Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Efficient Wireless

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

UNIWERSYTET TECHNOLOGICZNO-PRZYRODNICZY

IM. JANA I JDRZEJA NIADECKICH W BYDGOSZCZY


TELECOMMUNICATIONS AND ELECTRONICS. SCIENTIFIC JOURNAL

TOWARD MORE EFFICIENT WIRELESS LAST MILE


SMART GRID COMMUNICATION SYSTEMS
Piotr Kiedrowski1
Gabriel A. Deak2
Kevin Curran2
Joan Condell2
1

Institute of Telecommunication, University of Technology and Life Science


ul. Kaliskiego 7, 85-789 Bydgoszcz, Poland
2
ISRC - University of Ulster, Magee Campus, Northland Rd.,
BT48 7JL, Derry, UK
Summary: Multi-hop techniques are very popular in Wireless Sensor
Networks as they are used to solve coverage problems. The drawback of these
techniques is the low reliability. The aim of this paper is to outline hardware
solutions that improve the efficiency of wireless networks intended for last mile
smart grid communication systems. Typical Wireless Sensor Networks cannot
take advantage of these solutions due to power constraints. Fast carrier detection
and the probability of collisions, forward error correction, data whitening, and
clear channel assessment are discussed as hardware solution to improve wireless
network efficiency for last mile smart grid communication systems. The
hardware solutions significantly improve the efficiency of the wireless network
intended for last mile smart grid communication systems.
Keywords: smart grid, wireless sensor network, last mile communication, PER vs.
RSSI

INTRODUCTION
Successful implementation of Smart Grid (SG) solutions depend mostly on the
adopted communication solutions [1]. Currently, the main problem in the
implementation of Smart Grid communication systems is the solution of the
"last mile", part of SG communication networks. This part of networks has a
decisive impact on the cost of implementation and operation of the whole
communication system. Thus very cheap sensor-based devices have been used
to deploy last mile wireless networks. Wireless Sensor Networks (WSNs) use
similar hardware solutions.
A typical node in a WSN consists of an antenna, a microcontroller, a
transceiver and a sensor. In Europe, microwaves with frequencies 433 MHz and

868 MHz (ISM Industrial, Scientific, Medical radio band) are very often used
for automatic meter reading (AMR) systems and WSNs. Nodes using these
frequencies have a maximum transmission power of approximately 10 dBm and
a short transmission range. In order to increase the transmission range a multihop technique is often used. Thus a node in a network may act as a source node,
destination node, or a transfer node. The only difference between typical WSN
nodes and SG nodes is the power supply. WSN nodes are powered by small
batteries whilst SG ones are powered directly from mains electricity monitored
by this node. If there are no power constraints we can implement different
system solutions e.g. communications protocols or topology control. System
solutions are mainly realized by software methods. However, the best results in
improving the efficiency of the network are obtained by using the software and
hardware solutions simultaneously.
In this paper we have proposed four hardware solutions such as fast
carrier detection, forward error correction, data whitening and clear channel
assessment. The hardware solutions significantly improve the efficiency of the
wireless network intended for last mile smart grid communication systems.
These solutions cannot be applied in typical WSN where the nodes have power
constraints.

1. FAST CARRIER DETECTING AND COLISSIONS PROBLEM


Using a multi-hop technique, the reliability of the path between source and
destination nodes is based on individual reliability of all nodes in the path. The
Packet Error Ratio (PER) value is high for communications between sensorbased devices compared to other types of communication media e.g. optical
fiber.
Multi-hop techniques are very popular in WSN. These techniques solve
the coverage problem. However wireless communications using these
techniques have low reliability. A solution to cope with the high PER value is
fairly easy to implement when we do not have to consider energy efficiency. A
multipath routing protocol [2], using more than one path between source and
destination nodes at the same time, could fix the problem. Multipath routing
protocol is wasting more energy but this is not a problem in proposed
application. However, generating many redundant transmissions still seems to
be a problem. This can be evaluated as the probability of sending two or more
packets at the same time, using the equation below:

T T
T

1 ... k 1
t
t

PC k , T , t 1
k
T

t

(1)

where:
- k is the number of nodes in a cluster, transferring a packet in a defined
hop,
- t shows how long the CD (Carrier Detect) status cannot be monitored,
- T is a maximum value of random time delay in the process of packet
reliability.
The value of Pc could be decreased by increasing the value of T if there was no
influence on the communication time. Similarly, Pc could be decreased by
decreasing the value of k, but we have no influence on the network topology.
The only possibility to decrease the value of Pc is to reduce the time t. Figure 1
presents the probability of collisions for three values of t based on the number
of nodes k.

k
1
1

11

21

31

41

51

61

71

81

91

0.1
0.01

Pc

0.001

0.0001
0.00001

t=5 ms
t=0,033 ms
t=0,002 ms

0.000001

Fig.1. Probability of collisions vs. the number of nodes in transferring cluster

Equation 1was used to obtain the results in Figure 1 for a constant value of T
equal to 1 s. The value of t is the time needed to detect status of carrier (if
channel is busy or free). Usually radio modules have clocks factory set only to
32 kHz. This is because radio modules are mainly intended to work in WSNs,

where the power conservation is the main issue. In this case, when clock is
slow, the time needed to detect CS is long e.g. 5 ms if it is readout via SPI
(Serial Peripheral Interface). When the same module is overclocked up to 5
MHz the information about CS is achievable in about 33 s. The last example,
which is also presented in Fig. 1, concerns the same module with both the
overclocked microcontroller and additionally with the CD output pin of the
transceiver connected directly to the IRQ input pin of the microcontroller. With
such modifications t=2 s. Assuming that the probability of a collision cannot
be greater than 0.01, the maximum number of nodes transferring the same
packets should not exceed 9 if hardware solutions allow microcontrollers to
detect a carrier with sense of 5 ms. The main reason of using the 0.01 threshold
value is the fact that many applications use this value as the unreliability
measure in order to calculate threshold parameters. In [3] the value PER=10-2
was obtained using this threshold value as the probability of collision. This time
is too short considering that 9 nodes or bigger clusters occur very often. By
using a module with an overclocked microcontroller, we can detect carrier
faster. In this case we have the possibility to obtain a probability of collision
less than 0.01 if the size of transferring cluster is smaller than 27 nodes. The
case when the number of nodes in a cluster is greater than 27 is rare. The
whole subnet (managed by one sink) contains at most a few hundred nodes
spread over a large area. Thus, it can be concluded that in this case, when the
network is merely energy meters, using an overclocked radio module is
sufficient when multi-hop and multipath transmission technique are used. In the
future if SG systems usage will increase then the number of nodes in the
network will increase several times. In this case, the third solution will be
necessary to ensure a low probability of collisions in clusters consisting of more
than one hundred nodes.

2. HARDWARE FORWARD ERROR CORRECTION


Most of the one-chip transceivers on the market have built in support for
forward error correction (FEC). By using this option the energy consumption
increases. This is the main reason why FEC is not a very popular technique for
improving the transmission parameters in typical WSN solutions. FEC is a good
choice when energy efficiency is not a concern.
In order to see if this option improves the efficiency of transmission we
ran an experiment where we alternate sending and receiving packets on the
network - odd packets with FEC, even without FEC.
The efficiency of transmission can be assessed by various methods. The
most popular methods are the quality of service assessment and the reliability of
links measurement. The quality of service method is a direct method and is
dedicated to specific solutions e.g. implemented communication scheme. The

second method is more general and estimates the quality of service when the
implemented solutions are known. The efficiency of transmission with FEC was
assessed using these two methods.
As a measure of the quality of service we have considered the data
collection speed (number of readings from nodes per time interval) for all the
nodes in our test network. The speed increased with 9% using FEC compared to
the case when FEC was off.
We have also measured the PER versus Received Signal Strength
Indicator (RSSI) in the neighbourhood of the sink, based on all data received by
the sink. We cannot plot both PERFEC(RSSI) and PERnoFEC(RSSI) in the same
graphic as curves overlap. During long-term tests the results were always better
using FEC compared to the case without FEC. The global values of PER
(regardless of RSSI) were as follow: PERFEC=9.9210-2and PERnoFEC=1.0810-1.
We expected that the use of FEC would improve the transmission.
However the difference compared to the case without FEC was not large
enough to help us assess the transmission based on PER. Such a situation is the
effect of the implemented communication scheme on the function of the relative
PER growth versus RSSI, expressed by the following equation:
(

(
(

)
)

(2)
Figure 2, obtained using (2), shows the influence of FEC on the transmission
reliability.

PERnoFEC, FEC(RSSI)

0.2
0.15
0.1
0.05
0
-115 -110 -105 -100

-95

-90

-85

-80

-75

-70

-65

-60

-0.05

Received Signal Strength Indicator [dBm]


Fig. 2. The influence of FEC on the transmission reliability - relative PER
versus RSSI

Four power ranges can be distinguished analysing the data in Fig. 2.The
first power range is for RSSI values less than -113 dBm. In this range all
packets have errors with or without using FEC.
The second power range is for RSSI values between -113 dBm and 99 dBm. In this range the value of the relative PER is less than zero. This means
the FEC function does not improve transmission but worsens it instead. This
situation can be easily explained taking into account the range of RSSI. The
signal received by a sink is very weak which means the nodes are located on the
edge of their radio range. When the FEC is switched off, nodes located on the
edge do not receive packets from the sink or receive packets with errors, thus do
not send copies. However, when the FEC is switched on, some nodes receive
error free packets from the sink and send copies to other nodes. This improves
the communication but does not improve the relative PER measured at the sink
point.
The third power range is for RSSI values between -99 dBm and 82 dBm. In this range the value of the relative PER is greater than zero and it
goes up to nearly 20% for -90 dBm, thus the FEC function improves
transmission.
The fourth power range is for RSSI values greater than -82 dBm. In this
range all packets are error free and do not depend on the status of the FEC
function.
Concluding, we can say that only in the power range between -99 dBm
and -82 dBm is the FEC function significantly effective. We can also say that in
the power range between -113 dBm and -99 dBm, FEC is only seemingly
ineffective. In other power ranges FEC function does not affect the PER.
Nevertheless, the FEC function can be very useful under conditions of
interferences especially in the fourth power range. This phenomenon was
explained based on the observations presented in [4].

3. DATA WHITENING
In the ideal case, over many different types of media, random DC-free (Direct
Current) data is transmitted. In many cases the data transmitted contains long
sequences of ones and/or zeros. Performance can then be improved by
whitening the data (DW) before transmitting and de-whitening at reception.
The data is prepared by applying exclusive-or (XOR) with a pseudorandom sequence before transmission. Results in this paper were obtained using
pseudo-random bit sequences generated by the polynomial function
G(x)=x9+x5+1. This method is called data whitening. When the data is received,
it is XORd with the same pseudo-random sequence to obtain the original data.
Figure 3 presents a circuit for data whitening:

Fig. 3. Data whitening circuitry [5]

Figure 3 shows a circuit for data whitening similar to the scrambler circuit
implemented in SONET/SDH optical communication technology to realize the
same function as a line coding i.e. avoiding of long sequences of zeros or ones.
Using a similar technique as in the examination of the FEC, i.e.
alternate sending and receiving packets we have analysed the improvement on
the transmission efficiency by alternating the usage of DW with the case when
DW is not used. However the DW function had no effect on the transmission
efficiency, either on readings speed or PER
One should not consider our results as general due to the specific
telemetric data organized in packets. The packets, which carried telemetric data,
consist of the overhead and the data field. The data fields are encrypted so they
are already random and do not need to be whitened. The overhead consist
mostly of long addresses which rarely contain long sequences of zeros and
never contain long sequences of ones, where an address is a real serial number
of a device coded in BCD.

4. CLEAR CHANNEL ASSESSMENT


The Clear Channel Assessment (CCA) is used to indicate if the channel is free
or busy, during the transmitting process. There are different criteria for
determining whether the channel is free or busy such as: RSSI is below
threshold, receiver currently is receiving a packet etc.
The CCA function is very useful if the point-to-point transmission
mode is used with a half-duplex transmission channel. The CCA function can
be also useful when we use not advanced routing protocols e.g. flooding [6] or
gossip [7], whilst when we use advanced routing protocols e.g. energy greedy

quasi-flooding (EGQF) routing protocol [8], which is used in our test network,
the usage of the CCA is not recommended.
This is because, even though, during the transmitting process with CCA
enabled, the receiver is in RX state, it cannot receive any packet but just wait
for an opportunity to transmit the own, hold in TX buffer, packet as soon as
channel is free. Using advanced transmission techniques, the information which
is waiting to be sent can be out-of-dated by received one. So, we recommend
using the CCA feature only for a simple e.g. ping-pong transmission systems in
order to shorten the time of an idle state required between receiving and
transmitting states.
Examining the CCA in our test network, the CCA function had no
effect on the transmission efficiency but had a negative effect of increasing the
average rate of transmitted packets per query.

CONCLUSION
Most of the single chip transceivers have many internal modules which can be
programmatically switched on or off. However the drawback of using additional
modules is increased power consumption, therefore developers often have to opt
out of using additional/optional possibilities offered by modern communication
chips.
In the case of systems with no energy deficit (e.g. SG) one can use other
solutions, both hardware and software. Here, however, the risk is that the
designers decide to include all internal modules aimed at improving the
transmission parameters. The article shows that by deciding to include any
additional function, one has to take into account not only energy consumption
but also the protocols used by the system such as communication protocols.
Transmission parameters and quality of service were compared for two
cases, the first with optional functions enabled and the second with optional
functions disabled. Comparison of transmission parameters and quality of
service with all features enabled or disabled is not presented. Such an
experiment was not performed, because a solution that has the greatest impact
on transmission parameters and network performance is a fast carrier sensing.
Other functions have a significant effect only when the CS detection is not fast
enough. Therefore a fast carrier detecting solution is the most important and
must be taken into account at the stage of hardware design. Usage of other
solutions depends strongly on the system to be implemented and the
characteristics of this system should be considered before switching on optional
features.

REFERENCES
[1] IEEE SmartGridComm 2011 Concludes, Successful 2nd Annual Event in
Brussels, Belgium, IEEE Communications Magazine, - Conference Raport,
December 2011, pp. 18-19
[2] Al-Karaki A.N., Kamal A.E., Routing Techniques in Wireless Sensor
Networks: a Survey, IEEE Wireless Comm., Dec. 2004 (2004), 6 28
[3] Texas Instruments (2006) CC1100 Single Chip Low cost Low Power RF
Transceiver. Texas Instruments Incorporated: 89-90
[4] Boryna B., Dubalski B., Kiedrowski P. and Zabudowski A.: Image
Processing and Communications Challenges 2 (ed. Ryszard S. Chora), Errors
Nature in Indoors Low Power 433 MHz Wireless Network, 2010 Springer
Verlag, Advances in Intelligent and Soft Computing 84, pp. 373-378;
[5] Grant Christiansen: Data Whitening and Random TX Mode, Texas
Instruments, 2010
[6] Durresi A., Paruchuri V.K., Iyengar S.S., Kannan R.: Optimized Broadcast
Protocol for Sensor Networks, IEEE Transactions on Computers, Vol. 54, No.
8, August 2005, pp. 1013-1024
[7] Boyd S., Ghosh A., Prabhakar B., Shah D.: Randomized Gossip Algorithms,
IEEE Transactions on Information Theory, Vol. 52, No. 6, June 2006, pp. 25082530
[8] Kiedrowski P., Dubalski B., Marciniak T., Riaz T. and Gutierrez J.: Image
Processing and Communications Challenges 3 (ed. Ryszard S. Chora), Energy
Greedy Protocol Suite for Smart Grid Communication Systems Based on Short
Range Devices, 2011 Springer Verlag, Advances in Intelligent and Soft
Computing 102, pp. 493-502;

You might also like