3.3 Modeling the Energy Consumptions of WSN nodes
3.3.3 Validation of the Proposed Communication Energy Cost Model
transmission times for each data packet,N(rcc,rcl), can be approximated as:
N(rcc,rcl)=
M axr etry
∑
i=0
(1−rccM axN B+1)i+1rcli (3.11) where M axr etry is the maximum retransmission times specified by the MAC protocols. For the detailed calculation ofrcc andrcl based on the number of contending nodes, please refer to [80]. In one data packet transmission attempt, a sensor node would send in average∑M axN B
i=0 rcci CSMA trials. In each trial, the node waits in average 12(2min(MinBE+i,M axBE) −1)tbk back-off duration and then executes the clear channel assessment (CCA) procedure in the followingtcca
period. Thus, the average channel access time for each transmission is:
tcc(rcc)=
M axN B
∑
i=0
rcci (1
2(2min(MinBE+i,M axBE)−1)tbk +tcca)
(3.12) where MinBE and M axBE are the initial and maximum values of the back-off exponent respectively,tbk is the duration of a back-off period and tcca is the time of one CCA operation.
Note that MinBE, M axBE, tbk andtcca are constant values specified by the MAC protocols.
On this basis, the idle listening timetil(rcc)can be formulated based ontcc(rcc)and it affects the number of overhearing bitskoh(rcc).
In addition, depending on the MAC protocols, the radio is either started once and kept active until the end of successful transmission, or it is turned off and restarted again with each retransmission. In other words, the radio startup times are either one or the average transmission times as shown by:
Nst =
⎧⎪
⎪
⎨
⎪⎪
⎩ 1
N(rcc,rcl)
(3.13)
Example 1: CC2530 + IEEE 802.15.4 non-beacon-enabled mode
In non-beacon-enabled networks, the receiver typically turns on its radio continuously, while the sender wakes up based on the application requirement. In this case, the communication process is always initiated by the sender. When it wants to transmit to the receiver, it first transmits a request command using unslotted CSMA/CA. The receiver responses with an acknowledge (ACK) frame and indicates whether there is pending data for it. To model and calculate the energy consumption in this process, this section refers to a simple application in [82], with only one sender and one receiver both based on CC2530. The communication process and the state diagram between the sender and the receiver without pending data are depicted in Fig. 3.3.
Sender
Request( ) Ack ( )
ST CC
TX TA RX CSMA/CA ( ) OH
IL
RX TA TX
TA Receiver
OH IL
TA RF
off
Figure 3.3: State diagram corresponding to one sender contacting one receiver without pending data in a non-beacon-enabled network.
According to [82, 84] and IEEE 802.15.4 non-beacon-enabled mode [85], the values of the parameters determined by the hardware and MAC protocols are listed in the uper portion of Table 3.3. The application related parameters are easily obtained (see lower part of Table 3.3).
The sender accesses the channel successfully at the first time and transmits without collision, hence N(rcc,rcl) equals one. According to Eq. (3.12), tcc(rcc) is about 1.25 ms. From the communication process the idle listening time of the receivertil(pcc) ≈3.7ms.
Using Eq. (3.8), it obtains that the average energy consumption of the sender is about 0.22 mJ, which is close to the measurement result of 0.25 mJ in [82]. The deviation is due to the slightly different transmitting and receiving power in our analysis and their experiments.
Table 3.3: Values of the related parameters in the proposed communication energy cost model when applying CC2530 with IEEE 802.15.4 non-beacon-enabled mode.
HW
est 50 µJ Under different test conditions, the values vary. This is taken from the measurement Pcc 61.5mW Receive sensibility is -50 dBm
et x(d) 0.344 µJ Radio in TX mode, 1 dBm output power, CPU idle, 28.7m A
er x 0.246 µJ Radio in RX mode, -50 dBm input power, CPU idle, 20.5m A
eta 8.64 µJ Assuming 15mAas the turnaround power
Pil 61.5mW Radio in RX mode, -50 dBm input power, CPU idle eoh 0.344 µJ Radio in RX mode, -50 dBm input power, CPU idle MAC
Lcr 18 bytes The control packet transmitted by the sender Lct 11 bytes The control packet received by the sender Nta 2-4 Depends on the scenario
Nst 1 Only startup once
APP Ld 0 Without data packet transmission
Tscd 5ms Defined by the users
APP&MAC
N(rcc,rcl) 1 Transmit only once rcc 0 No channel access failure
rcl 0 No collision
tcc(rcc) 1.25ms The first access average time til(rcc) 3.7ms Idle listening time
Loh(rcc) 0 No overhearing
Example 2: CC2430 + IEEE 802.15.4 beacon-enabled mode
In beacon-enabled networks, the receiver will periodically wake up to broadcast a beacon that specifies the superframe structure and keep active during this superframe duration. When a sender wants to transmit data to the receiver in a beacon-enabled networks, it first listens for the network beacon. When the beacon is found, the sender transmits its data frame, using slotted CSMA/CA, to the receiver. The receiver can send an optional ACK to confirm the successful reception. To measure the energy consumption for both of the sender and receiver in a beacon-enabled IEEE 802.15.4 network, the experiment in [83] sets up a simple scenario with only one sender and one receiver both based on CC2430. Fig. 3.4 illustrates the communication procedure between the sender and receiver.
In this case, the duration of one superframe of the receiver is 15.36 ms; the beacon and the data frame are 26 bytes and 50 bytes, respectively. As in the last example, N(pcc,pcl) = 1 andtcc(p7) ≈1.25 ms. During one superframe duration, the idle listening time of the receiver
Data ( )
Ack ( )
CC
TX TA RX Slotted CSMA/CA ( )
OH IL
RX TA TX
TA ST
TX
ST Beacon( ) RX
TA
OH
IL RF
off TA
Receiver Sender
Figure 3.4: State diagram corresponding to one sender transmitting data to one receiver in a beacon-enabled network.
til(pcc) ≈11.66ms. Finally, the hardware and the MAC parameters are set according to [86, 87]
and IEEE 802.15.4 beacon-enabled mode [85]; they are listed in Table 3.4.
The average energy consumption of the receiver calculated by Eq. (3.8) is about 1.284 mJ, while the measurement result in [83] is 1.368mJ; the estimated and the measured cost for the sender are 0.53 mJ and 0.91 mJ respectively. The deviation is due to the uncertainty of the synchronization process that causes 0.32 mJ additional energy cost in the sender node and a small additional processing energy that has not considered in the model.
According to the comparison results of the above two examples, it indicates that the commu-nication energy cost of WSN nodes by using the proposed model is very close to the realistic measurements.