FOUNTAIN CODES FOR RELIABLE DATA TRANSMISSION IN

LOW VOLTAGE POWER-LINE NETWORKS

Juan Pedro Mu˜noz-Gea, Pedro Jose Pi˜nero-Escuer, Josemaria Malgosa-Sanahuja

Pilar Manzanares-Lopez and Juan Carlos Sanchez-Aarnoutse

Department of Information Technologies and Communications, Polytechnic University of Cartagena

Campus Muralla del Mar, 30202, Cartagena, Spain

Keywords:

Fountain codes, Online codes, Power-line communication, HomePlug AV, Home networking.

Abstract:

Home appliances are becoming information devices which can be networked to exchange information. No-

new-wires networks are the most promising candidate networking technologies to provide both residential

and Small and Medium Enterprise (SME) networking services. Among them, broadband communication over

power line networks has attracted much interest in academy and industry recently. The HomePlug Powerline

Alliance has developed a new speciﬁcation for in-home networking called HomePlug AV. The HomePlug AV

MAC layer provides a connectionless, prioritized contention service based on CSMA/CA to support best-effort

applications. When there are connections competing for the channel utilization, UDP connections lose a big

amount of packets. In addition, the bidirectional nature of TCP protocol may not be the adequate transmission

mechanism in a half-duplex channel. In this scenario, the use of Fountain codes could be a good alternative

to UDP and TCP to transmit reliable data. Online Codes are a free-software Fountain codes version. In this

work, it is evaluated the feasibility to use Online Codes for binary data transmission in a low-voltage PLC

network.

1 INTRODUCTION

Nowadays, home appliances are becoming informa-

tion appliances and they can be networked to ex-

change their information. Therefore, it is necessary

a home network able to provide support for video

and data transmission from a variety of sources in the

home. Candidate networking technologies to provide

convenient and widespread residential networking

services may be categorized as wireless, wired and

no-new-wires networks. For the no-new-wires net-

works category, broadband communication over low

voltage (220v) power lines or PLC (Power Line Com-

munications) has attracted much interest in the aca-

demic and industrial context recently. The HomePlug

Powerline Alliance (Afkhamie et al., 2005) has de-

veloped a new speciﬁcation for in-home networking

called HomePlug AV (HomePlug Audio and Video,

or simply HPAV). HomePlug AV employs advanced

physical and medium access control (MAC) technolo-

gies that provide a 200 Mbps power line network.

The physical layer utilizes this 200 Mbps rate to pro-

vide a 150 Mbps information rate. Nowadays, there

are about 70 certiﬁed HomePlug products (HomePlug

Certiﬁed Products, 2009). They range from a simple

HomePlug Ethernet adapter (which connects the Eth-

ernet devices to the HomePlug network) to 1 Gbps

high performance low-voltage PLC modem (Gigle,

2009). On the other hand, nowadays some companies

are adding HomePlug circuits directly into multime-

dia home entertainment equipment. By this way, it

will not be necessary any additional equipment (like

Ethernet adapters) to connect the electrical appliances

to the PLC network

However, it is necessary to take into account that

there are several aspects of the PLC medium that

make it difﬁcult to share resources fairly. For exam-

ple, all the electronic or electrical equipments con-

nected to the power lines are considered as noise re-

sources on the power grid (Jensen and Kjarsgaard,

2007). Therefore, our ﬁrst goal in this work is to char-

acterize the network behavior in presence of electrical

equipment of a typical home when using a commer-

cial HPAV modem.

135

Pedro Muñoz-Gea J., Jose Piñero-Escuer P., Malgosa-Sanahuja J., Manzanares-Lopez P. and Carlos Sánchez-Aarnoutse J. (2009).

FOUNTAIN CODES FOR RELIABLE DATA TRANSMISSION IN LOW VOLTAGE POWER-LINE NETWORKS.

In Proceedings of the 4th International Conference on Software and Data Technologies, pages 135-140

DOI: 10.5220/0002252001350140

 SciTePress

Another important aspect in PLC networks is

that, although the HPAV MAC layer provides a

connection-oriented service based on TDMA to sup-

port QoS requirements, almost all commercial HPAV

modems only support traditional connectionless, pri-

oritized contention service based on CSMA/CA to

transmit best-effort applications and applications that

rely on prioritized QoS. In other words, the PLC chan-

nel is broadcast (i.e. shared) and half-duplex by na-

ture. This fact made us think that the use of unidi-

rectional rateless codes to transmit the information

may be advantageous as opposed to traditional solu-

tions based on bidirectional (full-duplex) TCP pro-

tocol. Rateless codes are also known as Fountain

codes because the sender is theoretically always send-

ing coded packets. Some of them are lost in the noisy

channel but when the destination node has received

enough of them, the information can be completely

restored. They are mainly used in multimedia applica-

tions, such as IPTV or radio broadcasting, because in

these cases there is usually no feedback channel, and

it is impossible to take advantage of any type of er-

ror or ﬂow control. Online Codes are a free-software

Fountain codes alternative, which achieve linear cost

for encoding and decoding operations. In this work,

the feasibility to use Online Codes for reliable data

transmission in an in-home PLC network is evaluated.

The remainder of the paper is organized as fol-

lows. Section 2 presents the HomePlug AV speciﬁca-

tion. Section 3 introduces the encoding and decoding

procedure used by Online Codes. Section 4 evaluates

the use of Online Codes for data transmission in a real

scenario. Finally, Section 5 concludes the paper.

2 HOMEPLUG AV

As it was previously introduced, there are several as-

pects of the PLC medium that make it difﬁcult to

share resources fairly. In order to solve these prob-

lems, advanced coding and modulation mechanisms

are used. The Physical Layer (PHY) operates in the

frequency range of 2 - 28 MHz and provides a 200

Mbps channel rate and a 150 Mbps information rate.

It uses OFDM and a powerful Turbo Convolutional

Code (TCC). OFDM is a spectrum efﬁcient modu-

lation technique, which uses simultaneous transmis-

sion of a large number of narrow band carriers. These

carriers divide a large frequency channel into a num-

ber of subchannels. Subchannels can differ greatly

in their quality, deﬁned by their signal to noise ratio-

SNR-. Adaptive coding modulation for each subchan-

nel solves this problem by giving each subchannels an

appropriate capacity, and by switching off those with

a poor channel condition.

HomePlug AV provides two kinds of commu-

nication services: a connection-oriented contention

free service, based on periodic Time Division Mul-

tiple Access (TDMA) allocations of adequate dura-

tion, to support the QoS requirements of demand-

ing applications; and a connectionless, prioritized

contention service, based on Collision Sense Multi-

ple Access/Collision Avoidance (CSMA/CA), to sup-

port both best-effortapplications and applications that

rely on prioritized QoS. To efﬁciently provide both

kinds of communication service, HomePlug AV im-

plements a ﬂexible, centrally-managed architecture.

The central manager is called a Central Coordinator

(CCo). The CCo establishes a beacon period and a

schedule which accommodates both the cotention free

allocations and the time allotted for contention-based

trafﬁc. The Beacon Period is divided into 3 regions:

beacon region, CSMA region and TDMA region.

3 FOUNTAIN CODES

3.1 Description

The main idea behind Fountain codes is that the trans-

mitter is represented like a fountain of water that is

able to produce an inﬁnite number of water drops.

The receiver represents a bucket that needs to col-

lect a number of these water drops to obtain the in-

formation. The main advantage of these codes is that

the receiver can obtain the information irregardless

of which drops it has collected. Therefore, Fountain

codes should have the following properties:

• A transmitter can generate a potentially inﬁnite

amount of encoded packets from the original data.

• A receiver can decode a message that would re-

quire K packets from any set of K

′

encoded pack-

ets, for K

′

slightly larger than K.

The most important implementations of Foun-

tain codes are LT codes (Luby, 2002), Raptor codes

(Shokrollahi, 2006) and Online Codes (Maymounkov

and Mazi`eres, 2003). LT codes were the ﬁrst prac-

tical realization of a fountain code. The only draw-

back of these codes is that their encoding and decod-

ing costs scale as Klog

K, where K is the ﬁle size.

Raptor codes are an evolution of LT that achieve lin-

ear cost for encoding and decoding. Finally, Online

Codes are a free-software alternative to Raptor codes

that also achieve linear cost for both operations.

There are lots of application of Fountain codes

in digital communications. They are mainly used

in multimedia applications, such as IPTV or radio

ICSOFT 2009 - 4th International Conference on Software and Data Technologies

136

Figure 1: Structure of Online Codes.

broadcasting because in these cases there is no feed-

back channel. Therefore, it is impossible to take ad-

vantage of any type of error or ﬂow control. Another

kind of service that could also use these codes is mul-

ticast transmission. When a transmitter sends a ﬁle to

different receivers, each of these receivers could ex-

perience independent losses, delays, jitter, etc. With-

out the Fountain codes facility, the number of control

packets needed to maintain the multicast connection

could be very high.

3.2 Online Codes

Online Codes are characterized by two parameters ε

and q. The ﬁrst parameter is related to the number

of coded blocks (also called check blocks) that the

receiver needs in order to decode the original mes-

sage, and the second one has an effect on the proba-

bility of succesful decoding. In particular,the receiver

can recover the original message from any (1+ 3ε)K

check blocks with a success probability determined

by 1− (ε/2)

q+1

The structure of Online Codes is depicted in Fig.

1. The encoding process is divided into two layers,

the inner code and the outer code. The inner code is

in charge of generating the check blocks. Every check

block is computed as the XOR operation of d blocks

uniformly chosen from the message (d represents the

degree of the check block). The probability that d = i

is given by the probability distribution ρ

= 1−

(1+ 1/F)

(1+ ε)

(1)

(1+ ρ

(F − 1)i(i− 1)

i = 2,3,...,F (2)

where F is given by

F =

ln(ε

/4)

ln(1 − ε/2)

(3)

However, due to the random selection of the mes-

sage blocks, some of them may not be selected in the

inner coding process. One solution to this problem is

to add a preliminar coding process (called outer cod-

ing) which generates 0.55qεK auxiliary blocks from

the original message. The message blocks that do not

participate in the inner process will be able to be de-

coded thanks to this redundancy. In fact, the input

blocks of the inner coding are the message blocks plus

auxiliary blocks. All this set is called composite mes-

sage.

From all the available procedures to gener-

ate redundancy (Reed-Solomon, Cyclic Redundancy

Check, Parity Bits, etc.), Online Codes uses one of

the simplest: for each block of the original message,

q auxiliary blocks are chosen. Each auxiliary block is

computed as the XOR operation of the original mes-

sage blocks assigned to it.

The original message can be decoded from the

check blocks, with the success probability showed be-

fore. The decoding process is also divided into two

steps. In the ﬁrst step a 1 − ε/2 fraction of compos-

ite message blocks should be recovered. The knowl-

edge of this fraction of blocks is enough to decode the

original message: the composite message has such a

property thanks to the redundancy added by the outer

code. The second step consists of recovering the orig-

inal message from the composite message blocks re-

covered in the ﬁrst step.

In order to successfully recover the needed frac-

tion of composite message blocks, it is necessary to

get to know the degree of each check block and the

composite message blocks from which it is made up

(also called adjacent blocks). A way to send this in-

formation to the receiver must be implemented on

the transmitter side. However, if the receiver uses

the same random number generation algorithm as the

transmitter, it will only be necessary to send the seed

to reach this objective. Next, the decoding process

can start. It has the following steps:

1. Find a check block with only one adjacent block

(d = 1) and recoverthis composite message block.

2. Remove this recovered block from other check

blocks that also have this recovered block as adja-

cent (by simple substracting it; that is, computing

the XOR again). After this, some check blocks

can become degree-one blocks.

3. Continue with this process until a 1−ε/2 fraction

of composite message blocks is recovered.

The process can fail if in some of these steps there

are no degree-one blocks.

When all the needed composite message blocks

are recovered, the same process can be used to obtain

the original message blocks. In this case the success

probability is close to one because only the auxiliary

blocks have a degree higher than one.

FOUNTAIN CODES FOR RELIABLE DATA TRANSMISSION IN LOWVOLTAGE POWER-LINE NETWORKS

137

4 EVALUATION

In this section we want to evaluate two characteris-

tics of HomePlug AV speciﬁcation: First, the variable

capacity model of the physical layer; and second, the

contention-based service of the MAC layer. All the

evaluations presented in this section have been made

using a laboratory test-bed. The lab has three phases

of 220 volts, and all the PLC adapters and PC comput-

ers used in the evaluation are connected to the main

phase. We have used the PLE200 HomePlug AV Eth-

ernet adapters of Linksys (Cisco-Linksys, 2009). This

adapter connects the Ethernet device of a computer to

the 220 V power line.

4.1 Evaluation of the Variable Capacity

of the Physical Layer

Our ﬁrst objective is to evaluate the adaptation of the

data transmission rate according to the noise level.

For this aim, we use two computers connected to the

HomePlug AV network using the corresponding Eth-

ernet adapters. Their power suppliers are connected

to another electrical phase of the lab, different from

the one used to implement the HomePlug AV net-

work. Knowing that in Europe the common point of

two distinct phases is located in the low voltage trans-

former, the distance between two elements connected

to different phases is around 300 meters. This dis-

tance is large enough to ensure the absence of inter-

ference among devices connected to different phases.

In order to measure the capacity of the network,

we used a UDP trafﬁc generator which transmits traf-

ﬁc to the maximum capacity allowed by the physical

network. We achieved the results presented in Fig.

2. In particular, it can be observed that the network

is able to achieve a maximum capacity around of 80

Mbps (which is nearly to the maximum capacity as-

signed to the CSMA region). Next, after 8 seconds,

we connected the power suppliers of all the avail-

able computers in the lab to the electrical phase of

the HomePlug AV network. In Fig. 2 it can be ob-

served that the connection of all the computers to the

power line causes the reduction of the network capac-

ity to 60 Mbps. After 17 seconds, we disconnected

all the computers from the electrical line and we de-

tected that the transmission rate increased up to 80

Mbps again. The reason for this phenomenom is that

the powerful coding and modulation technique used

in HomePlug AV is able to adapt the transmission

speed according to the noise level generated by the

electrical devices in order to avoid the losses of pack-

ets.

5e+07

5.5e+07

6e+07

6.5e+07

7e+07

7.5e+07

8e+07

8.5e+07

0 5 10 15 20 25

Throughput [Bits/sec]

Time [secs]

Figure 2: Evaluation of the network capacity.

4.2 Evaluation of the Contention-based

Service of the MAC Layer

As it was previously introduced, the HomePlug AV

MAC layer offers a contention service based on

CSMA/CA. In packet communication networks, con-

tention is a media access method that is used to share

a broadcast medium. In this method all the hosts con-

nected to the medium compete in order to transmit

to the broadcast medium and they can only do that

when the channel is “idle”. In order to evaluate this

characteristic of HomePlug networks, in our next ex-

periment we establish three simultaneous (compete

against one another) ﬁle transmissions using UDP

connections, and we evaluate the performance of one

of them. The size of the transmitted ﬁle varies from 1

up to 20 Mbytes and the distance between the PLC de-

vices of the measured connection is about 45 meters.

This is aproximately the longest distance between two

PLC devices in a real in-home scenario (worse case).

The most remarkable result is the big proportion of

lost packets, represented in Table 1. It means that

the UDP receiver is not able to receive the full ﬁles.

These losses are mainly due to the buffer overﬂow at

the PLC interface because the incoming packet rate

is greater than the outgoing rate at which CSMA/CA

can transmit packets. Therefore, in this environment

the use of UDP applications which do not implement

any ﬂow control is not recommended for reliable data

transmission.

In order to avoid the losses of packets in a con-

tention environment,like a HomePlug networks, there

are two options: ﬁrst, using a protocol that imple-

ments a ﬂow control mechanism, for example TCP

(Transmission Control Protocol); second, using an ap-

plication that implements some kind of forward error

correction, for example, the use of Fountain codes. In

the following experiments, we want to compare the

ICSOFT 2009 - 4th International Conference on Software and Data Technologies

138

Table 1: Proportion of lost packets in UDP connections.

File size [MB] Losses [%]

1 0

2 16.66

3 36.5

4 38.17

5 30.84

6 40.77

8 34.73

10 41.45

15 37.25

20 37.73

performance achieved by two ﬁle tranmission appli-

cations, one of them uses TCP as transport protocol,

and the other one uses a kind of Fountain codes (con-

cretely Online codes).

We used the scp program (scp, 2009) as a TCP-

based ﬁle transmission application. On the other

hand, in order to implement the application based on

Online codes, we used the Online Codes implemen-

tation available in (Implementation of Online Codes,

2009). UDP transmission capability has been added

to this implementation. In contrast with some real-

time multimedia applications, in a ﬁle transmission

application the packet loss probability must be zero

(i.e., the ﬁle must be fully received). In order to

achieve this with Online Codes, the decoding fail-

ure possibility must be eliminated. This is obtained

by trying to decode the original information for ev-

ery received block. When the process fails, the re-

ceiver waits for the next check block and it tries to de-

code it again. With this method the number of check

blocks that need to be decoded is a little bigger than

(1+3ε)K, but the decoding failure probability is zero.

However, our application does not really decode

the received blocks, it only deduces whether the re-

ceived packets are enough to decode the original ﬁle,

and it ﬁnishes when it has already received enough

packets in order to decode it. This program imple-

ments a “light decoding” process and it gives as a

result the previous amount of packets. The decod-

ing mechanism can be implemented by another con-

current program, or it may be performed when the

previous program has ﬁnished, that is, when it has

received the necessary blocks to decode the original

ﬁle. On the other hand, the packets can also be stored

in a coded way. This second option could be a good

choice if the coded packets are going to be retrans-

mited (e.g. in a P2P network) or if a OS implements

a facility that is able to recognize this kind of coded

ﬁles.

Both ﬁle transmission applications are evaluated

in a contention scenario, with other two simultaneous

ﬁle transmission connections. The size of the trans-

mitted ﬁle varies from 1 up to 20 Mbytes. In this

case, both applications are able to receive the full ﬁle

without errors. Therefore, in the evaluation we are

going to compare them taking into account the neces-

sary time to receive the full ﬁle. The distance between

the PLC devices of the measured connection is about

45 meters, like in the UDP scenario. Fig. 3 shows

the average duration of the scp and Online Codes ses-

sions, extracted from 5 different sessions. In addition,

the ﬁgure also represents the 95% conﬁdence inter-

val. The ﬁrst result that can be concluded from the

previous ﬁgure is that the Online Codes sessions are

always faster than the scp sessions. In addition, the

increase of the scp sessions with respect to the Online

Codes sessions is approximatelyconstant and equal to

1 second, though for big ﬁles this increment increases

up to 2 seconds. Therefore, we can draw that in a con-

tention scenario the performance achieved by an On-

line Codes-based ﬁle transmission application is bet-

ter than the performanceachieved by a TCP-based ﬁle

transmission application. Although not represented,

we have also tested the FTP protocol, which also uses

TCP as transport protocol. In this case, the sessions

duration are always higher than with scp, and there-

fore the difference between the Online Codes appli-

cation and FTP is also higher.

Next, both ﬁle transmission applications are eval-

uated again in a contention scenario, but in this case,

with four simultaneous ﬁle transmission connections.

Fig. 4 shows the average duration of the scp and On-

line Codes sessions, extracted again from 5 different

sessions, and the associated 95% conﬁdence interval.

In this scenario, the Online Codes sessions are also

faster than the scp sessions. However, in this case the

increase of the scp sessions with respect to the On-

line Codes sessions is higher than in the previous sce-

nario, approximately equal to 2 second, and for big

ﬁles this increment increases up to 4 seconds. On

the other hand, if we compare these results with those

presented in Fig. 3, we can observe that the increase

of simultaneous ﬁle transmission connections causes

an increase in the length of the sessions. This result is

absolutely obvious because in this case the broadcast

medium has to be shared by a bigger number of users.

As a conclusion, in multiple access networks (like

in-home PLC based networks), the access control

mechanism produces losses in unidirectional (like

UDP) transmissions. In these cases, it is necessary

to add some type of ﬂow control (e.g. using TCP

as transport protocol), or implementing some kind of

forward error correction, for example, the use of On-

line Codes. We have compared the session lengths

FOUNTAIN CODES FOR RELIABLE DATA TRANSMISSION IN LOWVOLTAGE POWER-LINE NETWORKS

139

0 2 4 6 8 10 12 14 16 18 20

Time [secs]

File Size [MB]

TCP

Online Codes

Figure 3: Duration of TCP (scp) and Online Codes sessions

in a noisy channel with two (background) data ﬂows sharing

the channel. The conﬁdence interval has been set to 95%.

0 2 4 6 8 10 12 14 16 18 20

Time [secs]

File Size [MB]

TCP

Online Codes

Figure 4: Duration of TCP (scp) and Online Codes sessions

in a noisy channel with four (background) data ﬂows shar-

ing the channel. The conﬁdence interval has been set to

95%.

of a TCP ﬁle transmission application and an Online

Codes based application, and we have extracted that

the Online Codes sessions are always faster than the

TCP sessions. Therefore, we have proved that in con-

tention scenarios the performance achieved by Online

Codes is better than the performance achieved by a

TCP-based ﬁle transmission application.

5 CONCLUSIONS

Online Codes must be mainly used when the net-

work or the application is unidirectional (broadcast-

TV, satellite, IP live-TV, etc.) or in applications that

cannot directly use TCP (like Application Layer Mul-

ticast). However, in a multiple access in-home net-

work (HPAV, wireless 802.11, PhonePNA, etc.) the

Online Codes are a good alternative to TCP for reli-

able data transmission.

ACKNOWLEDGEMENTS

This research has been supported by project grant

TEC2007-67966-C03-01/TCM(CON-PARTE-1) and

it is also developed in the framework of “Programa de

Ayudas a Grupos de Excelencia de la Regi´on de Mur-

cia, de la Fundaci´on S´eneca, Agencia de Ciencia y

Tecnolog´ıa de la RM (Plan Regional de Ciencia y Tec-

nolog´ıa 2007/2010)”. Pedro Jos´e Pi˜nero-Escuer also

thanks “Fundaci´on S´eneca” for a S´eneca Program FPI

pre-doctoral fellowship.

REFERENCES

Afkhamie, K. H., Katar, S., Yonge, L., and Newman, R.

(2005). An overview of the upcoming homeplug av

standard. In Proceedings of 2005 Internacional Sym-

posium on Power Line Communications and its Appli-

cations.

Cisco-Linksys (2009). Cisco-linksys. http://www.linksys

bycisco.com/.

Gigle (2009). Gigle semiconductor. http://www.gigle.biz/.

HomePlug Certiﬁed Products (2009). Homeplug certiﬁed

products. http://www.homeplug.org/kshowcase/view.

Implementation of Online Codes (2009). Implementation

of online codes. http://sourceforge.net/projects/online

codes/.

Jensen, B. and Kjarsgaard, S. (2007). Benchmarking and

qos of in-house powerline equipment under noisy con-

ditions. In Proceedings of 2007 Internacional Sympo-

sium on Power Line Communications and its Applica-

tions.

Luby, M. (2002). Lt codes. In Foundations of Computer

Science, 2002. Proceedings. The 43rd Annual IEEE

Symposium on, pages 271–280.

Maymounkov, P. and Mazi`eres, D. (2003). Rateless codes

and big downloads. In Peer-to-Peer Systems II, Sec-

ond International Workshop, IPTPS 2003, Berke-

ley, CA, USA, February 21-22,2003, Revised Papers,

pages 247–255.

scp (2009). scp-secure copy. http://www.mkssoftware.com/

docs/man1/scp.1.asp.

Shokrollahi, A. (2006). Raptor codes. IEEE/ACM Trans.

Netw., 14(SI):2551–2567.

ICSOFT 2009 - 4th International Conference on Software and Data Technologies

140