Efficiency Evaluation Metrics for Wireless Intelligent Sensors Applications

Published Online September 2014 in MECS

Wireless technologies have made significant progress in recent years, allowing many applications in addition to traditional voice communications and the transmission of high-speed data with sophisticated mobile devices and smart objects. In fact, they also changed the field of metrology especially the sensor networks and the smart sensors. The establishment of an intelligent sensor system requires the insertion of wireless communication which has changed the world of telecommunications. It can be used in many situations where mobility is essential and the wires are not practical. Today, the emergence of radio frequency wireless technologies suggests that the expensive wiring can be reduced or eliminated. Various technologies have emerged providing communication differently. This difference lies in the quality of service and in some constraints related on the application and it environment. The main constraints to be overcome in choosing a wireless technology revolve around the following conditions [1], [2]:

Table 1. Needs based applications

Types of application	Specifications and Needs
Environmental monitoring	■    Measurement and regular sending ■    Few data ■    Long battery life ■    Permanent connection
Event detection	■    Alert message ■    Priority ■    Confirmation status ■    Few data ■    Permanent connection
Tracking	■    Mobility ■    Few data ■    Localization ■    Permanent connection

Section II present the related works realized in this context. However section III summarize the main constraints of intelligent sensor systems. Moreover, in section IV the performance indicators are defined and discussed, in some cases, compared with those derived from a theoretical as well as simulation analysis results.

Section V provides performance comparison of the popular sensor motes. Finally, section VI concludes the paper and outlines some future directions of research.

• II.    Related Work

In the related work, many research studies in [3-8] have been focused on wireless sensor networks to improve communication protocols in order to solve the energy constraint, to increase the level of security and precision and to expand autonomy for accuracy, feasibility and profitability reasons. On the other side, the field of intelligent sensors remains fertile and opens its doors to research and innovation, it is a true technological challenge in so far as the topology and the infrastructure of the systems based on intelligent sensors are greatly different compared to wireless sensor networks, particularly in terms of size (number of nodes) and routing. In fact, to preserve the quality of these networks, it is very difficult even inconceivable to replace regularly the faulty nodes, which would result in a high cost of maintenance. The concept of energy efficiency appears therefore in communication protocols, [5-9]. Thus, it is very useful to search the optimization of data routing and to limit unnecessary data sending and the collisions [6], [9]. The aim challenge for intelligent sensors systems is to overcome the physical limitations in data traffic such as system noise, signal attenuation, response dynamics, power consumption, and effective conversion rates etc… This paper emphasis on the performance indicators for wireless protocols which stands for superior measurement, more accuracy and reliability. The object of this study is for realizing an advanced intelligent sensors strategy that offers many system engineering and operational advantages which can offer cost-effective solutions for an application.

must select an interface that provides the best integration of the sensor with the others components of the system taking in our account the costs and the constraints of reliability required for a particular application. There are others solutions to collect remote measurements such mobile and satellite communications. The main problems related to the quality of communications are: attenuation problems (distance, obstacles, rain ...), interference and multipath. The realization of the systems based on smart sensors dedicated to the applications mentioned in section I, requires the techniques and the protocols that take into account the following constraints [3]:

• •    The nodes are deployed in high numbers

• •    At any time, the nodes may be faulty or inhibited

• •    The topology changes very frequently

• •    The communication is broadcast

Fig. 1. Block diagram of an intelligent sensor communication

The sensors are limited in energy, in computing capacity, and in memory. In ad-hoc networks, energy consumption was considered as an aim factor but not essential because energy resources can be replaced by the user. These networks are more focused on the QoS than the energy consumption. Contrariwise, in sensor networks, the transmission time and energy consumption are two important performance metrics since generally the sensors are deployed in a vast inaccessible areas.

• III.    New Constraints Of Intelligent Sensor Systems

An intelligent sensor is an electronic device for taking measurements of a physical quantity as an electrical signal, it intelligence lie in the ability to check the correct execution of a metrology algorithm, in remote configurability, in its functions relating to the safety, diagnosis, control and communication. The intelligent sensor can be seen consisting of two parts [10-13]:

• 1)    A measuring chain controlled by microcontroller

• 2)    A bidirectional communication interface with the network, providing the connection of the sensor to a central computer

The communication part reflects all the information collected by an intelligent sensor and allows the user to configure the sensor for operation. It is therefore absolutely essential that this interface be robust and reliable. Figure 1 illustrates the intelligent sensor with its wireless communicating component. A variety of communication interfaces (wireless modules) is available, but not all sensors support these interfaces. The designer

• IV.    Performance Indicators

Intelligents wireless sensor networks are different from the traditional communication networks, and therefore different performance measures may also be required to evaluate them. Among them are the QoS metrics. Some applications in WSNs have real-time properties. These applications may have QoS requirements such as delay, loss ratio, and bandwidth. This section describes the different criteria to evaluate the performance of a wireless communication for intelligent sensors systems.

• A. Network Size and topology

The size and the topology of a sensors network can be adopted according to differents application requirements such as data packets size during traffic, transmission protocols implemented, interferences and dimensions of the monitored area. Table 2 summarizes the main differences in size and in topology between the mentioned protocols.

Table 2. Network s ize and basic topology

Protocol	Bl *	UWB	ZigBee	Wi-Fi	Wi-Max	GPRS
Basic Cell	P	P	Star	BSS	Single Cell	Single Cell
Extension of the Basic Cell	S	PP	Cluster tree, Mesh	ESS	PTMP, PTCM, Mesh	Cell System
Max number of Cell Nodes	8	236	>65000	2007	1600	1000
*Bl : Bluetooth

B. Transmission Time

The transmission time depends on the data rate, the message size, and the distance between two nodes. The formula of transmission time in (µs) can be described as follows:

T tx

N data

N..

data ovhd

N maxPld

N data the data size

bit     prop

NmaxPld the maximum payload size

Novhd  the overhead size

T bit     the bit time

Tprop  the propagation time between two nodes to be neglected in this paper

S

E i-

Ё

■

10-

Bleutooth

ZigBee

Wi-Max

GSM/GPRS

10-4

Data Payload Size (Bytes)

From the figure 2, it is noted that the transmission time for the GSM/GPRS is longer than the others, due to its low data rate (168 Kb/s) and its long range reasons, while UWB requires less transmission time compared to the others because of its important data rate. It clearly shows that the required transmission time is proportional to the data payload size data and it is not proportional to the maximum data rate.

Fig. 2. Comparison of transmission time relative to the data size

The typical parameters of the different wireless protocols used to evaluate the time of transmission are given in Table 3.

Table 3. Typical parameters of wireless protocols

Protocol	Max data rate (Mbit/s)	Bit time (μs)	Max data payload (bytes)	Max overhead (bytes)	Coding efficiency+ (%)
Bl	0.72	1.39	339 (DH5)	158/8	94.41
UWB	110	0.009	2044	42	97.94
ZigBee	0.25	4	102	31	76.52
Wi-Fi	54	0.0185	2312	58	97.18
Wi-Max	70	0.0143	2700	40	98.54
GPRS	0.168	5.95	1500*	52*	80.86
+ Where the data is 10 Kbytes.      * For TCP/IP Protocol

C. Transmission power and range

In wireless transmissions, the relationship between the received power and the transmitted power is given by the Friis equation as follows [1], [33], [36-40]:

p

^P t    ^tr

1 4 n D )

Pt  the transmitted power

Pr  the received power

G t   the transmitting omni basic antenna gain

G r   the receiving antenna gain

D the distance between the two antennas λ the wavelength of the signal

From (2) yields the formula the range of coverage as follows:

D =

4 п I P

T V P t G t G r

We note that as the frequency increases, the range decreases. The figure 3 shows the variation of signal range based on the transmission frequency for a fixed power. The most revealing characteristic of this graph is the non-linearity. The signals of GSM/GPRS with 900 MHz propagate much better than ZigBee, Wi-Fi, Bluetooth with 2.4 GHz and UWB with 3.1 GHz vice to vice coverage area.

Fig. 3. Range evolution according to the transmission frequency

x 10-4

Range (m)

Fig. 4. The energy consumption depending on the signal range

D. Energy consumption

The energy consumption for intelligent sensor involves three steps: acquisition, communication, computation and data aggregation. This consumption in the acquisition operation depends on the nature of the application [3]. Data traffic, particularly in the transmission, consumes more energy than the other operations. It also depends on the distance between the transmitter and receiver [4], [5].

According to the radio energy model, [6], [38-44] the transmission power of a k bit message to a distance d is given by:

The predicted received power by an intelligent sensor for each data packet according to the communication range d is given by the Two-Ray Ground and the Friss free space models [3], [35], [40] as follows:

Pr ( d ) = <

PGG,^L (4nd )2 L

^Pt^Gt^Gr^ht ^{2 h}r ²

d ⁴

k.8fs .d 2 + k.E^ k.8amp .d 4 + k.EEiec

d о =

⁸ fs

amp

E Elec      electronic energy

,  4n Lhhkd„ =         r -

^X                                    (7)

L the path loss ht the height of the transmitter antenna hr the height of the receiver antenna d the distance between transmitter and receiver

ε fs , ε amp   amplification energy

The electronic energy E_Elec depends on several factors such as digital coding, modulation, filtering, and signal propagation, while the amplifier energy depends on the distance to the receiver and the acceptable bit error rate. If the message size and the range of communication are fixed, the required energy to cover a given distance increases. The figure 4 illustrates the evolution of the energy consumption for ZigBee protocol based on the signal range. We can say that an increase in data packet size allows then an increase of the transmission energy. Equations (4) and (5) can be generalized for the all wireless mentioned protocols. The simulation parameters are given in table 4.

The figure 5 shows the evolution of the reception power based on the signal range for different studied protocols and for fixed data packet size:

Table 5. The simulation parameters

Parameters	Value
L	1
G t =G r	1
h t =h r	1.5 m

Table 4. The simulation parameters

Parameters	Value
E Elec	50 nJ/bit
ε fs	10 pJ/bit/m2
ε amp	0.0013 pJ/bit/m4

Protocols	Transmitted Power (Watt)
Bluetooth	0.1
UWB	0.04
ZigBee	0.0063
Wi-Fi	1
Wi-Max	0.25
GSM/GPRS	2

According to this figure, it is noted that when the distance between the transmitter and the receiver

increases, the received power decreases, this is justified by the power loss in the path. The ZigBee, UWB and Bluetooth have low power consumption while Wi-Max, Wi-Fi and GPRS absorb more power due to theirs high communication range reason.

-2

0     10    20    30    40    50    60    70    80    90    100

Range (m)

Fig. 5. The received power depending on the signal range with fixed message size

• E.    Chipset power consumption

  To compare practically the power consumption, we are presents in the table 6 the detailed representative characteristics of particular chipset for each protocol [4449]. The figure 6 shows the consumption power in (mW) for each protocol. Obviously we note that Bluetooth and ZigBee consume less power compared to UWB, Wi-Fi, Wi-Max and a GPRS connection. The difference between the transmission power and reception power for the protocols GPRS and Wi-Max is justified by the power loss due to the attenuation of the signal in the communication path since both of these protocols have a large coverage area.

Table 6. Power consumption characteristics of chipsets

Protocols	Chipset	V DD (volt)	I TX (mA)	I RX (mA)	Bit rate (Mb/s)
Bluetooth	BlueCore2	1.8	57	47	0.72
UWB	XS110	3.3	~227	~227	114
ZigBee	CC2430	3.0	24.7	27	0.25
Wi-Fi	CX53111	3.3	219	215	54
Wi-Max	AT86 RF535A	3.3	320	200	70
GPRS	SIM300	3	350*	230*	0.164*
* For GSM 900 DATA mode, GPRS ( 1 Rx,1 Tx )

Based on the data rate of each protocol, the normalized energy consumption in (mJ/Mb) is shown in the figure 7 for a data size equal to 1 Mb. This shows clearly in this figure that the UWB, Wi-Fi and Wi-Max have better energy efficiency.

Fig. 6. Comparison of chipset power consumption for each protocol

Fig. 7. Comparing the chipset normalized energy consumption for each protocol

In summary, we can say that Bluetooth and ZigBee are suitable for low data rate applications with a limited battery power, because of their low energy consumption which promotes a long lifetime. Contrariwise for implementations of high data rate, UWB, Wi-Fi and WiMax would be the best solution due to their low normalized energy consumption. While for monitoring and surveillance applications with low data rate requiring large area coverage, GPRS would be an adequate solution.

• F.    Bit error rate

The transmitted signal is corrupted by white noise AWGN (Additive White Gaussian Noise) to measure the performance of the digital transmissions (OQ-B-Q-PSK, 4PAM, 16QAM, GMSK, GFSK, 8DPSK, 8PSK and OFDM), used in the studied wireless protocols, by calculating the bit error probability. This latter is a very good tool to measure the performance of a modulation used in a communication module and therefore helps to improve its robustness. It is calculated by the following formula:

N

TXBits

N Err the number of errors

N TXBits the number of transmitted bits

The figure 8 shows the BER of the different modulations used in wireless technologies mentioned above according on signal to noise ratio E b /N 0 . The BER

Table 7. E b /N 0 values which cancels BER for the different modulations

Modulation Eb /N0 (dB) B.E.R B-OQ-QPSK 7,8 10-6 GMSK 12,7 10-6 FSK 13,3 10-6 8PSK 13,8 10-6 OFDM 14,3 10-6 16QAM 14,8 10-6 GFSK 15,7 10-6 4PAM 17,6 10-6 8DPSK 22,6 10-6 for all systems decreases monotonically with increasing values of Eb/N0, the curves defining a shape similar to the shape of a waterfall [36], [38]. The BER for QPSK and OQPSK is the same as for BPSK. We note that the higher order modulations exhibit higher error rates which thus leads to a compromise with the spectral efficiency. QPSK and GMSK seem the best compromise between spectral efficiency and BER followed by other modulations. These two robust modulations are used in Wi-MAX, ZigBee, Wi-Fi and in GPRS network. They can be employed in the noisy channels and in the noisy environments.

relationship between the transmission quality and the number of served users.

G. Data coding efficiency

The coding efficiency can be calculated from the following formula:

р cdeff

N

data

< N maxPld )

Bit Error Rate depending on

Eb/No

10-1

10-2

10-3

10-4

10-5

10-6 0

Eb/N0(dB)

Q-OQ-B-PSK 16QAM GFSK GMSK 8DPSK 4PAM FSK 8PSK OFDM

Fig. 8. Bit Error Rate for different modulations

Based on the figure 9, the coding efficiency increases when the data size increase. For small data size, Bluetooth and ZigBee are the best solutions while for high data sizes GPRS, UWB, Wi-Max and Wi-Fi have efficiency around 94%. For point of view application, the automation industrial systems based on intelligent sensors, since most data monitoring and industrial control have generally a small size because they don't require an important data rate such as pressure or temperature measurements which don't exceeds 4 bytes, Bluetooth, ZigBee and GPRS can be a good choice due to their coding efficiency and their low data rate. On the other hand, for applications requiring a large cover zone as the borders monitoring, the persons tracking or the environmental monitoring or the event detection, GPRS and Wi-Max are an adequate solution, whereas for the multimedia applications requiring an important data rate such the video monitoring, Wi-Fi, UWB and Wi-Max form a better solution.

However, because of their sensitivity to noise and nonlinearities, the modulations 4PAM and 8DPSK remain little used compared to other modulations. Concerning the QAM modulation, it uses more efficiently the transmitted energy when the number of bits per symbol increases. As for the frequency hopping FSK modulations, the increase of the symbols will enable reduction of the BER but also increase the spectral occupancy. The main fault of these FSK modulations is their low spectral efficiency. On the other side, the GMSK modulation has been developed in order to increase the spectral efficiency [50]. It has a satisfactory performance in terms of BER and noise resistance [9], [35-41]. Furthermore, the lower bit error probability is obtained to the detriment of the number of users. It can be helpful to investigate the

Fig. 9. Coding efficiency depending on data size

• H.    Packet error probability

The packet error probability is one indicator which can be useful to know QoS level, if packet error is detected an error-correction mechanism is activated by retransmission of the faulty packet at the MAC layer. This process improves system reliability, but increases delay. It can be calculated as follows [52]:

b e bit error probability L packet length in bits

Packet Length (Bytes)

Fig. 10. Packet error depending on packet length

Referring in figure 10, it is obviously clear that increasing L may lead to higher packet error probability and therefore a higher number of retransmissions. Depending on the bit-error probability and packet overhead, packet length L can be optimized so that an optimal system lifetime can be achieved.

• I.    Energy index

0.01 0.002

Energy index is used to measure the transmission energy cost of a sensor node and therefore to determine life span of the sensors network. It is given as follows [52]:

be=0.01

be=0.005

be=0.0001

0 10 20 30 40 50 60 70 80 90 100110120130140150

Packet Length (Bytes)

Fig. 11. Energy index versus packet length

It can be observed that in fact there is an optimal value of packet length L which maximizes E i . The optimal value of L increases with a lower bit error rate and/or with an increase in the packet overhead O. When designing and deploying a wireless sensor network, it is possible to choose the appropriate values of the studied parameters that will extend system lifetime.

• J.    Real time throughput

The real time throughput is defined to know the number of octets per second transmitted on a specific link exclusively relevant to real-time traffic [52]. It analytical expression relevant to the link between the devices is the following:

n r number of retransmissions

Th =----m---- frame + backoff

O packet overhead et transceiver energy em collision, idle and overhearing energy

m     the amount of data to be transmitted

T frame time required to transmit a MAC data frame

T backoff the average backoff time

Figure 11 presents numerical results of E i as a function of packet length when:

*

O=2 bytes

em =200 nJ

et =100L nJ       ec =100 nJ

к

m=512 bytes m=1024 bytes

Ш

0.01

0.008

0.006

0.004

0.002

0 10 20 30 40 50 60 70 80 90 100110120130140150

Packet Length (Bytes)

Fig. 11. Energy index versus packet length

$ \

p

20       40       60       80       100

Backoff Time (ms)

Fig. 12. Throughput versus backoff time

Figure 12 shows the simulation results for two values of transmitted data m=512bytes and m=1024bytes with a fixed sensor node bandwidth equal to 38,4Kbits/s and with T frame =11,39ms, It seems that real time throughput needed to convey data decrease when transmitted data size decrease. The behavior of Th has a high bound and decrease suddenly for the lowest value of T backoff ,

thereafter it reduce asymptotically with growing values of T backoff . This fall which occurs in throughput is due to the bandwidth waste caused mainly by collisions and backoff delays when two or more sensor nodes attempt to transmit measurements at the same time which can produce saturation phenomena.

when designing a wireless microsensor mote because energy is wasted while no work is done.

K. Microcontrollers energy efficiency

Energy efficient computing during the sensing algorithm process depends on the number of instructions executed by the software implemented into microcontroller of the sensor motes [53]. This energy is the sum of energy required to switch between the internal states of the MCU given by (14) and the leakage energy which refers to the energy lost while the MCU is idle modeled by (15).

Е , = Е .. . +Е, , total       switch       leakage

switch

V ² total dd

leakage

dd

N1 f)

Where N is the number of cycles which the program takes to be executed, f is the MCU clock frequency and V T is the thermal voltage. Figure 13 shows comparison between the behaviors of the energy consumed per instruction in (J) depending on the number of instruction cycles for types of sensor motes. The energy consumption increases with the augmentation of the number of instruction cycles. This consumption is very important for an MCU working with lowest clock frequency. MCU leakage energy is an important parameter to optimizing

V. Sensors Motes Comparaison

While the particular sensor types vary significantly depending on the application, a variety of wireless modules and hardware platforms have been realized to facilitate developing applications in intelligent sensors networks [51]. Figure 13 summarizes a comparison of popular sensor motes that were designed in the recent few years. As can be observed, the capabilities of these platforms vary significantly. However, this will assist researchers, developers and users to select the most appropriate platform for their purposes.

Fig. 12. Popular sensor motes comparison

• VI. Conclusion

We have presented in this paper a comparative performance analysis of six wireless protocols: Bluetooth, UWB, ZigBee, Wi-Fi, Wi-Max and GSM/GPRS. Choosing a wireless protocol for an intelligent sensing application must do compromise between the energy cost, QoS and real time execution. However, robust sensing quantitative evaluation indicators permitted us to determine the suitable protocol for an application based on intelligent sensor. Furthermore, the adequacy of the sensor networks is influenced strongly by many others factors as the network reliability, the link capacity between several networks having different communication protocols, the security, the chipset price, conformity and installation cost. The challenge is to develop a gateway (multi-standard transceiver) that enables data exchange between these heterogeneous infrastructures with a good QoS. As perspective of this work, it is proposed to extend this study to investigate the impact of modulation scheme on the sensing coverage, on the bandwidth, on the number of served users and on energy consumption.