Ratio of the highest goodput against the lowest goodput.

You might also like